Non-Pharmacological Approaches for Migraine

Neurotherapeutics. 2018 Apr; 15(2): 336–345. Published online 2018 Apr 3. doi: 10.1007/s13311-018-0623-6 PMCID: PMC5935652 PMID: 29616493 Francesca Puleddacorresponding author1 and Kevin Shields2 1Department of Basic and Clinical Neuroscience, Institute of Psychiatry, Psychology and Neuroscience, King’s College London, London, UK 2Headache Service, The National Hospital for Neurology and Neurosurgery, Queen Square, London, UK Francesca Puledda, Phone: +44-203-299-6386, Email: ku.ca.lck@addelup.acsecnarf. corresponding authorCorresponding author. Author information ▼ Copyright and License information ► Disclaimer This article has been cited by other articles in PMC. Go to: Abstract Migraine is one of the most common and debilitating neurological disorders. However, the efficacy of pharmacological therapies may have unsatisfactory efficacy and can be poorly tolerated. There is a strong need in clinical practice for alternative approaches for both acute and preventive treatment. Occasionally, this need might arise in the context of low-frequency migraneurs who are not keen to use medication or fear the potential side effects. At the opposite end of the spectrum, clinicians might be faced with patients who have proven refractory to numerous medications. These patients may benefit from invasive treatment strategies. In recent years, promising strategies for migraine therapy have emerged alongside a progressively better understanding of the complex pathophysiology underlying this disease. This review discusses the most recent and evidence-based advances in non-pharmacological therapeutic approaches for migraine, offering alternatives to drug treatment for both the commonly encountered episodic cases as well as the more complex migraine phenotypes, which are capable of challenging even the headache specialist. Go to: Electronic supplementary material The online version of this article (10.1007/s13311-018-0623-6) contains supplementary material, which is available to authorized users. Keywords: Migraine, Therapy, Non-pharmacological, Neuromodulation, Nutraceuticals Go to: Introduction Migraine is one of the most common neurological diseases with a possible cumulative lifetime incidence of up to 50% in women and 20% in men [1]. Its high prevalence and frequency in young people at the peak of their productive years causes it to rank as the sixth cause of disability in the world [2]. The currently available oral pharmacological migraine treatments may be poorly tolerated by some patients. Unpleasant side effects and lower than hoped for efficacy [3, 4] can lead to low treatment compliance and other complications, such as headache chronification and acute medication overuse. However, recent years have seen a proliferation in new therapeutic strategies for tackling migraine, as well as novel techniques that offer promising areas of development. This review will focus on acute and preventive non-pharmacological approaches to migraine therapy. We will particularly concentrate on new strategies backed by robust evidence, prioritizing the attention of the reader on randomized controlled trials (RCTs), when these are available. The treatments that are analyzed in this article range from commonly used and easily accessible nutraceuticals, like magnesium and riboflavin, to well-established psychological therapies such as cognitive-behavioral therapy (CBT). Finally, we will focus on the relatively novel area of neuromodulation. With respect to the latter, we will analyze in detail the available non-invasive techniques (the most significant of which are summarized in Table ​Table1).1). These may potentially offer an alternative for episodic migraine patients who hope to avoid the side effects of pharmacological therapies. We will also briefly review the invasive strategies—represented by stimulation of the occipital nerve, the sphenopalatine ganglion, or the cervical spinal cord—which can be considered in chronic migraine patients who have repeatedly failed previous approaches. Table 1 Main randomized controlled trials for non-invasive neuromodulation in migraine An external file that holds a picture, illustration, etc. Object name is 13311_2018_623_Tab1_HTML.jpg Open in a separate window The importance in clinical practice of being able to offer patients who struggle with classic drug treatments relatively low risk and well-tolerated alternatives is obvious. Furthermore, it is of particular interest for the headache specialist to become familiar with the expanding world of neuromodulation, which aside from allowing strategies that bypass regular medication side effects, may offer insight into pathophysiological mechanisms of migraine and its underlying biology. Go to: Nutraceuticals Nutraceuticals are defined as food or dietary supplements that provide medicinal or health benefits. Their use is becoming increasingly popular in the general population. They have a particular appeal to patients with chronic diseases who hope to avoid the issues associated with long-term prescription treatments. In patients with migraine, the use of this kind of non-pharmacological therapy is growing and is likely to be widely underestimated [5]. The most commonly used nutraceuticals which have shown some evidence in migraine prevention are riboflavin (vitamin B2), coenzyme Q10 (CoQ10), magnesium, butterbur root extract (Petasites hybridus), and feverfew (Tanacetum parthenium). Riboflavin is a precursor of flavin mononucleotide and flavin adenine dinucleotide. These coenzymes are required for several energy-related cellular functions and electron transport within the Krebs cycle, and so play an important role in energy production inside the mitochondrion. The rationale for using riboflavin in migraine emanated from magnetic resonance spectroscopy studies [6–8]. These suggested that there may be mitochondrial dysfunction within the migranous brain. To date, five randomized controlled trials have assessed the efficacy of riboflavin in migraine, with somewhat conflicting results. The first randomized controlled trial assessing the use of riboflavin in migraine was performed in Belgium [9]. In this study, the effect of 400 mg daily dose was tested in 55 episodic adult migraineurs, with and without aura. Riboflavin showed a significant effect in reducing the number of headache days and attack frequency, with only minor and rare side effects when compared with placebo. However, a randomized control trial on 48 children with episodic migraine failed to show any difference between a 50% lower dose of 200 mg riboflavin and placebo in reducing the number of headache days, their duration, or the severity of migraine attacks and associated symptoms [10]. These results are in contrast with the study by Athaillah et al. [11] who performed a randomized controlled trial on 98 adolescents between the ages of 12 and 19. In this study, episodic migraineurs both with and without aura were randomly assigned to receive either placebo or riboflavin 400 mg. Treatment subjects in the active group showed a significant reduction in number of migraine days per month, starting from the second month of treatment. Two trials have also tested the effect of riboflavin in combination with other supplements in migraine, with opposite outcomes. In a study by Maizels et al. [12], subjects were randomized to receive either a combination of riboflavin 400 mg, magnesium 300 mg and feverfew 100 or 25 mg of riboflavin daily. This was done in order to achieve similar levels of chromaturia (a typical side effect of riboflavin) in both the active and “placebo” groups to avoid unblinding. Results showed no observed difference in the two groups with regard to reduction in migraine frequency. Given the high placebo response rate seen in the study, it has been speculated by the authors that 25 mg riboflavin might have actually acted as an active compound, therefore confounding comparison. It should also be noted that up to one third of patients in the study were taking concomitant preventive treatments for migraine. In 2015, a large German multicenter study by Gaul et al. [13] compared a compound containing riboflavin 400 mg, coenzyme Q10 150 mg, magnesium 600 mg and multivitamins with placebo in 130 episodic migraneurs. The active group failed to show a significant reduction in number of migraine days with respect to the placebo group; secondary outcome measures such as pain intensity and burden of disease were however significantly reduced. Methodological differences make it difficult to draw definitive conclusions regarding the efficacy of riboflavin. There does appear to be sufficient evidence to recommend that it be used as a well-tolerated, low risk preventive treatment for adults with migraine. Intriguingly there is some evidence to suggest that variations in mitochondrial DNA may influence the response to riboflavin [14]. This might help explain some of the observed variability in efficacy but regrettably it does not allow tailoring of treatment to likely responders. Similarly to riboflavin, coenzyme Q10 has an essential role in the mitochondrial electron transport chain and energy metabolism. In migraine, it has shown efficacy in one randomized controlled trial to date in which 42 migraine patients received either placebo or a 300-mg daily dose of CoQ10 [15]. The treatment group showed a more pronounced (p = 0.05) reduction in attack frequency from baseline to month 4 with respect to placebo, and the product was generally well tolerated. In another RCT on 50 migraine children and adolescents, CoQ10 was given at the dose of 100 mg per day and compared to placebo, showing no difference between the two groups [16]. More evidence will be needed therefore to support the widespread use of CoQ10 as a preventive treatment. Previous studies have highlighted a reduction of magnesium levels in migraine; these observations lead to the hypothesis of a consequent neuronal hyperexcitability in the migranous brain, given the role of this compound in inhibiting glutamate expression through NMDA receptor binding [17–21]. Oral magnesium has been systematically studied as a preventive treatment for migraine in five randomized controlled trials. The first study, performed by Facchinetti et al., assessed the effect of 360 mg oral magnesium in menstrual migraine compared to placebo [22] in 20 subjects. Both the treatment and placebo group had a significant decrease in migraine frequencies and pain scores with respect to baseline; therefore, magnesium did not prove more effective than placebo with regard to these outcome measures. However, the active group showed significantly reduced pain scores respect to placebo after treatment. Limitations of this study include its small case number and the fact that it was only conducted in women. In a multicenter German study carried out in 43 migraineurs with or without aura, a daily dose of 600 mg magnesium was compared to placebo [23]. The study was designed as a crossover trial in which the 2-month treatment was interchanged. Interestingly, a significant reduction in attack frequency was demonstrated only in the placebo-verum sequence and not in the verum-placebo one, possibly because of a carry-over effect of magnesium due to the absence of a wash-out period in between treatments [24]. Pfaffenrath and colleagues [25] compared magnesium 486 mg daily to placebo in a study involving 69 migraine without aura patients, finding no statistically significant difference in headache days. On the other hand, Peikert et al. [26] compared magnesium 600 mg to placebo in 68 total migraineurs and found significantly decreased attack frequency, duration and intensity in the active group. Finally, a study by Koseoglu on 40 migraineurs without aura showed a significant decrease in attack frequency in the group treated with 600 mg daily magnesium, although a reduction could also be seen in the placebo group [27]. This study however had a net disproportion between group sizes, with 30 patients receiving magnesium and 10 placebo. In conclusion, there is limited evidence to suggest that magnesium constitutes an effective preventive treatment for migraine patients. Petasites hybridus or butterbur root is a herbal extract that has shown some efficacy in migraine prevention. Its name derives from the leaves of the plant, which due to their size were originally used to wrap butter. The extract of the butterbur root which is used in tablet form is called Petadolex and is manufactured in Germany. There have been some safety concerns related to possible liver toxicity with the use of butterbur in recent years [28]. In total, two placebo-controlled trials have been performed in migraine to date. Lipton et al. [29] compared the efficacy of butterbur 50 and 75 mg twice daily with placebo. A significant reduction in migraine attack frequency was seen with the 75 mg dose, with no effect for the lower dose. Diener performed an analysis of a previous RCT [30], showing a significant decrease of migraine attack frequency after treatment with butterbur 100 mg total daily dose compared to placebo after 12 weeks of treatment. Butterbur therefore appears to be effective but safety concerns mean that it cannot currently be recommended as a preventive treatment. Tanacetum parthenium or feverfew has been studied in migraine prophylaxis, although results are still unconvincing and have low quality of evidence. One large RCT [31], conducted on 170 migraine patients, showed overall good tolerability and a reduction in migraine attacks with 6.25 mg of feverfew extract. A previous study using the same extract however had failed to show a significant effect in migraine prophylaxis [32]. Go to: Behavioral Techniques and Acupuncture Behavioral techniques comprise a series of strategies—relaxation, thermal and electromyographic biofeedback and cognitive behavioral therapy—which have been used in migraine therapy mostly with the aim of teaching patients to better cope with symptoms and identifying potential triggers for headache. Relaxation techniques include progressive muscle relaxation, autogenic training and meditation. Biofeedback training uses electronic devices to help the patient understand and monitor certain physiological processes associated with the experience of pain, such as muscle tension, blood pressure and heart rate changes. Cognitive behavioral therapy is a form of brief and symptom-oriented psychotherapy focused on managing stress. These techniques have some degree of evidence for their use in migraine [33, 34], particularly when there is low tolerance to medical strategies and in specific cases when medication is not indicated, such as pregnancy, medical comorbidities or evidence of previous medication overuse [35]. Additionally, they can be effective in conjunction with classic pharmacological therapies [36, 37] and allow a certain degree of self-management, which can be advantageous for some patients. However, one must consider that these strategies do not target the migraine biology or the actual pain mechanisms. A recent randomized trial [37] compared cognitive behavioral therapy plus amitriptyline to headache education plus amitriptyline in 135 children and adolescents with chronic migraine. The active CBT therapy group showed a significant reduction in headache days of 11.5 respect to the control group at 20 weeks. Another study on 61 transformed migraine and medication overuse headache patients showed that biofeedback-assisted relaxation combined with pharmacological therapy was more effective in reducing headache days and reduced consumption of analgesic than pharmacological therapy alone [38]; these positive results were confirmed by a more recent pilot randomized study on similar patient groups [39]. The use of acupuncture in migraine has yielded conflicting results. There have been a few RCTs [40] in which acupuncture is compared with either sham acupuncture or standard of care, showing some effect in headache improvement. Nonetheless blinding proves difficult in these studies and overall the strength of the evidence is low. One recent trial showed some minor effect over sham acupuncture [41] albeit with a high likelihood of unblinding, with others showing no difference [42]. The largest RCT investigating the effects of acupuncture in migraine was performed in Germany, on a total of 960 patients (n = 794 in the intention-to-treat population) [43]. This study compared the effect of verum acupuncture to sham acupuncture and standard therapy, showing that all three treatments were effective in reducing the number of migraine days respect to baseline, but that there was no significant difference in between the three groups. This allows to speculate that the biological effect of acupuncture might not depend on the positioning of the needles themselves. Go to: Non-Invasive Neuromodulation Non-invasive neuromodulation is a burgeoning field in migraine research and treatment. The classic techniques act by stimulating the nervous system centrally or at the periphery; this can be done through the skin either with and electric current or with a fluctuating magnetic field, ultimately modulating pain mechanisms involved in headache. Both modalities can have immediate effects, making them suitable for acute symptomatic treatment, while chronic administration may have longer-term preventive actions. These devices show potential as they offer an alternative to oral or invasive therapies while also having favorable side effect profiles; however, caution is required. As we shall see the randomized placebo-controlled trials published to date have tended to be small, and questions have been raised regarding the degree of blinding. It is therefore difficult to give definitive guidance currently on the efficacy of these treatments. Transcutaneous Cranial Nerve Stimulation Supraorbital nerve stimulation (STS) with the Cefaly® device (Cefaly Technology, Grâce-Hollogne Belgium) is a form of transcutaneous cranial nerve stimulation. It was first studied as a preventive treatment in a pilot study on episodic migraine, resulting in an average reduction of 1.3 headache days in ten participants [44]. This study was followed by an RCT on 67 migraine subjects. Sham stimulation in this trial was obtained with an identical looking and sounding machine. The impulse of the sham stimulation however was of 1 mA intensity and 1 Hz frequency (respectively 16 mA and 60 Hz for verum) and could not be perceived by the subjects. The results of the study showed that a once daily treatment session with Cefaly® for 3 months caused a significant 30% reduction in migraine days. This result was more evident in the active group, even though the difference between the two groups was just above statistical significance. The verum stimulation group also showed a greater 50% responder rate than the sham (38.2 vs 12.1%) with a 26% therapeutic gain [45]. A post marketing survey of 2313 patients who rented the Cefaly® device for 40 days showed that 54% of users thought that the device was beneficial and were satisfied with it [46]. The most frequently reported side effect was paresthesia in the area of stimulation. This side effect, albeit mild and fully reversible, can be intolerable in some and can lead to treatment interruption. It may also have caused a certain degree of unblinding during the randomized controlled trial, but it is still reasonable to consider Cefaly® as a preventive treatment. An RCT on the use of Cefaly® in the acute treatment of migraine (NCT02590939) and an open-label study in chronic migraine (NCT02342743) have recently been completed and results are awaited. The device is currently available for purchase in Europe, North America and Australia. Transcutaneous occipital nerve stimulation (tONS) has been trialed for migraine prevention in a recent randomized controlled trial [47]. A HANS TENS machine delivering three different frequencies (2 Hz, 100 Hz, 2/100 Hz) over the occipital area was compared with sham stimulation and 100 mg topiramate daily in 110 subjects. A significant decrease in headache frequencies was observed in the medication group and the 100 Hz stimulation group compared to the sham group, showing a potential for this new strategy in migraine. The Cefaly® device has also been recently used as a transcutaneous occipital nerve stimulator for the treatment of chronic migraine [48], with promising results. Twenty-three patients treated with the device over 3 months had a 17% decrease in headache days and 22% in migraine days; furthermore, abnormal VEP habituation reversed to an episodic migraine pattern in these subjects. Finally, one randomized controlled trial combined occipital and supraorbital transcutaneous nerve stimulation with an OSTNS Neurostimulator (NCT02438553) for acute migraine relief; results of this study are awaited. Non-Invasive Vagus Nerve Stimulation (nVNS) The gammaCore® device (electroCore, LLC; Basking Ridge, NJ, USA) is a handheld electrical stimulator that delivers a transcutaneous current to the cervical branch of the vagus nerve. Initially, vagus nerve stimulation (VNS) was an invasive procedure used in the treatment of refractory epilepsy and depression. Anecdotal observations of migraine improvement in patients being treated with implantable devices [49], spurred interest in using VNS to treat headache disorders. This leads to the development of the handheld non-invasive gammaCore® device. The first open-label single-arm study examining the efficacy of gammaCore® for the acute treatment of migraine was performed on 30 (27 in the full analysis set) USA patients, showing a pain free rate at 2 h of 21% [50]. Similar results were achieved in the open-label study by Barbanti and colleagues [51], which showed 2-h pain free rates of 23% in 48 high-frequency episodic and chronic migraine patients. The side effect profile in both studies showed a high tolerability for the device, with the most common adverse events being twitching and a tingling sensation at the stimulation site. Following these results, nVNS was studied initially with two open label studies as a preventive treatment for migraine. Magis et al. [52] performed a small study on 12 migraine subjects which showed a high improvement, particularly in one patient with medication overuse. However, a high percentage of subjects did stop treatment. More recently, one study in menstrual related migraine showed that a 12-week treatment period of non-invasive vagus nerve stimulation, started 3 days before the estimated onset of menses, was effective in reducing menstrually related migraine of 2.5 days per month, with more than one third of subjects showing a 50% reduction rate [53]. Unfortunately, the only randomized controlled study to test the efficacy of nVNS in migraine prevention to date—the EVENT study [54]—has failed to show a significant difference after 2 months of treatment in the active vs. sham group. Treatment was administered three times a day unilaterally with two 90 s doses in 59 chronic migraine patients. The trial consisted of a randomized phase lasting 2 months followed by an open label phase of 6 months. It is interesting to note, however, that at the end of the open label treatment, participants who had been assigned to nVNS during the randomized phase and therefore completed 8 months of treatment, had a significant reduction of almost eight headache days per month. nVNS therefore has shown some promise as an acute treatment but evidence is lacking to support its use as a preventive therapy. Currently two RCTs are studying gammaCore® both as an acute (NCT02686034) and preventive (NCT02378844) strategy in migraine and their results will be eagerly awaited. The Nemos® device (Cerbomed, Erlangen, Germany) is a recently developed transcutaneous stimulator of the aurical branch of the vagus nerve. The electrode is worn in the ear. In a German-based randomized controlled trial on 46 subjects, patients were randomly assigned to receive either 25 Hz or 1 Hz stimulation, the latter being intended as a sham stimulation [55]. Patients were asked to stimulate for a total of 4 h per day, in sessions of 1 to 4 h. Interestingly, patients in the “sham” treatment group had a more significant reduction in headache days than the ones in the “active” stimulation group. The response to sham stimulation may possibly indicate some biological activity but there is at present insufficient data to support its regular use in clinical practice. Treatment-related side effects in this study were mostly characterized by local pain, paresthesia and ulcers at the stimulation site. Single-Pulse Transcranial Magnetic Stimulation (sTMS) Single-pulse transcranial magnetic stimulation is a safe and non-invasive technique that has been used in the field of neuroscience for decades. It acts by creating a fluctuating magnetic field which in turn induces an electric current capable of modifying the excitability of cortical neurons and thalamocortical circuits. Animal studies have shown that sTMS is also capable of inhibiting cortical spreading depression, as well as the firing of nociceptive thalamic neurons projecting to the cortex [56], thus justifying the growing interest for this technique in migraine therapy. The first open-label study of sTMS in migraine was performed on 42 subjects who treated acute migraine attacks with brief pulses of TMS [57]. The encouraging efficacy and safety results of this study lead to the development of a large randomized control trial performed in the USA using a hand-held transcranial magnetic stimulator for the treatment of acute migraine with aura. A total of 164 patients treated at least one aura episode with two sTMS pulses or sham. The 2-h pain free rates were significantly higher in the group using sTMS (39%) compared to the sham group (22%) [58]. A UK open-label study [59] was performed to review the post-market use of SpringTMS ®device (eNeura, Baltimore, USA) in a large number of episodic and chronic migraineurs, both with and without aura. A total of 190 patients, using the device for acute and preventive purposes, reported an overall 60% pain relief rate as well as significant reduction of headache days after 3 months of continuous treatment. An economic comparison between sTMS and Botox® treatment in chronic migraine in the UK also suggests that sTMS offers a cost effective treatment for this group of patients [60]. The ESPOUSE trial (NCT02357381) is an ongoing, USA-based post-marketing study for sTMS treatment in the acute and preventive setting. Initial reports showed similar data to the UK post market survey, with a reduction of three and eight headache days per month, respectively, in episodic and chronic migraine patients. Transcranial Direct Current Stimulation (tDCS) Transcranial direct current stimulation may modulate cortical excitability by an anodal (excitatory) or cathodal (inhibitory) electric current applied to the scalp. This modifies the membrane potential of underlying cortical neurons. The effect of tDCS in migraine prevention has been tested using both anodal and cathodal currents. The two studies testing cathodal inhibitory currents to date have failed to show an effect of this technique in preventive migraine treatment. The first was a randomized controlled study on 26 migraine patients in which the active electrode was placed on the visual cortex. The study showed no significant reduction in migraine attacks and no difference between the two groups [61]. A smaller, more recent pilot study performed on 15 migraineurs similarly showed no significant difference in number of attacks, pain intensity and duration between sham stimulation and inhibitory tDCS over the area corresponding to the visual cortex [62]. Opposite outcomes seem to be achieved when applying an anodal activating current to the scalp with tDCS. In 2012, Auvichayapat et al. performed a randomized controlled trial on 37 episodic migraine patients in which a 20 min 1 mA anodal current over the primary motor cortex was compared with sham treatment [63]. Results showed a significant reduction in attack frequency and pain intensity in the active stimulation group. This result however was only seen at 4 and 8 weeks after treatment and not at 12 weeks, suggesting a possible short-term effect. Using a similar anatomical approach and a 2 mA current, DaSilva et al. also showed a significant reduction in headache intensity and a trend of reduced frequency in 13 patients with chronic migraine [64]. In a proof-of-concept open label study by Vigano’ et al, anodal current was applied over the visual cortex in 10 episodic migraine patients, with a significant reduction of 38% in migraine frequency respect to baseline [65]. An ongoing RCT will hopefully confirm these promising results for anodal tDCS applied overt the visual cortex (NCT02122757). Percutaneous Mastoid Stimulation The percutaneous mastoid stimulation (PMES) device administers an electric current through the skin behind the ear and acts by inducing fastigial nucleus stimulation in the cerebellum. Its neuroprotective effect has been previously trialed in stroke medicine [66] and more recently the device was used for migraine prevention in a randomized, double-blind, sham-controlled trial [67]. The study showed a significant reduction in migraine days (71.3%) in the treatment group, as well as a significantly higher 50% response rate, although the actual blinding degree in this study is difficult to assess. If repeated, this could be a promising new treatment modality but more evidence is needed. Non-Painful Brachial Electric Stimulation Yarnitsky et al. [68] studied the effects of electrical cutaneous stimulation on the arm as an acute treatment for migraine. Seventy-one patients treated at least one attack in this prospective, randomized, double-blind, sham controlled, crossover trial. Electrical stimulation was controlled by the patient’s smartphone and delivered by electrodes mounted on an armband. Various pulse widths were used but the best clinical effect was found with the 200us stimulus. A relatively modest therapeutic gain of 24% for pain freedom at 2 h was observed over sham stimulation. Active stimulation was rated as either painful or unpleasant by 39% of participants vs. 14% for sham. Compared to other stimulator devices this device has the added advantage of being discrete—it can potentially be hidden under clothing and controlled remotely. This result needs to be replicated but remote brachial stimulation does show promise as an acute treatment. Go to: Invasive Neuromodulation Occipital Nerve Stimulation (ONS) Dural and cervical sensory afferent neurons converge upon common second order neurons in the trigeminocervocal complex. Electrical stimulation of the greater occipital nerve may modulate central pain transmission [69, 70]. A total of three RCTs have been performed in chronic migraine. All studied ONS for a 12-week double blind phase, followed by open label phases. These lasted between 1 and 3 years depending on the study. The PRISM trial was published only in abstract form and failed to show a significant improvement in migraine frequency after ONS treatment [71]. The ONSTIM study, performed on 75 subjects, compared ONS to sham stimulation and medication management [72]. Results showed a higher responder rate of 39% (defined as the percentage of subjects who achieved a 50% or greater reduction in number of headache days per month or a three-point or greater reduction in average overall pain intensity compared with baseline) in the active adjustable stimulation group, compared to 6% of the preset stimulation group and 0% of the medically managed group. There was a high frequency of side effects, most commonly lead migration. Finally, a large RCT on 157 patients failed to show a significant difference in 50% responder rate (defined as a patient with a reduction from baseline of 50% or greater together with no increase in average headache duration) in the active ONS group, even though the 30% headache pain reduction was significantly higher with treatment [73]. At the moment, an ongoing randomized controlled trial aims at comparing the effects of ONS associated with medical treatment to sham stimulation in migraine prevention (OPTIMIZE trial - NCT01775735). Overall, what emerges from these trials is that the effect of ONS in migraine prevention is at best modest or not significant. It is possible to argue that for the most severe and refractory cases, even mild to modest reductions in headache frequency and pain rates might still prove clinically important. This may be especially so in the limited number of individuals for whom improvement lasts for years [74]. However adverse events—mostly pain, infections and lead migration—are quite common and this, coupled with the high cost, means that widespread use of the technique is not realistic. Sphenopalatine Ganglion Stimulation (SNS) The sphenopalatine ganglion (SPG) forms part of the parasympathetic outflow of the cranial trigeminal-autonomic reflex. Activation of this reflex arc is responsible for the autonomic manifestations seen in trigeminal autonomic cephalalgias [75] and it also plays a role in the pathophysiology of migraine [76]. Application of lidocaine to the SPG can abort an attack of migraine [77]. The effect of electrical SPG stimulation was therefore tested as acute migraine therapy in 11 patients [78]. Unfortunately, this study had unsatisfactory results but a randomized controlled trial is currently underway to evaluate the effect of an implanted SPG stimulator in chronic migraine (Pathway M-1 study - NCT01540799). High Cervical Spinal Cord Stimulation Cervical spine stimulation delivers an electrical current to the trigeminocervical complex through the use of implantable leads. This technique was initially trialed in refractory chronic migraine in a single center open label retrospective study on 17 patients [79]. Results showed a significant reduction in both pain intensity—of more than 50% in at least two thirds of patients—and median number of migraine days—from 28 to 9. Less than 20% of subjects had major adverse events, mostly lead migration and infection. More recently, 17 individuals with refractory chronic migraine and medication overuse took part in an open-label study evaluating the efficacy of high-frequency (10 Hz) stimulation of the cervical spinal cord [80]. Results were quite promising, with 8 of the 14 patients still implanted at 6 months experiencing a > 30% reduction in headache days and 6 a > 50% reduction; these rates are higher than what is usually achieved by sham stimulation in neuromodulation RCTs. An advantage with the high stimulation frequency of this particular device is that it avoids the paresthesias caused by other implantable neurostimulators. However randomized placebo controlled trials will be required before any firm recommendations can be made for this treatment. Go to: Conclusions The use of non-pharmacological treatments for migraine represents an expanding clinical practice and interesting area of research. Whenever confronted with headache patients with a complex disease or who are unwilling to accept possible drug-induced adverse events, clinicians should consider this rapidly growing armamentarium of treatment strategies and choose from the different options on the base of the desired clinical indications as well as availability. Non-invasive neuromodulation constitutes a valuable approach with strong backing evidence, particularly in the case of sTMS and transcutaneous cranial nerve stimulation. It is however available only in specialized centers, of several, but not most, countries. Even if blinding is extremely difficult to achieve in these studies, it will be interesting to observe the future developments of ongoing trials, which hopefully will confirm the initial positive results. Nutraceuticals, especially riboflavin and magnesium, represent a good option for patients with comorbidities and who are taking concomitant medications, or in patients who cannot tolerate drug side effects. They are also cheap and easily available. Behavioral treatment approaches can be considered as an add-on to ongoing treatments and often offer a positive clinical improvement. They are not as available as nutraceuticals, but they can be obtained in primary and secondary care settings. Implantable neuromodulation devices, on the other hand, should only be tentatively considered for the most serious and refractory patients who have failed multiple preventive attempts. Large clinical trials are especially needed in this field and will hopefully allow a better understanding of headache disorders, as well as a more individual-based approach to treatment in the future. Go to: Electronic supplementary material ESM 1(712K, pdf) (PDF 711 kb) Go to: Required Author Forms Disclosure forms provided by the authors are available with the online version of this article. Go to: Compliance with Ethical Standards Go to: Disclosure Kevin Shields is a paid scientific advisor to eNeura, Sunnyvale, California. Go to: Footnotes Electronic supplementary material The online version of this article (10.1007/s13311-018-0623-6) contains supplementary material, which is available to authorized users. Go to: References 1. Stewart WF, Wood C, Reed ML, Roy J, Lipton RB. Cumulative lifetime migraine incidence in women and men. Cephalalgia : an international journal of headache. 2008;28(11):1170–8. doi: 10.1111/j.1468-2982.2008.01666.x. [PubMed] [Cross Ref] 2. Global Burden of Disease Study 2013 Collaborators Global, regional, and national incidence, prevalence, and years lived with disability for 301 acute and chronic diseases and injuries in 188 countries, 1990-2013: a systematic analysis for the Global Burden of Disease Study 2013. Lancet (London, England) 2015;386(9995):743–800. doi: 10.1016/S0140-6736(15)60692-4. [PMC free article] [PubMed] [Cross Ref] 3. Hepp Z, Dodick DW, Varon SF, Gillard P, Hansen RN, Devine EB. Adherence to oral migraine-preventive medications among patients with chronic migraine. Cephalalgia : an international journal of headache. 2015;35(6):478–88. doi: 10.1177/0333102414547138. [PubMed] [Cross Ref] 4. Goadsby PJ, Sprenger T. Current practice and future directions in the prevention and acute management of migraine. The Lancet Neurology. 2010;9(3):285–98. doi: 10.1016/S1474-4422(10)70005-3. [PubMed] [Cross Ref] 5. Wells RE, Bertisch SM, Buettner C, Phillips RS, McCarthy EP. Complementary and alternative medicine use among adults with migraines/severe headaches. Headache. 2011;51(7):1087–97. doi: 10.1111/j.1526-4610.2011.01917.x. [PMC free article] [PubMed] [Cross Ref] 6. Welch KM, Levine SR, D'Andrea G, Schultz LR, Helpern JA. Preliminary observations on brain energy metabolism in migraine studied by in vivo phosphorus 31 NMR spectroscopy. Neurology. 1989;39(4):538–41. doi: 10.1212/WNL.39.4.538. [PubMed] [Cross Ref] 7. Barbiroli B, Montagna P, Cortelli P, Funicello R, Iotti S, Monari L, et al. Abnormal brain and muscle energy metabolism shown by 31P magnetic resonance spectroscopy in patients affected by migraine with aura. Neurology. 1992;42(6):1209. doi: 10.1212/WNL.42.6.1209. [PubMed] [Cross Ref] 8. Watanabe H, Kuwabara T, Ohkubo M, Tsuji S, Yuasa T. Elevation of cerebral lactate detected by localized 1H-magnetic resonance spectroscopy in migraine during the interictal period. Neurology. 1996;47(4):1093–5. doi: 10.1212/WNL.47.4.1093. [PubMed] [Cross Ref] 9. Schoenen J, Jacquy J, Lenaerts M. Effectiveness of high-dose riboflavin in migraine prophylaxis A randomized controlled trial. Neurology. 1998;50(2):466–70. doi: 10.1212/WNL.50.2.466. [PubMed] [Cross Ref] 10. MacLennan SC, Wade FM, Forrest KM, Ratanayake PD, Fagan E, Antony J. High-dose riboflavin for migraine prophylaxis in children: a double-blind, randomized, placebo-controlled trial. Journal of child neurology. 2008;23(11):1300–4. doi: 10.1177/0883073808318053. [PubMed] [Cross Ref] 11. Athaillah YD, Saing JH, Saing HB, Lelo A. Riboflavin as migraine prophylaxis in adolescents. Paediatr Indones. 2012;52:132–7. doi: 10.14238/pi52.3.2012.132-7. [Cross Ref] 12. Maizels M, Blumenfeld A, Burchette R. A combination of riboflavin, magnesium, and feverfew for migraine prophylaxis: a randomized trial. Headache. 2004;44(9):885–90. doi: 10.1111/j.1526-4610.2004.04170.x. [PubMed] [Cross Ref] 13. Gaul C, Diener HC, Danesch U. Improvement of migraine symptoms with a proprietary supplement containing riboflavin, magnesium and Q10: a randomized, placebo-controlled, double-blind, multicenter trial. J Headache Pain. 2015;16:516. doi: 10.1186/s10194-015-0516-6. [PMC free article] [PubMed] [Cross Ref] 14. Di Lorenzo C, Pierelli F, Coppola G, Grieco GS, Rengo C, Ciccolella M, et al. Mitochondrial DNA haplogroups influence the therapeutic response to riboflavin in migraineurs. Neurology. 2009;72(18):1588–94. doi: 10.1212/WNL.0b013e3181a41269. [PubMed] [Cross Ref] 15. Sandor PS, Di Clemente L, Coppola G, Saenger U, Fumal A, Magis D, et al. Efficacy of coenzyme Q10 in migraine prophylaxis: a randomized controlled trial. Neurology. 2005;64(4):713–5. doi: 10.1212/01.WNL.0000151975.03598.ED. [PubMed] [Cross Ref] 16. Slater SK, Nelson TD, Kabbouche MA, LeCates SL, Horn P, Segers A, et al. A randomized, double-blinded, placebo-controlled, crossover, add-on study of CoEnzyme Q10 in the prevention of pediatric and adolescent migraine. Cephalalgia : an international journal of headache. 2011;31(8):897–905. doi: 10.1177/0333102411406755. [PubMed] [Cross Ref] 17. Welch KM, Ramadan NM. Mitochondria, magnesium and migraine. Journal of the neurological sciences. 1995;134(1–2):9–14. doi: 10.1016/0022-510X(95)00196-1. [PubMed] [Cross Ref] 18. Gallai V, Sarchielli P, Coata G, Firenze C, Morucci P, Abbritti G. Serum and salivary magnesium levels in migraine. Results in a group of juvenile patients. Headache. 1992;32(3):132–5. doi: 10.1111/j.1526-4610.1992.hed3203132.x. [PubMed] [Cross Ref] 19. Gallai V, Sarchielli P, Morucci P, Abbritti G. Red blood cell magnesium levels in migraine patients. Cephalalgia : an international journal of headache. 1993;13(2):94–81. doi: 10.1046/j.1468-2982.1993.1302094.x. [PubMed] [Cross Ref] 20. Mauskop A, Altura BT, Altura BM. Serum ionized magnesium levels and serum ionized calcium/ionized magnesium ratios in women with menstrual migraine. Headache. 2002;42(4):242–8. doi: 10.1046/j.1526-4610.2002.02075.x. [PubMed] [Cross Ref] 21. Taylor FR. Nutraceuticals and headache: the biological basis. Headache. 2011;51(3):484–501. doi: 10.1111/j.1526-4610.2011.01847.x. [PubMed] [Cross Ref] 22. Facchinetti F, Sances G, Borella P, Genazzani AR, Nappi G. Magnesium prophylaxis of menstrual migraine: effects on intracellular magnesium. Headache. 1991;31(5):298–301. doi: 10.1111/j.1526-4610.1991.hed3105298.x. [PubMed] [Cross Ref] 23. Taubert K. Magnesium in migraine. Results of a multicenter pilot study. Fortschritte der Medizin. 1994;112(24):328–30. [PubMed] 24. von Luckner A, Riederer F. Magnesium in Migraine Prophylaxis-Is There an Evidence-Based Rationale? A Systematic Review. Headache. 2018;58(2):199–209. doi: 10.1111/head.13217. [PubMed] [Cross Ref] 25. Pfaffenrath V, Wessely P, Meyer C, Isler HR, Evers S, Grotemeyer KH, et al. Magnesium in the prophylaxis of migraine--a double-blind placebo-controlled study. Cephalalgia : an international journal of headache. 1996;16(6):436–40. doi: 10.1046/j.1468-2982.1996.1606436.x. [PubMed] [Cross Ref] 26. Peikert A, Wilimzig C, Kohne-Volland R. Prophylaxis of migraine with oral magnesium: results from a prospective, multi-center, placebo-controlled and double-blind randomized study. Cephalalgia : an international journal of headache. 1996;16(4):257–63. doi: 10.1046/j.1468-2982.1996.1604257.x. [PubMed] [Cross Ref] 27. Koseoglu E, Talaslioglu A, Gonul AS, Kula M. The effects of magnesium prophylaxis in migraine without aura. Magnesium research. 2008;21(2):101–8. [PubMed] 28. Rajapakse T, Pringsheim T. Nutraceuticals in Migraine: A Summary of Existing Guidelines for Use. Headache. 2016;56(4):808–16. doi: 10.1111/head.12789. [PubMed] [Cross Ref] 29. Lipton RB, Gobel H, Einhaupl KM, Wilks K, Mauskop A. Petasites hybridus root (butterbur) is an effective preventive treatment for migraine. Neurology. 2004;63(12):2240–4. doi: 10.1212/01.WNL.0000147290.68260.11. [PubMed] [Cross Ref] 30. Diener HC, Rahlfs VW, Danesch U. The first placebo-controlled trial of a special butterbur root extract for the prevention of migraine: reanalysis of efficacy criteria. European neurology. 2004;51(2):89–97. doi: 10.1159/000076535. [PubMed] [Cross Ref] 31. Diener HC, Pfaffenrath V, Schnitker J, Friede M, Efficacy H-v ZHH. safety of 6.25 mg t.i.d. feverfew CO2-extract (MIG-99) in migraine prevention--a randomized, double-blind, multicentre, placebo-controlled study. Cephalalgia : an international journal of headache. 2005;25(11):1031–41. doi: 10.1111/j.1468-2982.2005.00950.x. [PubMed] [Cross Ref] 32. Pfaffenrath V, Diener HC, Fischer M, Friede M, Henneicke-von Zepelin HH. The efficacy and safety of Tanacetum parthenium (feverfew) in migraine prophylaxis--a double-blind, multicentre, randomized placebo-controlled dose-response study. Cephalalgia : an international journal of headache. 2002;22(7):523–32. doi: 10.1046/j.1468-2982.2002.00396.x. [PubMed] [Cross Ref] 33. Goslin RE, Gray RN, McCrory DC, Penzien D, Rains J. Hasselblad. Agency for Health Care Policy and Research (US: AHRQ Technical Reviews. Behavioral and Physical Treatments for Migraine Headache. Rockville (MD; 1999. 34. Penzien DB, Irby MB, Smitherman TA, Rains JC, Houle TT. Well-Established and Empirically Supported Behavioral Treatments for Migraine. Current pain and headache reports. 2015;19(7):34. doi: 10.1007/s11916-015-0500-5. [PubMed] [Cross Ref] 35. Silberstein SD. Practice parameter: evidence-based guidelines for migraine headache (an evidence-based review): report of the Quality Standards Subcommittee of the American Academy of Neurology. Neurology. 2000;55(6):754–62. doi: 10.1212/WNL.55.6.754. [PubMed] [Cross Ref] 36. Holroyd KA, Cottrell CK, O'Donnell FJ, Cordingley GE, Drew JB, Carlson BW, et al. Effect of preventive (beta blocker) treatment, behavioural migraine management, or their combination on outcomes of optimised acute treatment in frequent migraine: randomised controlled trial. BMJ (Clinical research ed). 2010;c4871:341. [PMC free article] [PubMed] 37. Powers SW, Kashikar-Zuck SM, Allen JR, et al. Cognitive behavioral therapy plus amitriptyline for chronic migraine in children and adolescents: A randomized clinical trial. Jama. 2013;310(24):2622–30. doi: 10.1001/jama.2013.282533. [PMC free article] [PubMed] [Cross Ref] 38. Grazzi L, Andrasik F, D'Amico D, Leone M, Usai S, Kass SJ, et al. Behavioral and pharmacologic treatment of transformed migraine with analgesic overuse: outcome at 3 years. Headache. 2002;42(6):483–90. doi: 10.1046/j.1526-4610.2002.02123.x. [PubMed] [Cross Ref] 39. Rausa M, Palomba D, Cevoli S, Lazzerini L, Sancisi E, Cortelli P, et al. Biofeedback in the prophylactic treatment of medication overuse headache: a pilot randomized controlled trial. J Headache Pain. 2016;17(1):87. doi: 10.1186/s10194-016-0679-9. [PMC free article] [PubMed] [Cross Ref] 40. Linde K, Allais G, Brinkhaus B, Fei Y, Mehring M, Vertosick EA, et al. Acupuncture for the prevention of episodic migraine. The Cochrane database of systematic reviews. 2016(6):Cd001218. [PMC free article] [PubMed] 41. Zhao L, Chen J, Li Y, et al. The long-term effect of acupuncture for migraine prophylaxis: A randomized clinical trial. JAMA Internal Medicine. 2017. [PubMed] 42. Linde K, Streng A, Jurgens S, Hoppe A, Brinkhaus B, Witt C, et al. Acupuncture for patients with migraine: a randomized controlled trial. Jama. 2005;293(17):2118–25. doi: 10.1001/jama.293.17.2118. [PubMed] [Cross Ref] 43. Diener HC, Kronfeld K, Boewing G, Lungenhausen M, Maier C, Molsberger A, et al. Efficacy of acupuncture for the prophylaxis of migraine: a multicentre randomised controlled clinical trial. The Lancet Neurology. 2006;5(4):310–6. doi: 10.1016/S1474-4422(06)70382-9. [PubMed] [Cross Ref] 44. Gérardy PY, Fabry D, Fumal A, Schoenen J. A pilot study on supra-orbital surface electrotherapy in migraine. Cephalalgia : an international journal of headache. 2009;29:134. 45. Schoenen J, Vandersmissen B, Jeangette S, Herroelen L, Vandenheede M, Gerard P, et al. Migraine prevention with a supraorbital transcutaneous stimulator: a randomized controlled trial. Neurology. 2013;80(8):697–704. doi: 10.1212/WNL.0b013e3182825055. [PubMed] [Cross Ref] 46. Magis D, Sava S, d'Elia TS, Baschi R, Schoenen J. Safety and patients' satisfaction of transcutaneous supraorbital neurostimulation (tSNS) with the Cefaly(R) device in headache treatment: a survey of 2,313 headache sufferers in the general population. J Headache Pain. 2013;14:95. doi: 10.1186/1129-2377-14-95. [PMC free article] [PubMed] [Cross Ref] 47. Liu Y, Dong Z, Wang R, Ao R, Han X, Tang W, et al. Migraine prevention using different frequencies of transcutaneous occipital nerve stimulation: A randomized controlled trial. The journal of pain : official journal of the American Pain Society. 2017. [PubMed] 48. Schoenen J, D'Ostilio K, Cosseddu A, Nonis R, Sava S, Magis D. Transcranial Direct Current Stimulation and Transcutaneous Occipital Nerve Stimulation in Chronic Migraine: A Pilot-Comparison of Therapeutic and Electrophysiological Effects. Neurology. 2016;86(16 Supplement P2.200). 49. Sadler RM, Purdy RA, Rahey S. Vagal nerve stimulation aborts migraine in patient with intractable epilepsy. Cephalalgia : an international journal of headache. 2002;22(6):482–4. doi: 10.1046/j.1468-2982.2002.00387.x. [PubMed] [Cross Ref] 50. Goadsby PJ, Grosberg BM, Mauskop A, Cady R, Simmons KA. Effect of noninvasive vagus nerve stimulation on acute migraine: an open-label pilot study. Cephalalgia : an international journal of headache. 2014;34(12):986–93. doi: 10.1177/0333102414524494. [PubMed] [Cross Ref] 51. Barbanti P, Grazzi L, Egeo G, Padovan AM, Liebler E, Bussone G. Non-invasive vagus nerve stimulation for acute treatment of high-frequency and chronic migraine: an open-label study. J Headache Pain. 2015;16:61. doi: 10.1186/s10194-015-0542-4. [PMC free article] [PubMed] [Cross Ref] 52. Magis D, Gerard P, Schoenen J. Transcutaneous Vagus Nerve Stimulation (tVNS) for headache prophylaxis: initial experience. J Headache Pain. 2013;14(Suppl 1):198. doi: 10.1186/1129-2377-14-S1-P198. [Cross Ref] 53. Grazzi L, Egeo G, Calhoun AH, McClure CK, Liebler E, Barbanti P. Non-invasive Vagus Nerve Stimulation (nVNS) as mini-prophylaxis for menstrual/menstrually related migraine: an open-label study. J Headache Pain. 2016;17(1). [PMC free article] [PubMed] 54. Silberstein SD, Calhoun AH, Lipton RB, Grosberg BM, Cady RK, Dorlas S, et al. Chronic migraine headache prevention with noninvasive vagus nerve stimulation: The EVENT study. Neurology. 2016;87(5):529–38. doi: 10.1212/WNL.0000000000002918. [PMC free article] [PubMed] [Cross Ref] 55. Straube A, Ellrich J, Eren O, Blum B, Ruscheweyh R. Treatment of chronic migraine with transcutaneous stimulation of the auricular branch of the vagal nerve (auricular t-VNS): a randomized, monocentric clinical trial. J Headache Pain. 2015;16:543. doi: 10.1186/s10194-015-0543-3. [PMC free article] [PubMed] [Cross Ref] 56. Andreou AP, Holland PR, Akerman S, Summ O, Fredrick J, Goadsby PJ. Transcranial magnetic stimulation and potential cortical and trigeminothalamic mechanisms in migraine. Brain : a journal of neurology. 2016;139(Pt 7):2002–14. doi: 10.1093/brain/aww118. [PMC free article] [PubMed] [Cross Ref] 57. Clarke BM, Upton AR, Kamath MV, Al-Harbi T, Castellanos CM. Transcranial magnetic stimulation for migraine: clinical effects. J Headache Pain. 2006;7(5):341–6. doi: 10.1007/s10194-006-0329-8. [PMC free article] [PubMed] [Cross Ref] 58. Lipton RB, Dodick DW, Silberstein SD, Saper JR, Aurora SK, Pearlman SH, et al. Single-pulse transcranial magnetic stimulation for acute treatment of migraine with aura: a randomised, double-blind, parallel-group, sham-controlled trial. The Lancet Neurology. 2010;9(4):373–80. doi: 10.1016/S1474-4422(10)70054-5. [PubMed] [Cross Ref] 59. Bhola R, Kinsella E, Giffin N, Lipscombe S, Ahmed F, Weatherall M, et al. Single-pulse transcranial magnetic stimulation (sTMS) for the acute treatment of migraine: evaluation of outcome data for the UK post market pilot program. J Headache Pain. 2015;16:535. doi: 10.1186/s10194-015-0535-3. [PMC free article] [PubMed] [Cross Ref] 60. Bruggenjurgen B, Baker T, Bhogal R, Ahmed F. Cost impact of a non-invasive, portable device for patient self-administration of chronic migraine in a UK National Health Service setting. SpringerPlus. 2016;5(1):1249. doi: 10.1186/s40064-016-2924-8. [PMC free article] [PubMed] [Cross Ref] 61. Antal A, Kriener N, Lang N, Boros K, Paulus W. Cathodal transcranial direct current stimulation of the visual cortex in the prophylactic treatment of migraine. Cephalalgia : an international journal of headache. 2011;31(7):820–8. doi: 10.1177/0333102411399349. [PubMed] [Cross Ref] 62. Rocha S, Melo L, Boudoux C, Foerster A, Araujo D, Monte-Silva K. Transcranial direct current stimulation in the prophylactic treatment of migraine based on interictal visual cortex excitability abnormalities: A pilot randomized controlled trial. Journal of the neurological sciences. 2015;349(1–2):33–9. doi: 10.1016/j.jns.2014.12.018. [PubMed] [Cross Ref] 63. Auvichayapat P, Janyacharoen T, Rotenberg A, Tiamkao S, Krisanaprakornkit T, Sinawat S, et al. Migraine prophylaxis by anodal transcranial direct current stimulation, a randomized, placebo-controlled trial. Journal of the Medical Association of Thailand = Chotmaihet thangphaet. 2012;95(8):1003–12. [PubMed] 64. Dasilva AF, Mendonca ME, Zaghi S, Lopes M, Dossantos MF, Spierings EL, et al. tDCS-induced analgesia and electrical fields in pain-related neural networks in chronic migraine. Headache. 2012;52(8):1283–95. doi: 10.1111/j.1526-4610.2012.02141.x. [PMC free article] [PubMed] [Cross Ref] 65. Vigano A, D'Elia TS, Sava SL, Auve M, De Pasqua V, Colosimo A, et al. Transcranial Direct Current Stimulation (tDCS) of the visual cortex: a proof-of-concept study based on interictal electrophysiological abnormalities in migraine. J Headache Pain. 2013;14:23. doi: 10.1186/1129-2377-14-23. [PMC free article] [PubMed] [Cross Ref] 66. Wang J, Dong WW, Zhang WH, Zheng J, Wang X. Electrical stimulation of cerebellar fastigial nucleus: mechanism of neuroprotection and prospects for clinical application against cerebral ischemia. CNS neuroscience & therapeutics. 2014;20(8):710–6. doi: 10.1111/cns.12288. [PubMed] [Cross Ref] 67. Juan Y, Shu O, Jinhe L, Na Y, Yushuang D, Weiwei D, et al. Migraine prevention with percutaneous mastoid electrical stimulator: A randomized double-blind controlled trial. Cephalalgia : an international journal of headache. 2016. [PubMed] 68. Yarnitsky D, Volokh L, Ironi A, Weller B, Shor M, Shifrin A, et al. Nonpainful remote electrical stimulation alleviates episodic migraine pain. Neurology. 2017;88(13):1250–5. doi: 10.1212/WNL.0000000000003760. [PubMed] [Cross Ref] 69. Bartsch T, Goadsby PJ. The trigeminocervical complex and migraine: current concepts and synthesis. Current pain and headache reports. 2003;7(5):371–6. doi: 10.1007/s11916-003-0036-y. [PubMed] [Cross Ref] 70. Bartsch T, Goadsby PJ. Stimulation of the greater occipital nerve induces increased central excitability of dural afferent input. Brain : a journal of neurology. 2002;125(Pt 7):1496–509. doi: 10.1093/brain/awf166. [PubMed] [Cross Ref] 71. Lipton R, Goadsby PJ, Cady R, Aurora SK, Grosberg BM, Freitag F. PRISM study: occipital nerve stimulation for treatment-refractory migraine [abstract PO47]. Cephalalgia : an international journal of headache. 2009;29. 72. Saper JR, Dodick DW, Silberstein SD, McCarville S, Sun M, Goadsby PJ. Occipital nerve stimulation for the treatment of intractable chronic migraine headache: ONSTIM feasibility study. Cephalalgia : an international journal of headache. 2011;31(3):271–85. doi: 10.1177/0333102410381142. [PMC free article] [PubMed] [Cross Ref] 73. Silberstein SD, Dodick DW, Saper J, Huh B, Slavin KV, Sharan A, et al. Safety and efficacy of peripheral nerve stimulation of the occipital nerves for the management of chronic migraine: results from a randomized, multicenter, double-blinded, controlled study. Cephalalgia : an international journal of headache. 2012;32(16):1165–79. doi: 10.1177/0333102412462642. [PubMed] [Cross Ref] 74. Chen YF, Bramley G, Unwin G, Hanu-Cernat D, Dretzke J, Moore D, et al. Occipital Nerve Stimulation for Chronic Migraine—A Systematic Review and Meta-Analysis. PloS one. 2015;10(3). [PMC free article] [PubMed] 75. Goadsby PJ, Lipton RBA. review of paroxysmal hemicranias, SUNCT syndrome and other short-lasting headaches with autonomic feature, including new cases. Brain : a journal of neurology. 1997;120(Pt 1):193–209. doi: 10.1093/brain/120.1.193. [PubMed] [Cross Ref] 76. Goadsby PJ, Lipton RB, Ferrari MD. Migraine--current understanding and treatment. The New England journal of medicine. 2002;346(4):257–70. doi: 10.1056/NEJMra010917. [PubMed] [Cross Ref] 77. Maizels M, Geiger AM. Intranasal lidocaine for migraine: a randomized trial and open-label follow-up. Headache. 1999;39(8):543–51. doi: 10.1046/j.1526-4610.1999.3908543.x. [PubMed] [Cross Ref] 78. Tepper SJ, Rezai A, Narouze S, Steiner C, Mohajer P, Ansarinia M. Acute treatment of intractable migraine with sphenopalatine ganglion electrical stimulation. Headache. 2009;49(7):983–9. doi: 10.1111/j.1526-4610.2009.01451.x. [PubMed] [Cross Ref] 79. De Agostino R, Federspiel B, Cesnulis E, Sandor PS. High-cervical spinal cord stimulation for medically intractable chronic migraine. Neuromodulation : journal of the International Neuromodulation Society. 2015;18(4):289–96; discussion 96. [PubMed] 80. Arcioni R, Palmisani S, Mercieri M, Vano V, Tigano S, Smith T, et al. Cervical 10 kHz spinal cord stimulation in the management of chronic, medically refractory migraine: A prospective, open-label, exploratory study. European journal of pain (London, England). 2016;20(1):70–8. doi: 10.1002/ejp.692. [PubMed] [Cross Ref] Articles from Neurotherapeutics are provided here courtesy of Springer

Diverse effects of degree of urbanisation and forest size on species richness and functional diversity of plants, and ground surface-active ants and spiders

PLoS One. 2018; 13(6): e0199245. Published online 2018 Jun 19. doi: 10.1371/journal.pone.0199245 PMCID: PMC6007905 PMID: 29920553 Ramona Laila Melliger, Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Visualization, Writing – original draft, Brigitte Braschler, Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Visualization, Writing – original draft,* Hans-Peter Rusterholz, Conceptualization, Formal analysis, Methodology, Supervision, Visualization, Writing – review & editing, and Bruno Baur, Conceptualization, Funding acquisition, Methodology, Supervision, Writing – review & editing Petr Heneberg, Editor Section of Conservation Biology, Department of Environmental Sciences, University of Basel, Basel, Switzerland Charles University, CZECH REPUBLIC Competing Interests: The authors have declared that no competing interests exist. * E-mail: hc.sabinu@relhcsarb.ettigirb Author information ▼ Article notes ► Copyright and License information ► Disclaimer Go to: Abstract Urbanisation is increasing worldwide and is regarded a major driver of environmental change altering local species assemblages in urban green areas. Forests are one of the most frequent habitat types in urban landscapes harbouring many native species and providing important ecosystem services. By using a multi-taxa approach covering a range of trophic ranks, we examined the influence of degree of urbanisation and forest size on the species richness and functional diversity of plants, and ground surface-active ants and spiders. We conducted field surveys in twenty-six forests in the urban region of Basel, Switzerland. We found that a species’ response to urbanisation varied depending on trophic rank, habitat specificity and the diversity indices used. In plants, species richness decreased with degree of urbanisation, whereas that of both arthropod groups was not affected. However, ants and spiders at higher trophic rank showed greater shifts in species composition with increasing degree of urbanisation, and the percentage of forest specialists in both arthropod groups increased with forest size. Local abiotic site characteristics were also crucial for plant species diversity and species composition, while the structural diversity of both leaf litter and vegetation was important for the diversity of ants and spiders. Our results highlight that even small urban forests can harbour a considerable biodiversity including habitat specialists. Nonetheless, urbanisation directly and indirectly caused major shifts in species composition. Therefore, special consideration needs to be given to vulnerable species, including those with special habitat requirements. Locally adapted management practices could be a step forward to enhance habitat quality in a way to maximize diversity of forest species and thus ensure forest ecosystem functioning; albeit large-scale factors also remain important. Go to: Introduction Urbanisation is increasing globally and is considered a main driver of environmental change [1]. Urbanisation-related factors including reduced habitat size and increased spatial isolation change the dynamics of plant and animal populations in urban green areas [2, 3]. Several studies along urbanisation gradients also reported alterations in abiotic conditions in the remaining habitat patches caused by increases in temperature, precipitation and N deposition from the rural surroundings to the city centre [1, 4, 5]. These changes influence habitat quality and, consequently, the species richness, species composition and functional diversity of plants and animals [3, 6, 7], which in turn affect the functioning of ecosystems [8]. Furthermore, urbanisation can influence the population dynamics of animals and plants by altering the biology of hosts, pathogens and vectors [9]. Although urbanisation frequently reduces the abundance of many parasites and pathogens [10], transmission may also increase among urban-adapted hosts [9]. In some cases, invertebrates can serve as vectors of pathogens which otherwise are absent from urban environments [11]. Finally, plants and animals may be exposed to other chemicals (herbicides, fungicides, pesticides) and other types of pollution in urban environments than in rural agricultural landscapes [9]. Nonetheless, urban areas can harbour remarkably high species richness [12], in some cases exceeding that of their rural surroundings [6, 13]. Forests represent one of the most frequent types of green area in cities [14]. Urban forests provide a wide range of ecosystem functions including habitat for native species and recreation for residents [15, 16]. Both forests and orchards in cities can serve as refugia for rare and threatened specialist species and thus can be of high conservation value [12, 17]. Within urban landscapes, forest sites differ substantially in site history, management and disturbance intensity and consequently in species composition [3, 4, 17, 18]. Urban forests can be remnants of former continuous forests, a result of ongoing succession or actively planted [4]. They can include urban orchards, cemeteries overgrown by trees, or parks [4, 17, 18]. Not all species respond to environmental changes caused by urbanisation in the same way, because they have different requirements regarding their habitat and its surrounding landscape [7, 12, 19]. For example, specialist species may perceive the surrounding matrix as a stronger barrier than generalists, which are able to exploit a wide variety of resources from neighbouring green areas [1, 16, 20]. Thus, specialist species become frequently replaced by generalists [21, 22]. As a result, species composition in urban areas becomes more and more similar, which in turn may lead to a decrease in functional diversity–also called functional homogenisation ([20] and references within). Furthermore, groups of species at high trophic ranks such as herbivores and predators might also be more influenced by increased isolation and habitat loss because of their dependence on other species compared to groups of species at low trophic ranks such as plants [23, 24]. The majority of urban forest studies focused on a single taxonomic group, frequently plants, butterflies, carabids or birds (e.g. [22, 25, 26] and reviews of [27, 28]) or higher taxon or morphospecies levels [29, 30]. So far, few studies have examined the impact of urbanisation on the species diversity and/or functional diversity in forests using a multi-taxa approach. These studies often investigated either taxonomic groups at similar trophic ranks like carabids, rove beetles and spiders [31, 32, 33] or carrion-burying beetles, their phoretic mites, and muscoid flies [34] or focused on species with mutualistic or exploitative relationships [34, 35] or with similar life-history traits [20]. Most multi-taxa urban studies were conducted in openland habitats [19, 29, 35] or over a variety of habitat types [7]. To our knowledge, no studies were conducted in different-sized urban forests and considered species groups with different trophic ranks. In this study, we examined the impact of degree of urbanisation and forest size on the species and functional diversity of vascular plants and ground surface-active ants and spiders in forest sites in the city of Basel (Switzerland) and its suburban surroundings. The forests examined in our study are very small and embedded in a small-scattered landscape, where settlements and green areas are located within short distances. A rural–urban gradient approach extending over several kilometres is, therefore, not appropriate in our study area. Instead, we used the percentage cover of sealed area in the closer surroundings of the forests as a measure of degree of urbanisation as suggested by others (e.g. [31, 36]). The taxonomic groups considered in our study vary in trophic rank and thus in the use of resources available in the urban landscape. Ants are intermediate between the other two groups, as many ant species not only consume animal matter but also some plant material such as nectar or elaiosomes attached to seeds. Many species indirectly consume plant sap as excretion from sucking insects. In contrast, spiders are predators. Neither of the two arthropod groups depends on specific plant species as a resource. Hence, their responses to urbanisation can be expected to be independent of that of plants. In particular, we hypothesize that the diversity of plants, ants, and spiders (species richness, Shannon diversity and evenness and functional diversity) decrease with both increasing degree of urbanisation and decreasing forest size. These effects will be more pronounced for ants and spiders, because of their higher trophic rank, and for forest specialists due to their narrow habitat range. We further expect that small forests show lower species diversity and thus altered functional diversity and harbour lower percentages of forest specialists in highly compared to less urbanised forest areas. In contrast, the diversity in large forest sites should be less negatively affected by degree of urbanisation. Secondly, we hypothesize that species composition of plants, ants, and spiders will be altered by the degree of urbanisation and forest size. We expect that species composition in highly urbanised areas will be more similar than in less urbanised areas. Go to: Methods Study area The study was conducted in the canton Basel-Stadt (comprising the city of Basel and the municipalities Riehen and Bettingen; hereafter referred to as Basel, Fig 1), Switzerland (47°34’N, 7°36’E, elevation: 245–522 m a.s.l.). The study area covers 37 km2, consisting of 26.3 km2 (70.9%) residential area, 4.5 km2 (12.1%) agricultural land, 4.4 km2 (11.7%) forest and 1.7 km2 (4.5%) water bodies (Statistisches Amt Kanton Basel-Stadt: www.statistik-bs.ch). Basel has 196,471 inhabitants and a population density of 5320 inhabitants km-2 (www.statistik-bs.ch). Total annual precipitation averages 842 mm and annual mean temperature is 10.5°C (records from 1981 to 2010, www.meteoswiss.admin.ch). Most study sites were state owned and accessible to the public. Some forest was privately owned but managed by the forestry authorities. Permission for fieldwork was obtained from landowners, managers, and the authority responsible for the forests (Amt für Wald beider Basel). An external file that holds a picture, illustration, etc. Object name is pone.0199245.g001.jpg Open in a separate window Fig 1 Location of the study area in Northwestern Switzerland and the distribution of the forests examined in the area of Basel-Stadt. The investigation area is surrounded by dense settlements in Germany (north), France (northwest) and Switzerland (south-west). Characteristics of the forests To investigate the potential effects of degree of urbanisation and forest area on the species diversity of vascular plants, and soil surface-active ants and spiders, we chose 26 deciduous forests, belonging to the Fagetum association [37] and ranging in size from 258 m2 to 50,000 m2 (Fig 1; S1 Table; S2 Table). The forest sites examined differ in their historical development and consequently in age. Twenty of them are surrounded by settlements and agricultural lands and are no longer connected to large continuous forests (> 40 ha). These forest patches are either remnants of former large continuous forests (fragments) or a result of abandonment of orchards or planted after 1884 (planted; see S1 Table and S2 Table for detailed description of forests). For each of these twenty forests, we calculated the shape index following Gyenizse et al. [38]. A shape index of 1 corresponds to a circular area, which is considered as most stable and resistant against biotic and abiotic effects from the surrounding landscape [38]. Vegetation survey In each forest, we installed six sampling plots measuring 4 m × 4 m. Plots had a minimum distance of 1 m to the forest edge or permanent trails to minimize potential edge effects. We assessed species richness of vascular plants in the ground vegetation (≤ 40 cm) and cover of single species in a 2 m × 2 m subplot established in a randomly chosen corner of each 4 m × 4 m plot using the Braun-Blanquet scale [39]. To complete the plant species list in the entire sampling plot, additional species found in the other three 2 m × 2 m subplots were recorded. Ant and spider sampling We conducted pitfall trapping to sample ground surface-active ants and spiders. We installed a trapping grid in each of the forests examined. We arranged twelve pitfall traps (plastic cups: 5.8 cm diameter; fluid: 60 ml of water–detergent solution) in two rows with six traps each in a trap-grid system. The distance of the traps between and within the rows was 5 m. The size of the pitfall grid was determined by the smallest fragment, which was thus comprehensively sampled. A dummy of a grid of corresponding size was placed with closed eyes on a map showing forest cover and paths for the larger fragments thus avoiding prior knowledge of vegetation cover or topography when selecting the location for the pitfall trap grids. If necessary the grid was moved to be entirely within the forested area. To account for seasonal differences in activity among species, we operated pitfall traps once in spring, three times consecutively in summer, and once in autumn 2014. Traps were exposed for 7 days before being collected, which resulted in a maximum of 60 trap weeks per forest site (12 traps × 5 sampling weeks). We transferred trap contents to 70% ethanol for further processing. We identified individuals to the species level following the keys of Seifert [40] and Ward et al. ([41]; Colobopsis truncata) for ants and Roberts [42, 43] and Nentwig et al. ([44], , version 03.2017) for spiders. In ants, the winged reproductive castes (queens and males) were not considered in the analyses because in contrast to workers it is not clear whether they originated in the study site (123 of 16,465 individuals; 0.75%). We also excluded workers, which were too damaged to allow for species identification (0.13%). Three strictly arboreal species were likewise excluded, as they cannot be recorded in a representative way using pitfall traps. However, arboreal ants, which also have ground surface-activity, were retained [45]. Likewise, we excluded juvenile stages (1211 of 5327 individuals; 22.7%) and adult individuals of spiders, whose identification features (palpal bulbs, epigyne) were missing or destroyed (254 of 5327 individuals; 4.8%), from analyses. Trait data We assigned each plant species in one of the following two groups: forest species and non-forest species according to Delarze et al. [46]. For ant species, information on habitat specificity (forest specialist, generalist and open-land species) was designated from Seifert [40] and for spider species from Hänggi et al. ([47]; S3 Table). We called spider species forest specialists when they occur in deciduous forests. Edge species were excluded from this group. For each taxonomic group, we selected a set of species traits, which we considered to influence species’ response to urbanisation-related factors (Table 1 and S3 Table). Table 1 Species traits of plants, ants, and spiders. Trait Type Description Plants Life form1 Categorical Macrophanerophyte; nanophanerophyte; chamaephyte; hemicryptophyte; geophyte; therophyte Reproduction type1 Categorical Sexual; mixed Ecological strategy1 Categorical Following Grime (1979): C; CR; CS; CSR; S; SR Pollination syndrome1 Categorical Insects; wind Seed dispersal type2 Categorical Zoochory; anemochory; hemerochory; autochory; hydrochory Seed mass1 Continuous Mean of seed mass (mg) Ants Body size3, 4 Continuous Maximum of the total length of workers (mm) Main nest stratum4 Categorical Wood or litter; soil or crevices; both Number of queens4 Categorical Monogynous; oligogynous; polygynous Main food type4 Categorical Animal matter; animal matter and carbohydrates; carbohydrates; grains Spiders Body size5 Continuous Mean body size (mm) weighted by the proportion of males and females recorded in this study Hunting mode6 Categorical Web building; hunting (including active hunting and ambush) Open in a separate window Source 1 [37] 2 [38] 3 species descriptions in the taxonomic literature, sources listed under 4, and own measurements 4 [29] and three web-based resources (www.antwiki.org, www.ameisenwiki.de, www.antweb.org) 5 [33] 6 [39] Data of six plant traits (life form, reproduction type, ecological strategy following Grime [48], pollination syndrome, seed dispersal type and seed mass) were obtained from the database BiolFlor [49] and Müller-Schneider [50]. We obtained trait information for ants (body size of workers, main nest stratum, queen number, main food source) from Seifert [40], three web-based resources (, , ) and in a few cases from own measurements or taxonomic species descriptions (Table 1 and S3 Table). For spider species, we assembled data of body size, and hunting mode from literature [44, 51] (Table 1 and S3 Table). Environmental characteristics We estimated total cover of ground vegetation in each of the plots from the vegetation survey using the Braun-Blanquet scale [39]. Canopy closure was assessed based on three photographs in each plot and determined with the pixel counting function of Adobe Photoshop (version 10.0.1). To examine any potential influences of soil characteristics on plant diversity, three soil samples were collected in each vegetation plot using a metal cylinder (depth: 5 cm; diameter 5.05 cm; volume 100 cm3) in October 2014. We pooled and mixed the three soil samples of a plot and transported them to the laboratory, where they were sieved (mesh size 2 mm) and dried at 50°C for 96 h. We determined soil moisture content (%) using the fresh to dry weight ratio and assessed soil pH in distilled water (1:2.5 soil:water) [52]. We determined total soil organic matter content (SOM, %) as loss-on-ignition of oven dried soil at 750°C for 16 h [52]. We assessed total soil organic nitrogen content (orgN, %) using the standard method of Kjeldahl [53]. Finally, we determined total phosphorus content of soil (orgP, μg PO43– g–1) using the molybdenum blue method [54]. We measured biotic and abiotic environmental characteristics in the pitfall trap plots during the autumn pitfall trap survey. To assess the complexity of the vegetation structure and the amount of dead woody debris, we used a slight modification of the point intercept method [55]. In each grid of traps, we installed a transect line in the centre of the two rows. At the beginning of the transect line, we inserted a pin vertically into the ground and recorded the number of times the pin was touched by different plant specimens up to 2 m (hereafter referred to ‘vegetation structure’) and by dead woody debris on the forest floor (hereafter referred to ‘amount of dead wood’). We repeated this procedure at intervals of 1 m resulting in a total of 26 measuring points per forest site. To assess soil and litter characteristics, we divided the trap-grid system into three sections with each including four traps. In each grid section, we collected four soil samples. We pooled and mixed them to yield a total of three soil samples per trapping grid. In the centre of each grid section, leaf litter was collected in an area measuring 20 cm × 20 cm, dried and weighed. To assess the moisture content and pH of soil and litter and soil organic matter content, we applied the same methods as described above. Environmental factors were used to characterize the forest sites and to explain the patterns of diversity of the focal groups rather than to examine their own response to urbanisation and forest size. We assessed soil and litter variables, vegetation structure and amount of dead wood in autumn 2014. This is adequate for soil variables because soil pH, SOM, total soil organic nitrogen and total phosphorus content are relatively constant over the whole vegetation period in the forests examined [56]. For leaf litter the autumn sampling captures the year’s input. In addition to humidity also temperature can affect biodiversity or arthropod activity. We therefore measured soil temperature close to the surface (0–5 cm) hourly at the edge of the pitfall grid throughout the study period. As the study focused on the ground-surface active ants and spiders, soil surface temperature was considered to be the most appropriate measure for temperature, and air temperature higher up in the vegetation, where some species also forage can be expected to be correlated. However, due to high degrees of vandalism the temperature data were incomplete and could not be used in the models. A finer-scaled soil temperature survey conducted in nine of the forest sites, however, revealed only relatively small differences among the forests [56]. Landscape characteristics and recreational pressure For each forest, we derived land cover data of six landscape characteristics from satellite images (Google Earth, 2009). Around the most central sampling plot in each forest, we determined the percentage cover of built-up area and traffic infrastructure, urban green space (comprising gardens, parks and allotments), agricultural land and forest cover within radii of 200 m and 500 m using the pixel counting function of Adobe Photoshop (version 10.0.1). The percentage cover of sealed area (built-up area and traffic infrastructure) was used as a measure of the degree of urbanisation. Because the percentage cover of sealed area inter-correlated with the percentage covers of the other three landscape elements (all P < 0.008, S4 Table), we did not consider the percentage covers of these landscape elements for data analyses. We used two different measures to estimate the impact of recreational pressure in the forest sites: (1) path density expressed as the total length of paths and forestry trails per forest site (in m/ha), and (2) the total trampled area within a forest (expressed in percentage of forest area). Data analyses Statistical analyses were performed in R ver. 3.0.2 (www.r-project.org) and were carried out separately for the three taxonomic groups at the forest site level. Species richness consists of the total number of species recorded in all vegetation plots and pitfall traps, respectively, over the whole sampling period. In plants, we calculated Shannon diversity and evenness for each of the six vegetation plots separately and averaged them per forest site. In the ant and spider sampling, most of the forest sites were exposed to a variety of disturbances including vandalism, which caused the loss of several traps (72 out of 1560 traps, 4.6%; S5 Table). Therefore, we calculated sample-based rarefied species richness using the specaccum function in the package vegan in R. Due to positive correlations between observed and sample-based rarefied species richness (both, ants and spiders: rs = 1.00, n = 26, P < 0.001), we only used rarefied species richness in the subsequent analyses (hereafter referred to as ‘species richness’). For ants, where numbers can be inflated when a trap is close to a nest, we used the proportion of traps in which a species was present to calculate Shannon diversity and evenness instead of abundance data. We further used number of individuals per trap (individual density) instead of abundance data to compare Shannon diversity and evenness among forest sites for spiders. Preliminary analyses revealed correlations between the two radii of degree of urbanisation and the two measures of recreational pressure. In the vegetation plots, SOM further was positively correlated with soil orgN, while there were inter-correlations between soil and litter characteristics in the trap-grid system (soil moisture vs. litter moisture: r = 0.52, n = 26, P = 0.006; soil pH vs. litter pH: rs = 0.56, n = 26, P = 0.003). Therefore, we only considered degree of urbanisation within the 500-m radius, path density and soil orgN in plants and litter moisture content and litter pH in ants and spiders in the subsequent analyses. Furthermore, the historical development of forests was confounded with forest size (see S4 Table for further details). Forest size thus could not be considered independently from the historical development of the forests. Based on the percentage cover of sealed area in their surroundings, we classified the forests into areas with low (< 15%), medium (15–30%) or high (> 30%) degrees of urbanisation. We also divided forests into three size classes: small (< 4000 m2), medium-sized (4000–10,000 m2) or large (> 10,000 m2) forests (S2 Table). While these size classes also capture variations in forest history, for simplicity, we refer to these categories as forest size throughout the results section. The three size and three urbanisation classes were based on the distribution of available fragment sizes and percentages of sealed area following [56]. We considered the degree of urbanisation and forest size either as continuous variables (first approach) or as factors (second approach) in the statistical analyses to examine their potential effects on species diversity (species richness, Shannon diversity and evenness). However, because the two approaches revealed very similar results, we only present the results of the second approach. We applied generalized linear models (GLM) with quasi-Poisson distributed errors using log-link function to examine potential effects of the degree of urbanisation, forest size and the corresponding interaction on species diversity and the percentages of forest specialists, and ANCOVA for the functional dispersion of the three taxonomic groups. We used degree of urbanisation (three classes), forest size (three classes) and shape (three classes: continuous forests (no shape index), shape index 1–1.5, shape index > 1.5) and management of forest sites (‘time since last thinning’: ≤ 3 years, 4–10 years or > 10 years ago) as factors, and path density and canopy closure as cofactors in the GLM and ANCOVA models of all three taxonomic groups. In plants, we further included soil moisture content, soil pH, soil orgN and orgP and cover of ground vegetation as cofactors in the GLM and ANCOVA models (S2 Table). For ants and spiders, we used SOM, litter moisture content and litter pH, amount of litter biomass, vegetation structure and amount of dead wood as cofactors in the GLM and ANCOVA models (S2 Table). In ants and spiders, we further tested the impact of these factors on the percentages of generalist species. All the environmental factors listed above were included into models as covariables. The models were then reduced following a stepwise procedure, which resulted in the dropping of several covariables. We performed multiple comparisons (Tukey contrasts) to compare differences among degrees of urbanisation, forest size, forest shape and time since last thinning, respectively, using the glht function in the multcomp package in R [57]. To show whether degree of urbanisation affected species composition of plants, ants and spiders, we used non-metric multidimensional scaling (NMDS) with Bray-Curtis dissimilarity measures. Data were square-root-transformed and Wisconsin double standardization was applied. This type of transformation involves standardization of species maxima, followed by relativization of sample total [58] Species, which were recorded in only one site, were excluded from the analyses. Permutational multivariate analyses of variance (PERMANOVA) were used to test whether degree of urbanisation, forest size and local forest characteristics affected species composition of plants, ants and spiders [59]. The local forest characteristics were included as cofactors (S6 Table). For plants, soil moisture, soil pH, total soil phosphorous content, total soil nitrogen content, and ground vegetation cover were thus included in the analysis. For ants and spiders, path density, canopy closure, total soil organic matter content, litter moisture, litter pH, litter biomass, amount of dead wood, and vegetation structure were included as cofactors. For all three groups of organisms, we further included the shape index and the time since last thinning as factors. All PERMANOVA tests were based on 999 permutations of the untransformed raw data, using the adonis function in the vegan R-package [60]. As a measure for functional diversity we calculated functional dispersion for each taxonomic group according to Villéger et al. [61] using the dbFD function with Cailliez corrected distance matrices in the package FD in R [62]. As for the NMDS and PERMANOVA analyses, we only used those species that occurred in more than one forest site. We used ANCOVA to examine the impact of degree of urbanisation, forest size and local forest characteristics on the functional dispersion of plants, ants, and spiders. Go to: Results Across the 26 forests, we recorded a total of 130 vascular plant species (30.7 species per forest, range: 17–53 species; S5 Table). Eighty-three of the 130 plant species (63.8%) were forest specialists. The most common plant species in the ground vegetation layer were Hedera helix and Quercus robur, which occurred in 26 forests and Fraxinus excelsior and Geum urbanum, which were found in 25 forests. Overall, we collected 16,321 ants belonging to 28 species in the 26 forests examined. On average, we captured 10.0 ant species (range: 6–16 species) per forest (S5 Table). Among ant species, 10 were forest specialists or dependent on wood for their nest construction (35.7% of species found), while the reminder were habitat generalists (5 species; 17.9%) or even open-land species (13 species; 46.4%). Myrmica rubra, a generalist species, which is often found in urban habitats, comprised 41.7% of all ants collected. It occurred in 19 of the 26 sites, with 75.3% of individuals collected in a particular site. Six ant species were more widespread: Myrmecina graminicola and Temnothorax nylanderi (26 forests each), Lasius niger and Stenamma debile (23 each), Lasius brunneus (22) and Myrmica ruginodis (20). We collected 5,327 spiders belonging to 109 species. On average, 18.3 spider species (range: 10–31 species) were captured per forest (S5 Table). In spiders, 30 species were forest specialists (27.5%), 57 habitat generalists (52.3%) and 21 open-land species (19.3%). The most common spider species were Tenuiphantes flavipes (26 forests), Trochosa terricola (21), Diplostyla concolor and Pardosa saltans (19 each). Effects of degree of urbanisation on species diversity Plant species richness, the percentage of forest specialists and Shannon diversity of plants were affected by the degree of urbanisation (Table 2; Fig 2). While the species richness and Shannon diversity of plants decreased with increasing degree of urbanisation (Fig 2A and 2C), the percentage of forest specialists was slightly higher in forests located in areas with either a low or high degree of urbanisation compared to forests situated in areas with a medium degree of urbanisation (Fig 2B). Furthermore, Shannon evenness of plants tended to decrease in forests with increasing percentage cover of sealed areas in their surroundings (Fig 2D). An external file that holds a picture, illustration, etc. Object name is pone.0199245.g002.jpg Open in a separate window Fig 2 Plant species richness (a; mean ± SE), percentage of forest specialists (b), Shannon diversity (c) and evenness (d) in forests, which were located in areas with different degrees of urbanisation. Table 2 Summary of GLM analyses examining the effects of degree of urbanisation, forest size and shape, forest management (time since last thinning), disturbance (indicated by path density), canopy closure, soil characteristics (moisture, pH, soil orgN and orgP) and cover of ground vegetation on the species richness, percentage of forest specialists, Shannon diversity and evenness of vascular plants. Species richness1 Percentage of forest specialists Shannon diversity Shannon evenness df F P df F P df F P df F P Degree of urbanisation 2,23 8.43 0.004 2,23 4.59 0.029 2,23 7.71 0.004 2,23 3.44 0.056 Forest size 2,21 0.04 0.96 2,21 2.07 0.16 2,21 1.78 0.20 2,21 4.88 0.021 Shape index 2,19 1.43 0.27 2,19 18.59 <0.001 – – – – – – Time since last thinning – – – 2,17 1.47 0.26 2,19 2.64 0.10 2,19 5.57 0.014 Path density – – – – – – – – – – – – Canopy closure 1,18 7.84 0.015 1,16 9.53 0.008 – – – – – – Soil moisture content – – – 1,15 2.91 0.11 1,18 7.83 0.012 1,18 9.85 0.006 Soil pH – – – 1,14 2.63 0.13 – – – – – – Soil orgN1 – – – – – – 1,17 1.23 0.28 1,17 1.50 0.24 Soil orgP1 1,17 3.47 0.085 – – – – – – – – – Cover of ground vegetation – – – – – – † † † † † † Degree of urbanisation*forest size 4,13 1.41 0.28 – – – – – – – – – Open in a separate window Significant P-values (< 0.05) are in bold 1 log-transformed –Factor was excluded from the model by step-wise reduction † Factor was not included in the model In ants, the percentage of generalists was influenced by the degree of urbanisation, being slightly higher in forests with dense settlements in their surroundings than in forests located in areas with low or medium degrees of urbanisation (Table 3). In contrast, species richness, Shannon diversity and evenness of ants were not affected by the degree of urbanisation (Table 3). Table 3 Summary of GLM analyses examining the effects of degree of urbanisation, forest size and shape, forest management (time since last thinning), disturbance (indicated by path density), canopy closure, soil organic matter content, litter characteristics (moisture, pH) and structural diversity measures (litter biomass, vegetation structure and amount of dead wood) on the species richness, percentages of forest specialists and generalists, Shannon diversity and evenness of ants. Sample-based rarefied species richness Percentage of forest specialists Percentage of habitat generalists Shannon diversity Shannon evenness df F P df F P df F P df F P Df F P Degree of urbanisation 2,23 2.94 0.083 2,23 1.84 0.21 2,23 4.31 0.049 2,23 3.21 0.06 2,23 0.33 0.72 Forest size 2,21 2.71 0.10 2,21 6.09 0.018 2,21 0.71 0.51 2,21 4.71 0.023 2,21 1.57 0.15 Shape index – – – 2,19 1.20 0.34 2,19 2.56 0.14 – – – – – – Time since last thinning – – – 2,17 1.83 0.21 2,17 2.59 0.13 – – – – – – Path density 1,20 6.68 0.022 – – – – – – 1,20 7.45 0.014 – – – Canopy closure – – – 1,16 5.29 0.044 – – – – – – 1,20 1.57 0.23 Soil organic matter content1 – – – – – – 1,16 5.79 0.040 – – – – – – Litter moisture content – – – 1,15 2.46 0.15 1,15 6.31 0.033 – – – – – – Litter pH 1,19 7.12 0.018 – – – – – – 1,19 5.38 0.032 – – – Amount of litter biomass1 – – – – – – – – – – – – – – – Vegetation structure1 – – – 1,14 1.58 0.24 1,14 5.65 0.041 – – – – – – Amount of dead wood 1,18 1.36 0.13 – – – 1,13 1.61 0.24 1,18 6.47 0.020 – – – Degree of urbanisation*forest size 4,14 1.09 0.40 4,10 2.55 0.10 4,9 1.93 0.19 – – – – – – Open in a separate window Significant P-values (< 0.05) are in bold 1 log-transformed –Factor/Co-factor was excluded due to by step-wise model reduction procedure In spiders, both the percentages of forest specialists and generalists were influenced by the degree of urbanisation (Table 4). The percentage of forest specialists was lower in forests situated in areas with medium degree of urbanisation than in forests located in areas with high or low degree of urbanisation. In contrasts, the percentage of generalists was higher in forests located in areas with medium and high degrees of urbanisation than in forests surrounded by sparse settlements. However, the species richness, Shannon diversity and evenness of spiders did not differ among the urbanisation classes. We found an interaction between degree of urbanisation and forest size for the Shannon evenness of spiders: Small forests located in areas with a low degree of urbanisation had lower Shannon evenness indices than small forests located in areas with medium and high degrees of urbanisation, whereas medium-sized and large forests showed similar Shannon evenness indices in areas with different degrees of urbanisation. Table 4 Summary of GLM analyses examining the effects of degree of urbanisation, forest size and shape, forest management (time since last thinning), disturbance (indicated by path density), canopy closure, soil organic matter content, litter characteristics (moisture, pH) and structural diversity measures (litter biomass, vegetation structure and amount of dead wood) on the species richness, percentages of forest specialists and generalists, Shannon diversity and evenness of spiders. Sample-based rarefied species richness Percentage of forest specialists Percentage of habitat generalists Shannon diversity Shannon evenness df F P df F P df F P df F P df F P Degree of urbanisation 2,23 0.60 0.56 2,23 9.30 0.002 2,23 4.48 0.028 2,23 1.36 0.30 2,23 1.67 0.23 Forest size 2,21 0.02 0.98 2,21 3.96 0.039 2,21 0.29 0.75 2,21 1.90 0.20 2,21 3.77 0.051 Shape index – – – – – – – – – 2,19 1.33 0.31 2,19 3.38 0.066 Time since last thinning – – – – – – – – – 2,17 1.57 0.26 – – – Path density – – – 1,20 2.54 0.13 1,20 4.36 0.053 – – – 1,18 1.94 0.19 Canopy closure 1,20 4.33 0.052 – – – – – – 1,16 1.88 0.20 – – – Soil organic matter content1 – – – 1,19 2.67 0.12 1,19 1.78 0.20 – – – – – – Litter moisture content 1,19 5.29 0.034 – – – – – – 1,15 2.17 0.17 – – – Litter pH 1,18 1.23 0.28 1,18 7.20 0.016 1,18 2.64 0.12 – – – – – – Amount of litter biomass1 – – – 1,17 4.41 0.051 1,17 1.77 0.20 – – – – – – Vegetation structure1 – – – – – – 1,16 3.68 0.073 1,14 2.06 0.18 1,17 1.77 0.21 Amount of dead wood – – – – – – – – – – – – – – – Degree of urbanisation*forest size – – – – – – – – – 4,10 1.51 0.27 4,13 5.20 0.010 Open in a separate window Significant P-values (< 0.05) are in bold 1 log-transformed –Factor was excluded from the model by step-wise reduction Effects of the size and shape of forests on species diversity Shannon evenness of plants slightly increased with forest size, but was not affected by the shape of the forests (Table 2). In contrast, the percentage of forest specialists was influenced by the shape of forests (Table 2; S1 Fig), but did not differ among size classes (Table 2). A higher percentage of forest specialists was found in large continuous forests and forests with a shape index between 1.0 and 1.5 than in forests with a shape index higher than 1.5 (S1 Fig). The species richness and Shannon diversity of plants were neither influenced by the size nor shape of forests (Table 2). In ants, the percentage of forest specialists and Shannon diversity of ants were positively related to forest size (Table 3; S1 Fig), but were not influenced by the shape of forests. The species richness of ants and percentage of generalists of ants were neither influenced by the size nor the shape of forests (Table 3). Similar to ants, the percentage of spider forest specialists was higher in large than in medium-sized and small forests (Table 4; S1 Fig), but was not influenced by forest shape. Shannon evenness tended to be affected by forest size, being slightly higher in medium-sized than in small forests. Furthermore, Shannon evenness tended to be influenced by the shape index of forests. Large continuous forests and forests with a shape index between 1.0 and 1.5 exhibited a more even spider species distribution than forests with a shape index larger than 1.5. However, species richness, percentage of generalists and Shannon diversity of spiders were neither affected by the size nor shape of the forests (Table 4). Effects of forest site characteristics on species diversity measures Plant species richness decreased with increasing canopy closure of forests (rs = –0.61, n = 26, P < 0.001), while the percentage of forest specialists increased (rs = 0.47, n = 26, P = 0.017; Table 2). Both Shannon diversity and evenness of plants were positively related to soil moisture content (diversity: rs = 0.52, n = 26, P = 0.007; evenness: r = 0.42, n = 26, P = 0.031). Furthermore, Shannon evenness of plants was affected by the time since last thinning (Table 2). It was higher in forests, which were managed recently (≤ 3 years or 4–10 years) than in forests, which were thinned last time more than 10 years ago. However, path density, soil pH, soil orgN and orgP and the cover of ground vegetation did not influence any of the plant diversity measures examined (Table 2). In ants, species richness was negatively affected by litter pH (rs = –0.53, n = 26, P = 0.005) and tended to increase with path density (r = 0.38, n = 26, P = 0.058; Table 3). The percentage of forest specialists was influenced by canopy closure, being highest at moderate structural diversity of vegetation. The percentage of generalist ant species was positively affected by soil organic matter, litter pH, and vegetation structure (Table 3). However, the Spearman correlations for these covariables were not significant (all P > 0.2). Shannon diversity of ants tended to be positively affected by path density (rs = 0.37, n = 26, P = 0.062), and negatively by litter pH (rs = –0.35, n = 26, P = 0.077) and amount of dead wood (Table 3). However, the Spearman correlation for the latter was not significant. Shannon evenness of ants was not affected by any of the forest characteristics. The species richness of spiders decreased with litter moisture content (r = –0.41, n = 26, P = 0.038) and tended to be affected by canopy closure (Table 4). However, the latter was not a linear relationship. The percentage of forest specialists was influenced by litter pH with species richness highest at intermediate values of pH (Table 4). However, none of the forest characteristics examined had a significant impact on the percentage of generalist species, and Shannon diversity and evenness of spiders (Table 4). Species composition For plants, multivariate analysis using NMDS showed that plant species composition shifted from low to high degrees of urbanisation but with some overlap (Fig 3A). PERMANOVA confirmed that plant species composition was significantly affected by forest size (F2,19 = 2.42, P = 0.005). However, only a marginal tendency was found for degree of urbanisation (F2,19 = 1.47, P = 0.099). Plant species composition was also significantly affected by soil moisture (F1,19 = 2.82, P = 0.014) and total soil organic nitrogen content (F1,19 = 3.59, P = 0.001). Some common species showed marked differences in their frequencies depending on the degree of urbanisation or forest size (S7 Table). For example, the frequency of Arum maculatum and Duchesnea indica decreased with increasing degrees of urbanisation, while Alliaria petiolata was most frequent at intermediate degrees of urbanisation, and Tilia platyphyllos was most frequent in sites with high degrees of urbanisation. An external file that holds a picture, illustration, etc. Object name is pone.0199245.g003.jpg Open in a separate window Fig 3 NMDS of (a) plant, (b) ant and (c) spider species composition. Forests sites are grouped according to their degree of urbanisation (low, medium, high). Similar to the findings for plant species composition, ant species composition showed a shift from areas with a low degree of urbanisation to those with a high degree of urbanisation, though with some overlap (Fig 3B). Moreover, PERMANOVA showed that ant species composition was significantly affected by the degree of urbanisation (F2,15 = 1.86, P = 0.045). Ant species composition was also affected by forest size (F2,15 = 2.79, P = 0.005). Furthermore, canopy cover was also significantly affecting ant species composition (F2,15 = 2.30, P = 0.035). While many common species were similarly often present in sites with different degrees of urbanisation or of different size, some showed marked differences (S7 Table). For example, the generalist species Myrmica rubra occurred in all sites with high degrees of urbanisation, but only in three quarters of sites with a medium degree of urbanisation, and in just over half of sites with a low degree of urbanisation. Similarly, for spiders, there was a shift in species composition from highly to less urbanised areas (F2,17 = 2.63, P = 0.001; Fig 3C). Spider species composition was also affected by forest size (F2,17 = 1.62, P = 0.032). Furthermore, spider species composition was influenced by most forest characteristics examined: litter moisture content (F1,17 = 1.86, P = 0.028), SOM (F1,17 = 1.73, P = 0.049), vegetation structure (F1,17 = 1.72, P = 0.047), amount of dead wood (F1,17 = 1.98, P = 0.015). As for plants and ants, some spider species showed marked differences in their frequency of occurrence in forests with different degrees of urbanisation and of different size (S7 Table). Examples include Histopona torpida and Palliduphantes pallidus, which decreased in frequency with increasing degrees of urbanization, and Diplostyla concolor, which was most frequent in highly urbanized sites. Functional dispersion Plant functional dispersion was affected by the degree of urbanisation (F2,16 = 3.92, P = 0.041) and forest size (F2,16 = 3.68, P = 0.049; S8 Table). Furthermore, plant functional dispersion was influenced by the time since last thinning (S8 Table). Considering ants, functional dispersion tended to be influenced by forest size (F2,13 = 3.68, P = 0.054) (S9 Table). Furthermore, ant functional dispersion was significantly affected by litter moisture (F2,13 = 12.63, P = 0.004; S9 Table). In contrast spider functional dispersion was not significantly influenced by degree of urbanisation, forest size, or habitat characteristics (S9 Table). Go to: Discussion Our results showed that the response to the degree of urbanisation and forest size considerably varied among the three taxonomic groups. However, when we grouped species according to their habitat specificity, we observed a reduction in the percentage of forest specialist species with decreasing forest size in both arthropod groups. In addition to degree of urbanisation and forest size, species diversity and species composition of plants were determined by abiotic site characteristics and those of ants and spiders by the structural diversity of both leaf litter and vegetation. Effect of urbanisation on species diversity During the last decades, the worldwide urban sprawl and the subsequent destruction and isolation of green areas represent major drivers for local species extinction [3]. Hence, we expected a decrease in species diversity (species richness, Shannon diversity and evenness) with increasing degree of urbanisation. However, we only found this to be the case in plants. Cameron et al. [63] reported similar results for plant species richness, but did not find any effect on plant diversity. In contrast, McKinney [28] found the highest number of plant species in areas with medium degree of urbanisation, whereas Vallet et al. [25] did not detect any difference in total species richness of plants between urban and rural woodlands. These outcomes may be due to differences in the number of non-native plant species, the spatial dimension of the study areas and the degree of urbanisation associated with differences in habitat diversity ([28] and references therein). In our study sites we only found very few neophytes. In ground surface-active ants and spiders, the lack of response of species diversity to urbanisation contrasted our hypothesis and the findings of other studies conducted on soil arthropods in forests, which showed either a negative (carabids: [16]), hump-shaped (spiders: [32], carabids: [64]) or positive response (spiders: [65, 66]) on species richness in relation to the degree of urbanisation. However, similar results as in our study were reported by Alaruikka et al. [64], who argued that spiders might be more affected by local site characteristics (e.g. structural diversity) than by characteristics at the landscape scale. The higher sensitivity of plant species richness to degree of urbanisation compared to those of higher trophic rank ants and spiders did not confirm our hypothesis and contrasted findings of several multi-taxa studies (e.g. [28, 35, 67]). Comparisons with these studies, however, should be made with caution, as most of them were conducted in different habitat types and/or considered other taxonomic groups [16, 20, 35]. The taxonomic groups considered in those studies are also often closely related by showing specialised plant-herbivore interactions (e.g. [35]). Contrary to this, the majority of ant and spider species recorded in our study were food generalists and thus may better cope with the loss of some species at lower trophic rank compared to specialised herbivores or predators [23, 68], as long as primary productivity as a whole was sufficient. Another explanation for the observed pattern might be differences in mobility of the three focal groups. As plants are sessile, they are more strongly influenced by their immediate surroundings and can hardly evade unfavourable environmental conditions caused by urbanisation compared to ants and spiders. Furthermore, the seeds of most plant species recorded in the present study are dispersed by animals. Hence, negative impacts of urban sprawl on the behaviour, mobility and diversity of these seed dispersers, may have enhanced the vulnerability of plants to urbanisation. Several urban studies reported a replacement of forest specialists by generalist species with increasing degree of urbanisation suggesting that forest specialists are more sensitive to urbanisation-related disturbances [31, 32, 65]. While this was the case in ants, plants and spiders showed the lowest percentages of habitat specialists in forests located in areas with medium degree of urbanisation (15–30% sealed area). This finding was unexpected and may be a result of combined effects of differences in habitat diversity in the surroundings, which may be highest at medium levels of urbanisation, and of refugia effects of forests in highly urbanised areas. Effects of forest size and shape on species diversity As a consequence of proceeding urban development, many forest sites are characterised by intense isolation and small size. Thus, it is important to examine how habitat size affects biodiversity, and how this factor interacts with the degree of urbanisation in its surroundings. Theory of island biogeography predicts that small habitat patches contain less species than large habitat patches [69]. In this study, however, we did not find a species–area relationship for any of the three taxonomic groups examined. This result rejects our hypothesis and contrasts findings of previous studies on plants [70, 71] and web spiders [72] conducted in urban forests. Partly in line with our finding, Gibb and Hochuli [21] did not record a species–area relationship in spiders either and even reported an increase of ant species richness with decreasing forest size. Most studies, however, which failed to uncover area-related effects on species richness, were typically conducted in forests much larger (e.g. [21]: 4–80 km2) than those in our study. Even though forest size did not influence total species richness, we recorded higher percentages of forest specialists of ants and spiders in large than in small and medium-sized forests. Possible explanations might be a higher proportion of edge to different habitat types in small compared to large forests and, thus, a replacement of forest specialists by generalists and open-land species [21, 73]. Indeed, we found higher percentages of open-land species in small than large forests (ants: 32.3% vs. 20.1%; spiders: 11.0% vs. 4.9%). Regarding spiders, most open-land species were hunters in this study. We suggest that they may have temporarily visited forests for foraging rather than permanently living in them. Similarly, foraging ant workers from nests outside the fragments may have visited the edge zone of small forests. As forest size was not independent of forest history in our study, some of the observed differences in percentage forest specialists for ants and spiders may also be the result of some of the forest sites having previously been non-forested habitats. However, none of the forests were very recent in origin (all the study sites were marked as forest on old maps for at least 44–137 years), and even small fragments harboured forest specialist species. Indeed small forests were not per se less suitable habitats for forest specialists as demonstrated in plants. In our study, interestingly, the shape rather than the size/history of forests was the main predictor of the percentage of forest specialists. Forest sites, which were part of a continuous forests, and forests with a rather circular area (shape index 1–1.5) exhibited a higher percentage of forest specialists than forests with a more complex shape (shape index > 1.5). Hence, even small urban forest sites of comparably recent origin can serve as habitat for numerous forest specialists, if the proportion of edge to other habitat types and associated changes in the abiotic environment are minimized. However, most of the small forest sites in our study were dominated by a few plant species–independent of the degree of urbanisation in their surroundings. Effects of forest site characteristics on species diversity Plant species richness and the percentage of forest specialists were related to canopy closure considered as a proxy for light conditions, while soil moisture content was a key predictor of Shannon diversity and evenness, highlighting the importance of abiotic site characteristics for plant diversity. Similarly, in ants, canopy closure was important in explaining the percentage of forest specialists. Furthermore, leaf litter characteristics (litter moisture, litter pH) were important determinants for ant diversity. In urban forests, leaf litter biomass can be reduced as a result of recreational use. This would not only affect ant species with nests within this layer, but also the many species foraging there. In our study, the majority of both spider species and individuals belonged either to the family Linyphiidae (44.0% and 57.9%), which build their webs in leaf litter and mainly low vegetation, or Lycosidae (9.2% and 22.1%), which are active hunters. Hence, we expected a strong response of spiders to changes in the structural diversity of leaf litter and vegetation. Surprisingly, these two variables had no significant role in explaining variation in overall spider diversity. This lack of response may be partly explained by the habitat specificity of spider species, since we observed a trend towards an increase in the percentage of forest specialists with the amount of leaf litter biomass. This positive relationship may be also the reason for the high percentage of forest specialists recorded in large forests, which exhibited a higher amount of leaf litter biomass (mean: 335.2 g m-2) than small and medium-sized forests in this study (157.6 and 138.5 g m-2). Species composition Species composition may change even when species richness is maintained [19]. Urban communities can be a subset of the regional species pool, often biased towards generalists, which are better adjusted to a stressful environment [31, 32], or they may be novel by comprising many non-native species [3]. While we recorded few non-native species, the urban forests in this study harboured many generalist and open-land species, in line with other studies (e.g. [31, 32]). This is likely a consequence of differences in disturbance intensity and a small-scale habitat mosaic. Nevertheless, many forest specialists persisted including a few species listed as threatened for Switzerland. However, the red list for ants is out-dated and no such list exists for spiders, and we thus did not analyse threatened species separately. As our fragments were small compared to other studies on this topic (e.g. [21, 31, 71]), our findings highlight the sometimes-overlooked conservation value of even small, heavily disturbed habitats. PERMANOVA showed that as hypothesized, groups at higher trophic rank were more strongly affected by urbanisation. While this was not the case for species richness and diversity, the shift was visible in species composition. Plant species composition did only show a weak trend towards differences among the urbanisation classes, while species composition of the predaceous spiders significantly shifted with increasing degree of urbanisation. In line with our expectation, spider species composition was more similar in highly than in less urbanised areas. Ants fell between, with highly urbanised areas having a significantly changed species composition. Most spider species are generalist predators. We expected that urbanisation might affect specialised predators or parasites in our study area even more [23]. Indeed, none of the three social parasitic ant species, which use host species to found new colonies, were present in highly urbanised forest sites, even though one was common in seven other sites. Species at high trophic rank, therefore, should receive special attention when managing urban habitats. Interestingly, forest size was important for explaining species composition of all groups. This may have also partly reflected the effects of the history of the forest sites, as species composition may have not reached equilibrium yet in sites that had been previously non-forested habitats, or in forest fragments whose area has been reduced. This may also have affected some local environmental conditions such as soil-related factors, alongside current effects such as disturbance and forest management. However, none of the forests were very recent in origin. Local abiotic factors (soil moisture and soil orgN) were important drivers for plants species composition. In contrast, only canopy closure helped to explain ant species composition, while spider species composition was affected by both abiotic and structural forest characteristics. These results mirror the importance of local abiotic habitat characteristics as key drivers for plant species diversity measures. Combining results for species composition and diversity we find that both abiotic and structural forest characteristics are important in explaining arthropod diversity and species composition. Structural forest characteristics may be a surrogate for food availability. However, we did not directly measure food availability for arthropods, though e.g. SOM may be related to it, as it supports detrivores and thus potential prey [74, 75]. This finding indicates opportunities to increase the conservation values of urban forests, because local site characteristics are more amenable to management efforts than landscape factors. Functional dispersion It is expected that functional dispersion should decrease with increasing urbanisation because of an enhanced influence of environmental filtering in stressful urban environments. Some species fulfil unique roles, while others have similar functions within an ecosystem. Thus, local species loss or shifts in relative abundance can reduce the abundance and efficiency of functional traits in niche space and subsequently ecosystem functioning [76]. The observed changes in species composition in our study should thus translate to changes in functional diversity [61]. Indeed we observed that functional dispersion of plants decreased with increasing degrees of urbanisation. That this decrease in functional dispersion was a result of an increasingly stressful environment, was also supported by the finding that functional dispersion decreased with forest size. Small fragments with a high proportion of edge habitat were assumed to be exposed to most stress. In contrast to the situation found for plants, functional dispersion in the two arthropod groups was not influenced by these two main factors, with only that of ants showing a non-significant trend to be affected by forest size. Neither did functional dispersion change depending on most of the local environmental factors examined. Given the results from the PERMANOVAs, we would have expected the observed shifts in species composition to result in larger effects on functional dispersion also for the ground surface-active arthropod community. For example, the litter layers in some urban forest fragments were reduced as a consequence of the high levels of disturbance, which could have been expected to reduce habitat quality and thus the presence of functional groups associated with leaf litter. Go to: Conclusions Our results showed that species richness of the taxonomic groups was not an ideal indicator of biodiversity change in urban landscapes, as it masked shifts in species composition and relative abundance of species with different habitat specificity. Using a multi-taxa approach, we further found that the effect of urbanisation on species composition increased with trophic rank. This highlights the necessity to consider different taxonomic and functional groups in urban planning to maximize conservation value of urban green areas. In the short term, urban planners could focus on small-scale environmental factors, which proved to be important determinants of species diversity and species composition. For example, protection of litter layers and ground vegetation could be enhanced using simple management practices. However, the influence of large-scale factors like the proportion of sealed area in the surroundings and forest size on forest specialists indicates that also more complex changes at the landscape level are essential to maintain vulnerable elements of forest communities. Go to: Supporting information S1 Fig Forest specialists in relation to size and shape of urban forests. Percentage forest specialist species of a) ants and b) spiders in fragments of different size; size classes are small (< 4000 m2), medium-sized (4000–10,000 m2) and large (> 10,000 m2); and c) percentage of forest specialist species of plants depending on the shape of the fragment. The shape index was calculated following Gyenizse et al. [28]. A shape index of 1 corresponds to a circular area, which is considered as most stable and resistant against biotic and abiotic effects from the surrounding landscape. Classes are A: continuous forest, B: shape index between 1 and 1.5, C: shape index > 1.5. (DOCX) Click here for additional data file.(367K, docx) S1 Table Description of forest sites. Characteristics of the 26 forests examined in Basel (Switzerland) and its surroundings. (DOCX) Click here for additional data file.(98K, docx) S2 Table Landscape, forest and plot characteristics recorded during field surveys. (XLS) Click here for additional data file.(72K, xls) S3 Table Species and trait lists. Species list of (a) vascular plants, (b) ants and (c) spiders. Habitat specificity, conservation status (Red List) and a set of traits, which we considered to influence species’ response to urbanisation-related factors are shown. Traits not used for analyses are in parentheses. (DOCX) Click here for additional data file.(190K, docx) S4 Table Correlations within and among landscape and site characteristics. Results of Pearson’s (r) and Spearman’s rank (rs) correlation, Contingency table (χ2-test) and Kruskal-Wallis test examining the relationship between observed species richness and rarefied species richness (a) and among landscape and forest characteristics for all three taxonomic groups (b), in the vegetation plots (c) and in the trap-grid system (d) (DOCX) Click here for additional data file.(106K, docx) S5 Table Species–site matrices for plants, ants, and spiders. For the arthropod survey, the number of recollected traps per forest site is presented. (XLSX) Click here for additional data file.(760K, xlsx) S6 Table Data used to perform PERMANOVA. (XLSX) Click here for additional data file.(51K, xlsx) S7 Table Percentage of sites in which common species occur for different degrees of urbanisation and forest size classes. Common species are defined as occurring in at least 10 of the sites. Means are given for less common species. (XLS) Click here for additional data file.(50K, xls) S8 Table Functional dispersion: Summary of ANCOVAs of plants. Summary of ANCOVAs examining the effects of degree of urbanisation, forest size and shape, forest management (time since last thinning), disturbance (indicated by path density), canopy closure, soil characteristics (moisture, pH, soil organic nitrogen (orgN) and phosphorus (orgP) content) and cover of ground vegetation on functional dispersion of vascular plants. (DOCX) Click here for additional data file.(70K, docx) S9 Table Functional dispersion: Summary of ANCOVAs of ants and spiders. Summary of ANCOVAs examining the effects of degree of urbanisation, forest size and shape, forest management (time since last thinning), disturbance (indicated by path density), canopy closure, soil organic matter content, litter characteristics (moisture, pH) and structural diversity measures (litter biomass, vegetation structure and amount of dead wood) on functional dispersion of ants and spiders. (DOCX) Click here for additional data file.(91K, docx) Go to: Acknowledgments We thank A. Hänggi for help with spider identification and R. Neumeyer and B. Seifert for verifications of some ant identifications. We further thank D. Binggeli, J. Hart, D. Milner, K. Reinacher and R. Schneider for sorting of pitfall-trap content. We also thank the foresters (Amt für Wald beider Basel), Christoph Merian Stiftung, Industrielle Werke Basel, Stadtgärtnerei Basel and private owners for access to the study sites. We thank P. Heneberg and three anonymous reviewers for comments on an earlier draft of this manuscript. Go to: Funding Statement Stadtgärtnerei Basel (BBaur), Stiftung Emilia Guggenheim-Schnurr (BBraschler), and Basler Stiftung für biologische Forschung (BBraschler) provided funding for the research. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Go to: Data Availability All relevant data are within the paper and its Supporting Information files. Go to: References 1. Grimm NB, Faeth SH, Golubiewski NE, Redman CL, Wu J, Bai X, et al. Global change and the ecology of cities. Science. 2008; 319: 756–760. doi: 10.1126/science.1150195 [PubMed] 2. Niemelä J. Ecology and urban planning. Biodivers Conserv. 1999; 8: 119–131. doi: 10.1023/A:1008817325994 3. McKinney ML. Urbanization, biodiversity and conservation. BioScience. 2002; 52: 883–890. doi: 10.1043/0006-3568(2002)052(0883:UBAC)2.0.CO;2 4. Gilbert OL. Städtische Ökosysteme Radebeul: Neumann; 1994. 5. Pickett STA, Cadenasso ML, Grove JM, Boone CG, Groffman PM, Irwin E, et al. Urban ecology systems: scientific foundations and a decade of progress. J Environ Manage. 2011; 92: 331–362. doi: 10.1016/j.jenvman.2010.08.022 [PubMed] 6. Sukopp H. Urban ecology: scientific and practical aspects In: Breuste J, Feldmann H, Uhlmann O (eds), Urban Ecology. Berlin: Springer; 1998. pp. 3–16. 7. Concepción ED, Moretti M, Altermatt F, Nobis MP, Obrist MK. Impacts of urbanisation on biodiversity: the role of species mobility, degree of specialisation and spatial scale. Oikos. 2015; 124: 1571–1582. doi: 10.1111/oik.02166 8. Chapin FS, Walker BH, Hobbs RJ, Hooper DU, Lawton JH, Sala OE, et al. Biotic control over the functioning of ecosystems. Science. 1997; 277: 500–504. doi: 10.1126/science.277.5325.500 9. Bradley CA, Altizer S. Urbanization and the ecology of wildlife diseases. Trends Ecol. Evol. 2007; 22: 95–102. doi: 10.1016/j.tree.2006.11.001 [PubMed] 10. Sitko J, Zaleśny G. The effect of urbanization on helminth communities in the Eurasian blackbird (Turdus merula L.) from the eastern part of the Czech Republic. J. Helminthology. 2014; 88: 97–104. doi: 10.1017/S0022149X12000818 [PubMed] 11. Máximo HJ, Felizatti HL, Ceccato M, Cintra-Socolowski P, Beretta ALRZ. Ants as vector of pathogenic microorganisms in a hospital in Sao Paulo county, Brazil. BMC Research Notes 2014; 7:554 doi: 10.1186/1756-0500-7-554 [PMC free article] [PubMed] 12. Godefroid S, Koedam N. Urban plant species patterns are highly driven by density and function of built-up areas. Landscape Ecol. 2007; 22: 1227–1239. doi: 10.1007/s10980-007-9102-x 13. Kühn I, Brandl R, Klotz S. The flora of German cities is naturally species rich. Evol Ecol Res. 2004; 6: 749–764 14. Cvejić R, Eler K, Pintar M, Železnikar Š, Haase D, Kabisch N, Strohbach M et al. D3.1 A typology of urban green spaces, ecosystem services provisioning services and demands. Work package 3 2015; 15. Bolund P, Hunhammer S. Ecosystem services in urban areas. Ecol Econ. 1999; 29: 293–301. doi: 10.1016/S0921-8009(99)00013-0 16. Croci S, Butet A, Georges A, Aguejdad R, Clergeau P. Small urban woodlands as biodiversity conservation hot-spot: a multi-taxon approach. Landscape Ecol. 2008; 23: 1171–1186. doi: 10.1007/s10980-008-9257-0 17. Horák J, Rom J, Rada P, Šafářová L, Koudelková J, Zasadil P et al. Renaissance of a rural artifact in a city with a million people: biodiversity responses to an agro-forestry restoration in a large urban traditional fruit orchard. Urban Ecosyst. 2018; online available. doi.org/10.1007/s11252-017-0712-z 18. Kowarik I, Buchholz S, von der Lippe M, Seitz B. Biodiversity functions of urban cemeteries: Evidence from one of the largest Jewish cemeteries in Europe. Urban Forest. Urban Green. 2016; 19: 68–78. doi.org/10.1016/j.ufug.2016.06.023 19. McIntyre NE, Rango J, Fagan WF, Faeth SH. Ground arthropod community structure in a heterogeneous urban environment. Landscape Urban Plan. 2001; 52: 257–274. doi: 10.1016/S0169-2046(00)00122-5 20. Lizée M-H, Mauffrey J-F, Tatoni T, Dechamps-Cottin M. Monitoring urban environments on the basis of biological traits. Ecol Indic. 2011; 11: 353–361. doi: 10.1016/j.ecolind.2010.06.003 21. Gibb H, Hochuli DF. Habitat fragmentation in an urban environment: large and small fragments support different arthropod assemblages. Biol Conserv. 2002; 106: 91–100. doi: 10.1016/S0006-3207(01)00232-4 22. Magura T, Tóthmérész B, Molnár T. Changes in carabid beetle assemblages along an urbanisation gradient in the city of Debrecen, Hungary. Landscape Ecol. 2004; 19: 747–759. doi: 10.1007/s10980-005-1128-4 23. Holt RD, Lawton JH, Polis GA, Martinez ND. Trophic rank and the species–area relationship. Ecology. 1999; 80: 1495–1504. doi: 10.1890/0012-9658(1999)080[1495:TRATSA]2.0.CO;2 24. Steffan-Dewenter I. Importance of habitat area and landscape context for species richness of bees and wasps in fragmented orchard meadows. Conserv Biol. 2003; 17: 1036–1044. doi: 10.1046/j.1523-1739.2003.01575.x 25. Vallet J, Beaujouan V, Pithon J, Rozé F, Daniel H. The effects of urban or rural landscape context and distance from the edge on native woodland plant communities. Biodivers Conserv. 2010; 19: 3375–3392. doi: 10.1007/s10531-010-9901-2 26. Gaublomme E, Hendrickx F, Dhuyvetter H, Desender K. The effects of forest patch size and matrix type on changes in carabid beetle assemblages in an urbanized landscape. Biol. Conserv. 2008; 141: 2585–2596. doi: 10.1016/j.biocon.2008.07.022 27. Chace JF, Walsh JJ. Urban effects on native avifauna: a review. Landscape Urban Plan. 2006; 74: 46–69. doi: 10.1016/j.landurbplan.2004.08.007 28. McKinney ML. Effects of urbanization on species richness: a review of plants and animals. Urban Ecosyst. 2008; 11: 161–176. doi: 10.1007/s11252-007-0045-4 29. Bolger DT, Suarez AV, Crooks KR, Morrison SA, Case TJ. Arthropods in urban habitat fragments in Southern California: Area, age, and edge effects. Ecol. Appl. 2000; 10: 1230–1248. doi: 10.1890/1051-0761(2000)010[1230:AIUHFI]2.0.CO;2 30. Rota E, Caruso T, Migliorini M, Monaci F, Agamennone V, Biagini G et al. Diversity and abundance of soil arthropods in urban and suburban holm oak stands. Urban Ecosyst. 2015; 18: 715–728. doi: 10.1007/s11252-014-0425-5 31. Deichsel R. Species change in an urban setting–ground and rove beetles (Coleoptera: Carabidae and Staphylinidae) in Berlin. Urban Ecosyst. 2006; 9: 161–178. doi: 10.1007/s11252-006-8588-3 32. Vergnes A, Pellissier V, Lemperiere G, Rollard C, Clergeau P. Urban densification causes the decline of ground-dwelling arthropods. Biodivers Conserv. 2014; 23: 1859–1877. doi: 10.1007/s10531-014-0689-3 33. Fujita A, Maeto K, Kagawa Y, Ito N. Effects of forest fragmentation on species richness and composition of ground beetles (Coleoptera: Carabidae and Brachinidae) in urban landscapes. Entomol. Science. 2008; 11: 39–48. doi: 10.1111/j.1479-8298.2007.00243.x 34. Gibbs JP, Stanton EJ. Habitat fragmentation and arthropod community change: carrion beetles, phoretic mites, and flies. Ecol. Appl. 2001; 11: 79–85. doi: 10.1890/1051-0761(2001)011[0079:HFAACC]2.0.CO;2 35. Melliger RL, Rusterholz H-P, Baur B. Habitat- and matrix-related differences in species diversity and trait richness of vascular plants, Orthoptera and Lepidoptera in an urban landscape. Urban Ecosyst. 2017; 20: 1095–1107. doi: 10.1007/s11252-017-0662-5 36. McDonnell MJ, Hahs AK. The use of gradient analysis studies in advancing our understanding of the ecology of urbanizing landscapes: current status and future directions. Landscape Ecol. 2008; 23: 1143–1155. doi: 10.1007/s10980-008-9253-4 37. Burnand J, Hasspacher B. Waldstandorte beider Basel. Quellen und Forschungen zur Geschichte und Landeskunde des Kantons Basel-Landschaft, Band 72. Verlag des Kantons Basel-Landschaft; 1999. 38. Gyenizse P, Bognár Z, Czigány S, Elekes T. Landscape shape index, as a potential indicator of urban development in Hungary. Landsc Environ. 2014; 8: 78–88 39. Braun-Blanquet J. Pflanzensoziologie: Grundzüge der Vegetationskunde New York: Springer; 1964. 40. Seifert B. Die Ameisen Mittel- und Nordeuropas Görlitz: Lutra; 2007. 41. Ward PS, Blaimer BB, Fisher BL. A revised phylogenetic classification of the ant subfamily Formicinae (Hymenoptera: Formicidae), with resurrection of the genera Colobopsis and Dinomyrmex. Zootaxa. 2016; 4072: 343–357. doi: 10.11646/zootaxa.4072.3.4 [PubMed] 42. Roberts MJ. The spiders of Great Britain and Ireland, Volume 1: Atypidae to Theridiosomatidae Colchester: Harley Books; 1985. 43. Roberts MJ. The spiders of Great Britain and Ireland, Volume 2: Linyphiidae and checklist Colchester Harley: Books; 1987. 44. Nentwig W, Blick T, Gloor D, Hänggi A, Kropf C. Spiders of Europe. 2017; http://www.araneae.unibe.ch, ver. 03.2017. 10.24436/1 45. Schuler J. Baumbewohnende Ameisen mitteleuropäischer Auenwälder. Artenspektrum und Ökologie arborikoler Ameisen in naturnahen Hartholzauen an Rhein, Elbe und Donau Zürich: Bristol-Stiftung; Bern: Haupt; 2015. 46. Delarze R, Gonseth Y, Eggenberg S. Lebensräume der Schweiz: Ökologie–Gefährdung–Kennarten, 3rd edn. Bern: Ott Verlag; 2015. 47. Hänggi A, Stöckli E, Nentwig W. Lebensräume Mitteleuropäischer Spinnen Neuchâtel: Miscellanea Faunistica Helvetiae 4. Centre de cartographie de la faune (CSCF); 1995. 48. Grime JP. Plant strategies and vegetation processes Chichester: Wiley; 1979. 49. Klotz S, Kühn I, Durka W. BIOLFLOR–Eine Datenbank mit biologisch-ökologischen Merkmalen zur Flora von Deutschland Bonn: LandWirtschaftsverlag; 2002; http://www2.ufz.de/biolflor (accessed 9 May 2017) 50. Müller-Schneider P. Verbreitungsbiologie der Blütenpflanzen Graubündens, 85 Heft. Veröffentl. Geobot. Inst. ETH, Zurich: Stiftung Rübel; 1986. 51. Wiki der Arachnologischen Gesellschaft e.V. ‘Hauptseite’. 2017; https://wiki.arages.de/index.php?title=Hauptseite&oldid=91730 (accessed 23 May 2017) 52. Allen SE. Chemical analysis of ecological material, 2nd ed. Oxford: Blackwell Scientific; 1989. 53. Bremner JM. Total nitrogen In: Black C.A. (ed), Methods of soil analysis, Part 2. Madison: American Society of Agronomy; 1965. pp. 1149–1178. 54. Sparks DL, Page AL, Helmke PA, Loeppert RA, Soltanpour PN, Tabatabai MA, et al. Methods of soil analysis Part 3 –Chemical methods, 3rd ed. Madison: SSSA; 1996. doi: 10.2136/sssabookser5.3.c32 55. Godínez-Alvarez H, Herrick JE, Mattocks M, Toledo D, Van Zee J. Comparison of three vegetation monitoring methods: their relative utility for ecological assessment and monitoring. Ecol Indic. 2009; 9: 1001–1008. doi: 10.1016/j.ecolind.2008.11.011 56. Melliger RL, Rusterholz H-P, Baur B. Ecosystem functioning in cities: Combined effects of urbanization and forest size on early-stage leaf litter decomposition of European beech (Fagus sylvatica L.). Urban For. Urban Green. 2017; 28: 88–96. doi: 10.1016/j.ufug.2017.10.009 57. Hothorn T, Bretz F, Westfall P. Simultaneous inference in general parametric models. Biometr. J. 2008; 50: 346–363. doi: 10.1002/bimj.200810425 [PubMed] 58. Bray JR, Curtis JT. An ordination of the upland forest communities of Southern Wisconsin. Ecol. Monogr. 1957; 27: 325–349. doi: 10.2307/1942268 59. Anderson MJ. PERMANOVA: a FORTRAN computer program for permutational multivariate analysis of variance Department of Statistics, University of Auckland, New Zealand: 2005. 60. Oksanen J, Blanchet FG, Friendly M, Kindt R, Legendre P, McGlinn D, et al. Vegan: community ecology package. R package version 2.4–2. 2017. https://CRAN.R-project.org/package=vegan 61. Villéger S, Mason NW, Mouillot D. New multidimensional functional diversity indices for a multifaceted framework in functional ecology. Ecology. 2008; 89: 2290–2301. doi: 10.1890/07-1206.1 [PubMed] 62. Laliberté E, Legendre P, Shipley B. FD: measuring functional diversity from multiple traits, and other tools for functional ecology. R package version 1.0–12. 2014. 63. Cameron GN, Culley TM, Kolbe SE, Miller AI, Matter SF. Effects of urbanization on herbaceous forest vegetation: the relative impacts of soil, geography, forest composition, human access, and an invasive shrub. Urban Ecosyst. 2015; 18: 1051–1069. doi: 10.1007/s11252-015-0472-6 64. Alaruikka D, Kotze DJ, Matveinen K, Niemelä J. Carabid beetle and spider assemblages along a forested urban–rural gradient in southern Finland. J Insect Conserv. 2002; 6: 195–206. doi: 10.1023/A:1024432830064 65. Magura T, Horváth R, Tóthmérész B. Effects of urbanization on ground-dwelling spiders in forest patches, in Hungary. Landscape Ecol. 2010; 25: 621–629. doi: 10.1007/s10980-009-9445-6 66. Horváth R, Magura T, Tóthmérész B. Ignoring ecological demands masks the real effect of urbanization: a case study of ground-dwelling spiders along a rural–urban gradient in a lowland forest in Hungary. Ecol Res. 2012; 27: 1069–1077. doi: 10.1007/s11284-012-0988-7 67. Aubin I, Venier L, Pearce J, Moretti M (2013) Can a trait-based multi-taxa approach improve our assessment of forest management impact on biodiversity? Biodivers Conserv. 2013; 22: 2957–2975. doi: 10.1007/s10531-013-0565-6 68. Didham RK, Ghazoul J, Stork NE, Davis AJ. Insects in fragmented forests: a functional approach. Trends Ecol Evol. 1996; 11: 255–260. doi: 10.1016/0169-5347(96)20047-3 [PubMed] 69. MacArthur RH, Wilson EO. The theory of island biogeography New Jersey: Princeton University Press; 1967. 70. Hobbs ER. Species richness of urban forest patches and implications for urban landscape diversity. Landscape Ecol. 1988; 1: 141–152. doi: 10.1007/BF00162740 71. Godefroid S, Koedam N. How important are large vs. small forest remnants for the conservation of the woodland flora in an urban context? Global Ecol Biogeogr. 2003; 12: 287–298. doi: 10.1046/j.1466-822X.2003.00035.x 72. Miyashita T, Shinkaib A, Chidac T. The effects of forest fragmentation on web spider communities in urban areas. Biol Conserv. 1998; 86: 357–364. doi: 10.1016/S0006-3207(98)00025-1 73. Ivanov K, Keiper J. Ant (Hymenoptera: Formicidae) diversity and community composition along sharp urban forest edges. Biodivers Conserv. 2010; 19: 3917–3933. doi: 10.1007/s10531-010-9937-3 74. Bengtsson J, Lundkvist H, Saetre P, Sohlenius B, Solbreck B. Effects of organic matter removal on the soil food web: forestry practices meet ecological theory. Appl Soil Ecol. 1998; 9: 137–143. doi: 10.1016/S0929-1393(98)00067-5 75. Moore JC, Berlow EL, Coleman DC, de Ruiter PC, Dong Q, Hasting A, et al. Detritus, trophic dynamics and biodiversity. Ecol Lett. 2004; 7: 584–600. doi: 10.1111/j.1461-0248.2004.00606.x 76. Mason NWH, Mouillot D, Lee WG, Wilson JB. Functional richness, functional evenness and functional divergence: the primary components of functional diversity. Oikos. 2005; 111: 112–118. doi: 10.1111/j.0030-1299.2005.13886.x Articles from PLoS ONE are provided here courtesy of Public Library of Science