Hydrogen cyanide, a key plant defense, as a potential driver of root-associated nematode communities along urbanization gradients

Introduction Plant chemical defenses can influence the distribution, community composition, and abundance of soil biota. Urbanization plays a key role in shaping soil biotic communities either directly through changes in soil properties or indirectly via changes in plant characteristics such as defense traits. The effects of urbanization and plant defenses on the abundance and structure of aboveground plant-associated communities have been studied, yet their effects on belowground root-associated communities are poorly understood. Methods Here we sampled white clover (Trifolium repens L.) leaves and roots along urban–rural gradients in the cities of Antwerp and Ghent, Belgium. We measured production of hydrogen cyanide (HCN) in leaves, a known defense trait against herbivores, and abundances of different feeding guilds of nematodes associated with the roots. Results We found that HCN production decreased with increasing levels of urbanization in both cities. Urbanization was significantly correlated with shifts in root-associated nematode community structure in Antwerp but not in Ghent. Responses of nematode feeding guilds and trophic groups to urbanization were highly dependent on the clovers’ HCN production, especially in Ghent. Changes in nematode channel ratio in Antwerp indicated that urban root-associated nematode communities of white clover were more strongly dominated by fungivorous nematodes. Discussion Our results demonstrate that urbanization is driving changes in a plant phenotypic trait and in the community structure of root-associated nematodes, as well as that both changes interact. Plant defense mechanisms could thus help elucidate the effects of urbanization on root-associated biota communities. As strong differences existed between the two studied cities, the particular properties of cities should be taken into account to better understand the direction and strength of phenotypic trait changes driven by urbanization.


Introduction
Urbanization has become a global driver of both environmental and biodiversity change due to the human population boom and city expansion worldwide (United Nations, 2019 (McDonnell et al., 2009). Urban soils are thus dramatically altered, physically and chemically (Lehmann and Stahr, 2007). The composition and structure of the soil food web commonly aggregated in the rhizosphere are often directly affected by human-induced changes in soil properties , such as agricultural intensification and urbanization (Pothula et al., 2019), and indirectly through changes in plant traits, including plant defenses (Hawlena and Zaguri, 2016). Both urbanization and plant defenses can influence aboveground communities (Miles et al., 2019;Qu et al., 2022), yet much less is known about their effects on soil biota.
Nematodes dominate the soil food web and function as useful indicators of soil condition regarding environmental disturbance (Yeates, 2003;Ferris, 2010). Specifically, nematodes occupy various trophic positions and have diverse feeding habits. They include bacterivores, fungivores, herbivores, omnivores and predators (Yeates et al., 1993). Several studies have demonstrated that nematodes are affected by urbanization, but not all functional groups are affected in the same way (Pothula et al., 2019;Li et al., 2022). Generally, urbanization decreases the overall abundance of nematodes. Abundances of omnivorous and predatory nematodes at higher trophic levels decrease in urbanized sites, while nematodes at lower trophic levels (i.e., herbivores, bacterivores, fungivores) often display opposite trends (Pavao-Zuckerman and Coleman, 2007). This may be attributed to differences in the sensitivity to environmental perturbances, with higher sensitivity of nematodes at higher trophic levels (Bongers and Ferris, 1999). In addition, nematode indices based on the presence and abundance of taxa and trophic groups are useful to evaluate the structure and function of the soil food web (Ferris et al., 2001). For example, the Nematode Channel Ratio (NCR) indicates the predominant decomposition pathway in the soil. The NCR varies from zero to one, with increasing values indicating a relative shift from bacterial to fungal decomposition (Ferris et al., 2001;Yeates, 2003). NCR values were reported to not change (Li et al., 2011) or decrease with increased urbanization levels (Pouyat et al., 1994;Pavao-Zuckerman and Coleman, 2007). However, the underlying mechanisms in relation to specific urban drivers behind these patterns remain unclear.
There are, so far, few studies linking patterns in soil communities to drivers associated with urbanization. Urban land use has profound effects on soil physiochemical properties. Herbivorous nematodes can benefit from urban soils that are sandy with lower clay content but also from rural soils with higher content of organic materials (Li et al., 2011). Besides, urban soils experienced highest heavy metal pollution compared to rural and suburban soils due to industrial, construction and agroforestry development in the city (Pouyat et al., 1994;Li et al., 2011;Ugarte and Taylor, 2020). Metal pollution immediately reduced total nematode abundance and the abundance of omnivores and carnivores but increased the NCR (Li et al., 2006;Pen-Mouratov et al., 2010;Šalamún et al., 2012). Moreover, soil management practices such as fertilization and tillage are key drivers of soil nematode community composition (Biswal, 2022;Li et al., 2022). Furthermore, the movement of nematodes strongly depends on soil water content (Freckman et al., 1987). Additionally, any management related to soil moisture can have important effects on nematode communities. Importantly, the key drivers of nematode abundance differ among trophic groups, with bacterivores and fungivores sensitive to soil characteristics, while plant feeders are strongly affected by vegetation traits (van den Hoogen et al., 2019). However, except for Francini et al. (2018), who documented effects of plant identity on abundance and composition of soil nematode communities within an urban context, few studies have investigated whether the effect of urbanization on soil nematode communities is related to changes in plant traits. Here, we studied the role of plant traits, more specifically secondary metabolites, in structuring root-associated nematode assemblages.
Plant secondary metabolites are recognized as a major defense mechanism against herbivores and therefore play a key role in shaping the structure of biotic communities (Bezemer and van Dam, 2005;van Dam and Heil, 2011). The rhizosphere community can be influenced by secondary metabolites released from root exudates and from leaf litter returned to the soil (Chomel et al., 2016;Tsunoda and van Dam, 2017;Zhang et al., 2019). These secondary metabolites can act as attractants, repellents or inhibitors for herbivorous nematodes (Abgrall et al., 2018;Sikder and Vestergård, 2020), and further influence the overall nematode community through trophic cascades. Plant metabolites also have strong influences on soil microorganisms and therefore on microbes and microbivorous nematodes and, consequently, indirectly on the omnivorous and predatory nematodes that feed on them (Tsunoda and van Dam, 2017). Generally, responses of nematode abundance and community structure to leaf litter highly depends on plant species identity and associated differences in litter quality (Cesarz et al., 2013). However, secondary metabolites in leaves showed variation among populations along urbanization gradients Moreira et al., 2018;Santangelo et al., 2020). Such variation can have an impact on belowground multi-trophic interactions (van Geem et al., 2016). We therefore expected that urbanization can indirectly influence nematode communities through its effects on plant traits.
White clover (Trifolium repens L., Fabaceae) is a cosmopolitan herb and can be found in nearly all temperate cities. It is capable of cyanogenesis, that is the ability to release hydrogen cyanide (HCN) upon tissue damage (Hughes, 1991). HCN not only defends against small generalist herbivores aboveground, including snails, slugs and insects (Burgess and Ennos, 1987; but also suppresses nematodes in the soil (Dutta et al., 2019). For example, incorporating white clover leaf material into soil caused a 45% reduction in severity of root galling by Meloidogyne hapla (Widmer and Abawi, 2002). Furthermore, there was a strong correlation between the concentration of HCN in sudangrass leaves incorporated into soil and the reduction of M. hapla root-galling severity (Widmer and Abawi, 2002). Recent work reported that urbanization has driven parallel clines of HCN frequency in white clover populations in 47% of cities across 160 cities globally, with 39% of cities showing a declined frequency of HCN in urban populations (Santangelo et al., 2022). Such urban-rural clines were unrelated to leaf herbivory at least in some cities , but whether they were related to root-associated nematode communities is unknown.
In this study, we explored the relationships between rootassociated nematode communities and plant secondary metabolites along urban-rural gradients in the cities of Antwerp and Ghent, Belgium. We determined whether the abundance and community composition of these nematodes changed with HCN frequency in white clover leaves and urbanization levels to address four hypotheses. As urbanization drives homogenization of the environment among distinct cities (Groffman et al., 2014) and the intensity of urbanization is positively related to city size (Bettencourt and West, 2010), and given the prevailing trend of declined HCN frequencies in urban areas worldwide (Santangelo et al., 2022), we hypothesize that (1) the HCN Frontiers in Ecology and Evolution 03 frontiersin.org frequencies would decline with urbanization in the two cities, and that the strength of the decline would be stronger in the larger city (Antwerp). Considering that HCN possibly reduces litter quality with potential cascading effects across trophic groups, (2) HCN frequency would reduce the abundance of nematodes for all feeding types and trophic groups. Because of less-variable environmental conditions including impervious surface among urban populations within cities compared to rural populations (Santangelo et al., 2022), (3) we hypothesize a stronger negative effect of HCN in rural areas on rootassociated nematode communities. Furthermore, (4) we expect the more disturbance-sensitive predatory and omnivorous nematodes to respond more strongly to urbanization.

Study system
White clover (Trifolium repens) is a perennial herbaceous legume that is native to Eurasia and has been introduced to temperate regions globally as a nitrogen-fixing forage crop and soil stabilizer (Burdon, 1983;Kjaergaard, 2003). White clover is typically found in grazed or mowed pastures, lawns, meadows and roadsides where it can maintain large and dense populations (Burdon, 1983). In addition to sexual reproduction, T. repens reproduces clonally, with horizontal stems (i.e., stolons) that spread along the ground. A single clone can be up to 0.5 m in diameter (Burdon, 1983).
Cyanogenesis occurs as a discrete Mendelian polymorphism in white clover, with cyanogenic (HCN present) and acyanogenic (HCN absent) plants co-occurring in many natural populations (Daday, 1958). The molecular genetics of the cyanogenesis, controlled by two individual Mendelian genes, has recently been documented in detail (Olsen et al., 2007(Olsen et al., , 2013Olsen and Small, 2018). Ac/ac determines the synthesis of cyanogenic glucosides (linamarin and lotaustralin) in the cell vacuole; Li/li determines the synthesis of hydrolysing enzyme (linamarase/β-glucosidase) in the cell wall. For both Ac/ac and Li/li, the dominant allele expresses the cyanogenic precursor. Therefore, plants with at least one functional allele at each locus are cyanogenic and produce HCN upon tissue damage, for example due to herbivore attacks. Both cyanogenic glucosides and linamarase are synthesized during shoot growth and stored in the mature leaves (Collinge and Hughes, 1982). Hence, the roots and seedlings before shoot emergence are not cyanogenic.

Study sites and sampling design
The study was conducted in two cities (Antwerp and Ghent) in Belgium in north-west Europe (Supplementary Figure S1). Belgium is a highly urbanized country (average human population density: 380.8 inhabitants/km 2 ; Eurostat, 2021), and the landscape of both sites is composed of highly urbanized, agricultural, and seminatural areas.
From September to November 2018, we selected 40 white clover populations that were spaced at least 200 meters apart from each other along an urban-rural gradient within each city. For each population, we recorded the coordinates at the center using a hand-held GPS device (± 5 m accuracy) and collected 10 to 20 stolons that had three to four intact leaves. Stolons had a minimum distance of 1.5 m between each other to minimize the probability of collecting the same clonal genotype. Afterwards, roots were excavated at several locations and pooled to obtain a total fresh weight of up to 6 g. In total, we sampled stolons from 697 and 664 plants and roots from 24 and 27 populations in Antwerp and Ghent, respectively. Samples of stolons and roots (per population) were separately placed in plastic bags and preserved in a cooler with ice bags until being transported to a fridge at 4°C in the lab. Stolons were individually transferred to 2 ml microcentrifuge tubes and preserved at−80°C until HCN assay.
As urbanization involves many factors that characterize different physical environments, we used the proportion of built-up cover (%BUC) as a proxy for urbanization and BUC only included all buildings but not roads, parking infrastructure and pavements. This proxy was determined from digital maps in a GIS software for a nested series of spatial scales (50, 100, 200, 400, 800, 1,600, and 3,200 m radii, Supplementary Figure S1; Merckx et al., 2018) surrounding the center of each population and expressed as a continuous variable. Meanwhile, populations with BUC higher than 15% within 3,200 m surrounding each population were assigned to "urban, " otherwise to "rural. "

Cyanogenesis analysis
To assess the frequency of cyanogenesis in natural T. repens populations, we detected the absence/presence of HCN in each plant/ stolon using Feigl-Anger assays based on a color change reaction (Feigl and Anger, 1996;Gleadow et al., 2011). We took samples out of the freezer, let them thaw, and put a single intact leaf (three 1-cm-wide leaflets) of each stolon in wells of 96-well plates. Subsequently, we added 80 μl of distilled water to each well and macerated tissue samples using pipette tips. The plates were covered with Feigl-Anger test paper and put in an incubator at 37°C for 3 h. Cyanogenic plants were indicated by a blue color.

Nematode analysis
Roots were separated from bulk soil and rinsed with tap water. If fresh roots were more than 3 g, they were divided into two portions and separately weighed to the nearest 0.001 g. One portion was used to extract nematodes, and another one to determine the relationship between dry and fresh root weights. Otherwise, all roots were used for nematode extraction. Nine out of 24 populations in Antwerp, and 20 out of 27 populations in Ghent, had extra roots for determining the dry weight. These roots were oven-dried at 50°C for 72 h and weighed to the nearest 0.001 g. For Antwerp, this relationship was fitted well by a linear regression (dry root weight = 0.018 + 0.1547* fresh root weight, R 2 = 0.86, n = 9). For Ghent, among urban populations where estimates were needed, the best fitting relationship was dry root weight = 0.1743*1.3896 fresh root weight (R 2 = 0.62, n = 7). According to this relationship, we could predict dry weight of roots used for nematode extraction.
Nematodes were extracted using the Baermann funnel technique (Whitehead and Hemming, 1965). Extractions were collected three times with approximately 15 ml every day in three consecutive days and kept in a fridge at 4°C for 3 days until identification. Nematodes were counted and identified to each of five feeding guilds based on mouthpart, cuticle, and pharynx morphology (Yeates et al., 1993): Frontiers in Ecology and Evolution 04 frontiersin.org bacterial-feeder (BF), fungal-feeder (FF), plant-feeder (PF), predator (PR) and omnivore (OM). A small proportion (less than 2% across the two cities) of nematodes was difficult to identify and categorized as unknown. Total abundance of nematodes in each population was the sum of the five feeding guilds and the unknown group. Nematode Channel Ratio is estimated by the ratio of bacterial-feeders to bacterial-feeders plus fungal-feeders, NCR = BF/(BF + FF), and has values between 0 (totally fungal-dominated) and 1 (totally bacterialdominated) (Yeates, 2003). To depict trophic levels of the rootassociated food web based on its nematode community, we defined PF as the first trophic-level (TL1), BF and FF as the second trophic-level (TL2), and PR and OM as the third trophic-level (TL3) (Peralta et al., 2020).

Statistics
We performed all statistical analyzes (Supplementary Figure S2) in R 3.6.3 (R Development Core Team, 2018).

Spatial scale of urbanization and general approach
We tested for each response variable at which spatial scale of urbanization the corresponding model explained most variation. We plotted R 2 values from linear models (LMs) or PERMANOVA, and Nagelkerke's pseudo-R 2 from generalized linear models (GLMs) for each response variable (see below) against the seven scales (50-3,200 m radii) at which urbanization was evaluated ( Supplementary Figures S3-S5). For each response variable analyzed both across and within the cities of Antwerp and Ghent, we decided the best scale with the highest R 2 values for at least in two of these three analyzes. Afterwards, p values from either type-III Chi 2 Wald tests or F tests for effects of independent variables were estimated by the function ' Anova' ('car' package) for GLMs and LMs, respectively.

Patterns of frequency of HCN
The frequency of HCN was calculated as the proportion of all tested plants that was cyanogenic within each population. To examine whether the proportion of HCN-producing plants varied with urbanization levels and across cities, we fitted a series of GLMs at each of the seven spatial scales, with city, urbanization level, and their interaction as predictors. A 'quasibinomial' distribution was used to account for overdispersion of binomial data.

Nematode abundance and nematode channel ratio
We first tested whether the influence of urbanization and HCN frequency on nematode abundance differed between the cities of Antwerp and Ghent, with an individual GLM for each of the seven scales. Nematodes were tested in terms of feeding type, trophic level, and total abundance. Predictors included city, urbanization level, HCN frequency, and all possible interactions. Dry weight of roots used for nematode extraction was included in the models as a covariate. A negative binomial distribution was used to account for overdispersion of the count data. Similarly, the influence of urbanization level, HCN frequency and their interaction on NCR was tested by LMs. Afterwards, to better illustrate the interactions among the city effect and those of other predictors, we examined how urbanization, HCN frequency and their interaction affected nematode abundance and NCR within each city, performing a similar series of GLMs and LMs, respectively, but leaving city out as a predictor. To meet the assumptions of normality of the model residuals, NCR was square transformed while analyzed across cities and within the city of Ghent.

Nematode community composition
To detect differences in feeding-guild composition of rootassociated nematode community (expressed as abundance per gram of dry root weight) between the cities of Antwerp and Ghent, we conducted a permutational multivariate analysis of variance (PERMANOVA,9,999 permutations) with city as only predictor, based on Bray-Curtis dissimilarity distances using the function 'adonis2' from the package 'vegan' . We tested for multivariate homogeneity of dispersions via the function 'betadisper' ('vegan' package). Afterwards, to gain insights into the effects of urbanization and HCN frequency on nematode community composition within each city, a set of type II PERMANOVAs was conducted for each of the seven spatial scales. As the interactive effect between urbanization and HCN frequency was not significant for each city, only their main effects were retained in the final models. Finally, the patterns of the nematode community composition across cities and within each city were visualized using separate non-metric multi-dimensional scaling (NMDS) ordination, and urbanization and HCN frequency were overlaid on NMDS plots within each city via the function 'envfit' ('vegan' package).

Variation in HCN frequencies along the urbanization gradients
The frequency of HCN in the two cities consistently and significantly decreased with increasing urbanization levels measured within a radius of 800 m (χ 2 = 17.9, p < 0.001), but their slopes were not significantly different (χ 2 = 1.78, p = 0.182; Figure  1; Supplementary Table S1). On average 24% of Antwerp clover plants versus 18% of plants in Ghent expressed HCN. The difference between intercepts of both cities was statistically significant (χ 2 = 4.52, p = 0.034; Figure 1; Supplementary Table S1).

Characteristics of the root-associated nematode community
The main effect of city as well as its two-way and three-way interactions with urbanization and HCN frequency were significant for most variables, especially for root-feeding and total nematode abundance (Table 1), indicating that their patterns were specific to each city. In Antwerp, abundances of plant-feeding nematodes and total nematodes decreased with an increasing HCN frequency, and this decrease for plant feeders was only significant in locations at low urbanization levels (Figures 2A,B; Table 1).
In Ghent, the abundance of PF ( Figure 2C), FF ( Figure 2E) and total nematodes ( Figure 2H) showed similar responses to urbanization gradients and HCN frequencies (Table 1). Their abundance Frontiers in Ecology and Evolution 05 frontiersin.org significantly decreased with increasing frequency of HCN in more urbanized locations, whereas it changed very little in less urbanized areas. However, abundance of BF ( Figure 2D), TL2 ( Figure 2F), and TL3 ( Figure 2G) in general, did not vary significantly with HCN frequency but was greater at higher levels of urbanization (Table 1). Note that the patterns of BF and TL2, too, represented a decrease in abundance with increasing HCN frequency only in more urban sites, even though the interaction between the effects of urbanization and HCN frequency was not significant here. Except for FF across cities and within Ghent, dry weight of roots for nematode extraction did not affect nematode abundance (Table 1). The trend of NCR along the urban-rural gradient depended on the city (Supplementary Table S2). In Antwerp, NCR significantly decreased with increased urbanization levels and was independent of HCN frequency ( Figure 3A; Supplementary Table S2). In Ghent, neither urbanization level nor HCN frequency affected NCR among clover populations ( Figure 3B; Supplementary Table S2).

Discussion
We assessed clines of HCN frequency in T. repens along urbanrural gradients in the cities of Antwerp and Ghent and compared the relationships among urbanization, HCN frequency, and the functional structure of nematode assemblages associated with the clover roots. As expected, HCN frequency decreased along a gradient of increasing urbanization in both cities. However, the decline in Antwerp, which is the more urbanized and larger, was not significantly stronger than that in Ghent. We mainly observed negative relationships between HCN frequency and nematodes from the most urbanized clover populations in Ghent yet from the less urbanized populations in Antwerp. Conversely, predators and omnivores together were more abundant in urbanized sites in Ghent. Urban soils in Antwerp were less bacterially dominated as indicated by a lower nematode channel ratio.

Patterns of clines in cyanogenesis
The consistent decrease in cyanogenesis in white clover populations along gradients from rural to urban environments in Antwerp and Ghent was in line with the trend in 47% of cities throughout the world (Santangelo et al., 2022), confirming that urbanization is driving the evolution of clover populations. Besides, Santangelo et al. (2022) concluded that drought stress and vegetation cover were the major environmental factors driving the strength of clines. Due to the urban-heat-island effect, urban areas at such temperate latitudes are often warmer and drier than rural sites throughout the year (Merckx et al., 2018). Considering that cyanogenesis functions as an antiherbivore defense, an alternative explanation is that herbivory may play a role in generating urban-rural clines in HCN. As a low frequency of HCN production was found in urban clover populations, these clover populations may either experience less herbivory, if herbivores control HCN production, or more herbivory, if herbivores select undefended plants. Although  did not establish a causal link between leaf herbivory and HCN along an urbanization gradient in the city of Toronto, they highlighted that the role of herbivory in driving HCN clines should be further investigated.
In contrast with our hypothesis, the parallel declines along rural-urban gradients in both cities were statistically equally steep. Environmental changes such as impervious surface along ruralurban gradients adopted by this study were more parallel (Santangelo et al., 2022). Given the proximity of Antwerp and Ghent, other local environments rather than impervious surface itself are more likely to regulate the strength of the observed HCN clines. Even though the extent of urbanization in Antwerp is larger than in Ghent, this difference was apparently not strong enough to cause a significant difference in micro-environments and hence in HCN clines.

Root-associated nematode community composition
We frequently observed interactions between urbanization and HCN frequency in their relationship with the abundance of different feeding groups of root-associated nematodes. This highlights the important roles of both plant defense and urbanization in the trophic structure of these nematode communities.  Dry weight of roots (DRW) for nematode extraction was included as a covariate. Nematodes were tested in terms of the abundance of feeding types and trophic levels and of total abundance. Tested nematode feeding types include plant feeders (PF), bacterial feeders (BF), fungal feeders (FF) and omnivores (OM). Predators (PR) were rare across all samples, and hence excluded from the test. Tested nematode trophic levels comprised the first (TL1), second (TL2) and third (TL3) trophic levels. TL1 equals to plant feeders; TL2 is the sum of BF and FF; TL3 is the sum of PR and OM. Total nematode abundance is the sum of all feeding types. Numbers of samples in Antwerp and Ghent are 24 and 27, respectively. For each variable, model R 2 at the tested spatial scale, estimates (ES) of predictors and their χ 2 and p values from type-III Chi 2 Wald test are presented. P-values lower than 0.05 are marked in bold.
Frontiers in Ecology and Evolution 07 frontiersin.org

Urbanization effect
The effect of urbanization on the functional composition of rootassociated nematode communities was highly city-specific in our study. City size determines most properties of the city, such as the extent of urbanization and the characteristics of local environments (Bettencourt et al., 2007;Uchida et al., 2021). The larger the city, the stronger the intensity of urbanization and the more complex the environmental heterogeneity. The feeding-guild composition of rootassociated nematode communities was influenced by urbanization only in Antwerp. We speculate that this could be due to pollutants (Li et al., 2011), which are more common in relatively larger cities (Pavao-Zuckerman and Coleman, 2007). The shift in root-associated nematode community structure in Antwerp was possibly caused by the slightly larger dispersion of different feeding guilds, rather than the responses of individual feeding guilds to urbanization. Related environmental factors such as soil water content plays a key role in nematode assemblages (Amossé et al., 2016) but its role is highly guild-specific (Vandegehuchte et al., 2015). Besides, our findings of increased overall abundance of predators and omnivores (TL3) with increasing urbanization levels in Ghent indicate that urbanization can modify the functional composition of the root-associated food web through changes in abundances of trophic groups. This is contrary to our hypothesis (4) based on the general rule that nematodes at higher trophic levels are more sensitive to urban soils (Georgieva et al., 2002;Pavao-Zuckerman and Coleman, 2007;Li et al., 2011). A plausible explanation could be that these nematodes track the increased abundance of microbivorous nematodes (TL2) with urbanization, which are a main part of their diet. This would suggest that bottom-up Root-associated nematode abundance as a function of urbanization gradient and frequency of HCN in the city of Antwerp (A,B) and Ghent (C-H). Negative binomial generalized linear models were fitted, with dry weight of roots for nematode extraction as a covariate. Nematode abundances were tested regarding feeding types, trophic levels and total nematodes. Trophic level 2 is the sum of bacterial and fungal feeders; Trophic level 3 is the sum of predators and omnivores. Nematode abundance of tested groups whose models showed significant effects of at least one factor at p < 0.05 (see the detailed results, Table 1) is plotted against frequency of HCN, for a mean value of root dry biomass. To illustrate the interaction effect of urbanization and frequency of HCN, nematode abundance at low (the first quartile value, green lines) and high (the third quartile value, red lines) levels of urbanization is predicted. Similarly, data points with urbanization values lower than the first quartile, between the first and third quartile, and higher than the third quartile are indicated in green, blue, and red, respectively. The scale of urbanization at which nematode abundance was best explained by urbanization and frequency of HCN is shown in between parentheses.
Frontiers in Ecology and Evolution 08 frontiersin.org forces outweigh any top-down disturbance effects on these higher tropic levels. Urbanization had mixed effects on the nematode channel ratio. In Antwerp, root-associated food webs of white clover switched to a fungal-dominated decomposition system in more urbanized areas, which contrasts with studies reporting a less fungal dominance in urban soil food webs (Pouyat et al., 1994;Pavao-Zuckerman and Coleman, 2007). However, the nematode channel ratio did not vary across the urban-rural gradients in Ghent, which was in line with the work conducted by Li et al. (2011). Soils exposed to heavy industrial pollution have been shown to be dominated by bacterial-based decomposition processes (Pen-Mouratov et al., 2010), which suggests that urban soils in Antwerp did not experience heavy pollution, probably because industries were mainly close to ports of rural sites. In addition, frequent fertilization in agriculture and grassland ecosystems in rural areas can stimulate the growth of bacteria (Bittman et al., 2005) and, hence increase the number of bacterivorous nematodes (Hu and Qi, 2010;Pan et al., 2010), which could explain the higher NCR observed in less urbanized locations in Antwerp. Moreover, frequent and intensive mowing in more urbanized sites increases leaf litter in the soil and the resulted higher amount of decaying residues could support a higher fungal biomass and fungivorous nematodes (Nakamoto and Tsukamoto, 2006;Deguchi et al., 2007). Finally, unmeasured variables could have covaried differently with the urbanization gradients in Ghent and Antwerp, and resulted in the different responses of NCR to urbanization in both cities.

Plant defense
In line with our hypothesis (2), HCN frequency reduced the abundance of most of functional and trophic groups of nematodes in rural areas of Antwerp and in urban areas of Ghent, which suggests that the feeding-guild patterns of root-associated nematode communities could be partially attributed to plant chemical defense. HCN can have negative impacts on root feeders (Dutta et al., 2019). The components of cyanogenesis in white clover are also synthesized in shoots, but roots are not cyanogenic (Collinge and Hughes, 1982). As there is no evidence that HCN in shoots of white clover can be directly transported to roots, the possible effects of HCN on soil biota may occur when the defensive chemicals in the leaves of white clover could be transported into soil through degradation of litter. In this case, the decreased abundance of herbivores observed in both cities may have been the indirect consequence of such litter effects. We further speculate that in Ghent, the HCN-containing leaf litter may have decreased the abundance of bacteria and fungi in the rhizosphere, causing a decline in the abundance of bacterial and fungal feeders, or have directly affected these two nematode feeding types via fumigation with HCN (Dutta et al., 2019). Lower fungal and bacterial feeder abundances may in turn have caused predators and omnivores to increasingly feed on plant feeders. This hypothesis would need further experimental validation, and does not address the question why nematodes in Ghent seem to react most strongly to HCN in urban sites where the proportion of HCN-producing plants is lowest, in contrast with our hypothesis (3). Perhaps nematode communities become less sensitive to HCN through adaptation if it is returned to soil in larger amounts. Nevertheless, some evidence suggest that the quality of plant litter can drive belowground communities. For example, stem litter of the invasive Spartina alterniflora stimulated the growth of nematodes, particularly bacterivores, due to its higher quality than that of the native Phragmites australis (Chen et al., 2007). Besides, S. alterniflora that decayed faster had lower values of structure index and maturity index than P. australis, which means nematode communities around the former plants were dominated by colonizers with shorter longevity, smaller size and more sensitive to disturbance. This is in accordance Linear relationship of nematode channel ratio with urbanization (BUC) within a 200-meter radius in the cities of Antwerp (A) and Ghent (B). The effects of predictors are presented in Supplementary Table S3. Nematode channel ratio is plotted against BUC values and the size of dots corresponds to the proportion of HCN-producing plants. The fitted line for the significant predictor of BUC at p < 0.05 is based on its modeled intercept and slope.
Frontiers in Ecology and Evolution 09 frontiersin.org with the observation that labile litter induced a more bacterially dominated energy channel than recalcitrant litter (Vauramo and Setälä, 2010). Therefore, the litter input with high quality may cause bacterialbased and active belowground biota communities. However, in Antwerp urban white clover with low HCN frequency and correspondingly high-quality litter harbored fungal-dominated communities. This is not in line with the idea that HCN is more antibacterial than antifungal, because the activity of the cyanideinducible enzyme cyanide hydratase, which can detoxify HCN, appears to only occur in fungi not in bacteria (Fry and Evans, 1977;Wang et al., 1992). It thus seems that HCN frequency and NCR in Antwerp were driven by one or more unmeasured factors associated with urbanization.

Conclusion
This study showed that the relationship between urbanization and root-associated nematode feeding-guild composition was modified by the presence of a plant defense molecule, most likely via changes in the quality and palatability of leaf litter entering the soil system. Therefore, these results confirm that nematodes are good bioindicator of environmental changes. Although it is challenging to identify the specific urbanization drivers shifting nematode communities, we suggest that future studies incorporating plant traits and relevant environmental variables will contribute to interpreting the consequences of urbanization for soil communities and the ecosystem functions and services they provide. Nematode feeding-guild composition in the roots of Trifolium repens across the cities of Antwerp and Ghent and environmental factors for each of them. (A) Non-metric multidimensional scaling (NMDS) plot showing compositional difference of nematode feeding guilds between Antwerp and Ghent. Explanatory variables of urbanization gradient (BUC) within the radius of 1,600 meter and frequency of HCN production (HCN frequency) are superimposed for Antwerp (B) and Ghent (C). Sample size for Antwerp and Ghent is 24 and 27, respectively. Nematode feeding guilds are bacterial feeders (BF), fungal feeders (FF), plant feeders (PF), predators (PR) and omnivores (OM). Effects of both environmental variables on the nematode community composition were evaluated by PERMANOVA (Supplementary Table S3), and significant and non-significant effects are indicated in solid and dashed lines, respectively. The stress levels of NMDS 2D plots are shown.
Frontiers in Ecology and Evolution 10 frontiersin.org

Data availability statement
The original contributions presented in the study are included in the article/Supplementary material, further inquiries can be directed to the corresponding author.