https://doi.org/10.5061/dryad.fxpnvx0zh

Materials and Methods

Study area
Our study focused on urban A. thaliana populations in the city of Cologne, situated in midwestern Germany (about 50.9°N, 7.0°E). Cologne is Germany’s fourth most populous city, with more than one  million inhabitants (Ansmann et al., 2021). One of Germany's warmest cities, it has a temperate oceanic climate with a mean annual temperature of 11.7 °C during the day and 6.3 °C at night, and a mean annual precipitation of 840 mm.
Field surveys, phenology monitoring, and functional trait measurements

We surveyed eight A. thaliana populations within a 4.3 x 3.7 km area in southwestern Cologne. Study populations were selected based on their accessibility and on their isolation from other populations, with all habitat patches located at least 0.6 km from one another. Urban habitat patches where viable A. thaliana populations were observable included green spaces such as disturbed lawns; various habitats within gray spaces, such as sidewalks, wall tops, and bases; and vegetated roadsides (Figure 1a; Figure S1; Table S1). Depending on habitat characteristics, patch size ranged from 4 m2 for populations on wall tops to > 100 m² for populations in vegetated roadsides, and population size ranged from 15 to 1,000 individuals. All populations were studied throughout their growth period in 2017 (which extended between early March and the end of June), and revisited yearly during five subsequent growth periods (2018-2022) to assess population persistence. No field permissions were necessary for the collection of the plant samples in this study.
We recorded phenological stages and fitness-related plant functional traits of A. thaliana individuals distributed over the eight habitat patches. At each patch, we established 10 to 14 observation quadrats (10 x 10 cm), each at least 50-60 cm from one another. We marked from 1 to 3 A. thaliana individuals per quadrat (15-24 per habitat patch; 174 in total) and visited them twice a week during the study period to score their development. For all marked individuals, we recorded the day of bolting (i.e., when the inflorescence became visible) and, subsequently, the number of open flowers and the number of closed fruits. Rosette diameter and the number of rosette leaves were measured at the time of bolting. We used these observation data to derive five phenological stages and five fitness-related plant functional traits (“fitness proxies”; see Table S2). The selection of plant functional traits follows recommendations for the assessment of plant fitness (Gibson, 2014), while phenological stages were selected to represent life-history stages of the entire growth period, as recommended for herbaceous species (Huang et al., 2018; Nordt et al., 2021). Germination was generally not possible to determine in situ, except in populations growing on wall tops (SGY, KAD, HLU).

Assessment of environmental conditions of habitat patches
Abiotic environmental conditions and characteristics of disturbance regime were recorded at the  level of habitat patches, making use of the indicator function of plant communities (Kollmann & Fischer, 2003). We used species’ Ellenberg indicator values, EIVs (Ellenberg & Leuschner, 2010), as a proxy for abiotic conditions. EIVs, which express the optimal positions of Central European plant species along gradients of abiotic factors, provide information about species’ niche and habitat requirements (Diekmann 2003), in both rural and urban habitats (Fanelli et al. 2006). EIVs are expressed on an ordinal scale ranging from 1 to 9 and refer to light regime (L), temperature (T), and continentality of climate (K). Edaphic conditions are captured as soil moisture (F), pH (R), and nutrient availability (N). Likewise, species’ disturbance indicator values (DIVs) (Herben et al., 2016), were used to characterize the disturbance regimes of urban habitat patches. To this end, we extracted species-level indicators for disturbance severity (DS), disturbance frequency (DF), and disturbance effects on vegetation structure (DV) from the data compiled in Herben et al., 2016. DIVs are based on vegetation-plot records of the Czech flora, with a species’ DS defined as the mean disturbance severity of all vegetation classes weighted by its occurrence frequencies in these classes; DF calculated as the mean of logarithmic disturbance frequency of all vegetation classes weighted by a species’ occurrence frequencies in these classes; and DV calculated from vegetation structural parameters by summing covers and community-weighted means of all species’ growth height per vegetation class. We determined the composition of plant communities in autumn (late August and early September) 2017 by recording the presence of all vascular plant species in the herbaceous layer of the habitat patches. Based on species’ specific indicator values (see Table S4) and excluding A. thaliana, weighted average EIVs and DIVs were calculated for the plant community of each habitat patch.

Assessment of genomic variation
At the end of the 2017 growth period, we harvested seeds from 5 to 17 individuals per population and used single seeds from each individual to produce progenies. For four of the eight habitat patches (BER, BIS, KAD, and MIL), we also included seeds from individuals sampled in the year prior to field data collection, referring to these as population samples (Table S3). In order to broaden our view of the genomic diversity of A. thaliana in Cologne, we also produced progenies from seeds collected at 30 additional habitat patches distributed throughout the sampling area but not included in the in situ monitoring (hereafter called scattered samples). Field collected seeds were grown in a randomized setting, under conditions simulating winter, i.e., in growth chambers with 10 h light at 18 °C / 14 h dark at 16 °C, 60 % humidity, for 6 weeks; followed by 8 h light/16 h dark at 4°C, 80 % humidity, for 6 weeks; and ultimately ripened in the greenhouse in spring-like conditions with 16 h light, at about 20-24 °C / 8 h dark at about 18-20 °C until harvest. Seeds for progeny were harvested using Aracons (Betatech, Gent, Belgium).
Three leaves per plant were sampled for DNA extraction, deep frozen, and homogenized using a Precellys Evolution homogenizer (Thermo Fisher Scientific, Waltham, MA, USA). DNA was isolated using the NucleoSpin® Plant II kit (Macherey-Nagel, Düren, Germany). Genomic DNA was prepared for RAD-Seq as described by (Dittberner et al., 2019) using 10 pools with 20 individually barcoded plant samples each, and sequencing was performed at the Max Planck Institute of Plant Breeding Research. The bioinformatics pipelines used for read mapping, and identification of each of the 12 genotypes present in the studied habitat patches, population structure and admixture are described in the supplementary information.

Common garden experiments:
For each identified genotype, two individuals were randomly selected and amplified in the greenhouse. Since they were raised in the same maternal environment, variance among genotypes can be assigned to genetic effects and variance among replicate lines of each genotype allows testing whether all maternal effects were effectively removed when growing parents in common garden conditions. We then took eight individuals from each of these two lines and grew them in growth chambers (Johnson Controls, Milwaukee, WI, USA) in three trials under different growth conditions. Growth conditions were chosen to identify genetic variation in the major pathways regulating flowering time: constant long days (16 h light at 20°C/8h dark at 18°C); constant short days (8 h light at 18°C day/16 dark at 16°C night); and short days/vernalization/long days (8 h day light at 18°C /16 h dark at 16°C for 5 weeks; 8h day light/16h dark at 4 °C for 4 weeks, 16 h day light at 20°C/8h dark at 18°C until ripening). The A. thaliana genotypes Col-0 was included as a reference in all experiments. Because we did not detect differences between replicated lines, we considered the number of replicates to be 2x8= 16 replicates/genotype. Replicates were sown in individual 6-cm-diameter pots that were watered regularly by flooding the trays. Flowering time was determined as the time span between sowing and the first flower to open petals.

Ripe seeds from the short-day/vernalization/long-day experiment were harvested and stored at room temperature for the germination assay. Germination rates were determined after three months to quantify primary dormancy (Baskin & Baskin, 2004). Seeds were sown on wet filter paper and incubated in closed microtiter plates. Before incubation under long-day conditions (16 h light at 20°C/8h dark at 18°C), seeds were incubated in the dark either at 4 °C for 7 days to quantify their capacity to germinate when dormancy is released, or at -21 °C for 4 d and at 35 °C for 4 d to quantify variation in secondary dormancy. Germination was scored after 10 days.

To quantify genetic variation under outdoor field conditions, we conducted a set of common garden experiments. For this, seeds from 4 independent replicate lines per genotype were sown directly at a density of 8 seeds per pot. In total, we planted 32 pots per genotype and randomized 9-cm- diameter pots at a field site in the botanical garden of Cologne University. Again, because we found no significant differences between replicate lines, we determined that the experiment was complete with 32 replicates per genotype. Seeds were sown in late summer (August 23) and in early autumn (September 20), mid-autumn (November 8), and late winter (February 26). Rather than being watered artificially, pots were put on a fabric that retained water after rainfall. The site was protected against birds and rabbits by a net. Plants were inspected for germination twice a week. Flowering time was scored when the first flower showed open petals. For fertility measurement, siliques were counted at the end of each plant’s life cycle.
Phenotypic data was analyzed using the glm function in R. We used one-way ANOVA with habitat patch as the fixed factor to analyze differences in environmental conditions (EIVs and DIVs of plant species co-occurring with A. thaliana); and in the phenological stages and fitness proxies of A. thaliana plants observed in situ. Measurements of flowering time in common garden experiments were analyzed with two-way ANOVA, using genotype and cohort or growth condition as fixed factors.

For germination assays under controlled indoor conditions, the number of germinants per Petri dish was analyzed with the R function zeroinfl to correct for zero inflation. For germination rates in the outdoor common garden experiment, we proceeded as under controlled conditions, except that we used a quasipoisson distribution of error. Because the two replicate lines we used for each genotype never showed significant effects, the line effect was not included in any of the models. We used the function ANOVA with a Chi-squared test to determine the overall significance of effects. In all cases, the significance of pairwise differences was tested with a Tukey post-hoc test, using the function glht from the multcomp package in R. We used the Pearson coefficient of the cor.test function in R to quantify the correlation between phenotypes.