Pollen dispersal patterns differ among sites for a wind-pollinated species and an insect-pollinated species

MATERIALS AND METHODS We assessed pollen dispersal patterns of A. tuberculatus at one roof and one ground site in 2014 and of both A. tuberculatus and S. lycopersicum at two roof and two ground sites in 2015. Because methods and results were similar between the two years, and the 2015 study was more comprehensive, only methods and results for 2015 are presented here. A summary of the results from 2014 can be found in Appendix S1 (see the Supplementary Data with this article) and the results from the 2014 analysis can be found in Appendix S2. Study species – Amaranthus tuberculatus was chosen as the wind-pollinated species because amaranths grow well in urban areas (Del Tredici, 2010; Aloisio et al., 2016), it is dioecious (Hager et al., 1997), and there have been no individuals reported in the counties where our study sites were located (Moore et al., 2012; USDA NRCS, 2017). Seeds from both an Iowa, USA population (Hinz and Owen, 1997) and a Mississippi, USA population (Nandula et al., 2013) of A. tuberculatus were used in the 2014 experiment. In the 2015 experiment, only plants from the Iowa population were used due to low germination in the Mississippi population seeds. We chose S. lycopersicum as the insect-pollinated species because members of the Solanaceae grow well in urban environments (Del Tredici, 2010). S. lycopersicum is most commonly buzz-pollinated (Buckmann and Hurley, 1978) by bee species (e.g. bumble bees, sweat bees; Teppner, 2005), although bees that do not buzz pollinate (e.g. honey bees), have been shown to chew through the anther cone to obtain pollen (Deprá et al., 2014). Furthermore, although S. lycopersicum typically self-fertilizes, the NC4 S. lycopersicum breeding line (Panthee and Gardner, 2013) produces male-sterile and male-fertile offspring. Also, the ms-10 male-sterility gene is tightly-linked to the aa gene for stem color, with 90% of male-sterile plants having green stems and 90% of male-fertile plants having purple stems. As a result, we were able to separate most male-sterile and male-fertile plants prior to the onset of flowering and pollen production. Greenhouse methods – Seeds were germinated and seedlings were grown in a greenhouse at the Louis Calder Center (Fordham University) field station in Armonk, New York, USA during the spring of 2015 ((S. lycopersicum – April/May; A. tuberculatus – May/June), following Butcher et al. (2020). Briefly, seeds were germinated in 53 x 28 x 5 cm (l x w x d) potting trays (Griffin Greenhouse Supplies, Tewksbury, Massachusetts, USA) containing Sungro Sunshine Soil-mix 4 (Sungro Horticulture, Agawam, Massachusetts, USA) supplemented with Osmocote 14-14-14 fertilizer (Scotts Miracle-Gro, Marysville, Ohio, USA). Trays were bottom-watered every other day with well water until the soil was saturated. Seedlings were transplanted to 10-cm dia (0.57 L) pots (Griffin Greenhouse Supplies, Tewksbury, Massachusetts, USA) when all seedlings had reached 5 to 10 cm tall (measured from soil to apical meristem). Seedlings were grown in Sungro Sunshine Soil-mix 4, and watered to throughflow with well water every other day. Before onset of flowering, S. lycopersicum were separated into pollen receptors and donors based on stem color. Prior to placing the plants at study sites, pollen receptors and donors of both species were confirmed by flower morphology, inflorescences were removed to exclude seeds or fruits produced as a result of pollination in the greenhouse, and plants were transplanted to 11.4 L pots (Griffin Greenhouse Supplies, Tewksbury, Massachusetts, USA) with Sungro Sunshine Soil-mix 4 (Sungro Horticulture, Agawam, Massachusetts, USA). These plants served as the parent plants in the study. We collected leaf tissue (1-2 young leaves) from each plant for genotyping. Study sites – Parent plants were placed in experimental arrays at four sites in the New York metropolitan area, which is located in the northeast USA: 1) Fordham University’s Rose Hill Campus (Rose Hill) in Bronx, New York, USA, 2) the Javits Convention Center (Javits) in New York, New York, USA, 3) Fordham University’s Louis Calder Center (Calder) in Armonk, New York, USA and 4) the Queens Zoo (Queens Zoo) in Corona, New York, USA (Fig. 1). At each site, 12 pollen donor plants of each species (i.e. male A. tuberculatus plants and male-fertile S. lycopersicum plants) were placed in the main sector of each site (pollen donor group; Fig. 2). A. tuberculatus and S. lycopersicum pollen receptor plants were also placed in the main sector, in groups of two, by species, starting at 1 m and at then at increasing distances from the pollen donor group (Table 1; Fig. 2). Additionally, A. tuberculatus and S. lycopersicum pollen receptor plants were placed in groups of two, by species, in up to three nearby non-contiguous sectors. The number of pollen receptor plants, the maximum distance between the pollen donor group and pollen receptor plants, and the number of non-contiguous sectors depended on the species, as well as the layout of each site (Table 1). The experimental array layouts for both species at the Rose Hill site are shown in Fig. 3 as representative of the layout of arrays at one site. At the Rose Hill site, a college campus, the main sector and one of the non-contiguous sectors were located on two separate portions of the roof of a campus building, while the second non-contiguous sector was located on the roof of Rose Hill’s parking garage (Fig. 2A). The third non-contiguous sector was a ground-level garden site at the New York Botanical Garden (Fig. 2A). Surrounding areas at the Rose Hill site included urban lawn areas and additional buildings of Fordham’s Rose Hill campus, the New York Botanical Garden, and residential and commercial spaces. At the Javits site, a large commercial building, the main and non-contiguous sectors were located on two separate portions of the Javits Center’s Sedum-dominated green roof (Fig. 2B). The Javits site was surrounded by the Hudson River, the West Side (Rail) Yard, and residential and commercial spaces. At the Calder site, the main sector and one of the non-contiguous sectors were located in suburban lawns which were regularly mowed, while the second non-contiguous sector was located in an unmowed meadow (Fig. 2C). Surrounding areas at the Calder site included forested regions within and adjacent to the Calder Center and suburban residential spaces. Finally, at the Queens Zoo site, a zoological garden, the main sector and one of the non-contiguous sectors were located in urban lawns that were mowed frequently, while the second non-contiguous sector was located in an unmowed meadow (Fig. 2D). The Queens Zoo site was surrounded by urban green spaces (e.g. lawns, gardens) located within the Queens Zoo and residential and commercial spaces. Field methods – Site configuration was chosen assuming that pollen would dispersal evenly in all directions (i.e. isotropic dispersal) because the exponential distribution, which is the relationship we assumed during our statistical analyses, assumes isotropy (Austerlitz and Smouse, 2001). Thus, the design did not include pollen receptor plants at matching distances in each direction. Because S. lycopersicum plants required vertical support, they were staked with 0.91 m Bond bamboo stakes (Home Depot, Atlanta, Georgia, USA). Further, because Calder had a sizable white-tailed deer (Odocoileus virginianus) population, the parent plants of both species were caged with 137 cm tall Gilbert and Bennett tomato cages wrapped with 2.54 cm weave Everbilt poultry netting (Home Depot) to minimize herbivory. Pollen dispersal studies have used poultry netting to allow insect pollinator visits, but exclude herbivores or large pollinators (e.g. Price and Waser, 1979; Richards, 2000; Epps et al., 2015). Further, Wratten et al. (2003) found that fencing did not act as a barrier to hover flies, which are similar in size to the buzz-pollinating bee species that most frequently pollinate S. lycopersicum (e.g. bumble bees, sweat bees; Teppner, 2005). Flannery et al. (2004) also used poultry netting in a study of pollen dispersal in Brassica napus, which is both wind and insect-pollinated, suggesting that this herbivore barrier does not impede wind-dispersed pollen dispersal. Therefore, we assumed that the poultry netting would not impede pollen dispersal in our study. The GPS coordinates for the pollen donor group and each pair of pollen receptor plants were collected and the straight-line distance between each pair and the donor group was calculated using the measure tool in ArcMap version 10.3 (ESRI, 2014). The calculated distance was then used as the effective pollen dispersal distance. Additionally, we estimated percent vegetation cover and percent impervious surface for a 350 m radius buffer around each site in Google Earth Pro 7.3.3.7699 (Google, Inc., Mountain View, California, USA) using the circle and polygon tools (Johnson et al., 2018). The experimental arrays were maintained for eight weeks during the summer of 2015 (S. lycopersicum – June 15 through August 3; A. tuberculatus – July 15 through September 9). Plants were watered with untreated tap water three times per week to flow-through and allowed to pollinate naturally. During week 2, plants were watered to through-flow with 5 cm3 per 3.8 L tap water mixture of Miracle Grow Bloom Booster 10N-52P-10K (Scotts Miracle-Gro, Marysville, Ohio, USA) to promote flowering. Three of the 66 A. tuberculatus pollen receptor plants (one from Rose Hill, two from Queens Zoo) and two of the 94 S. lycopersicum pollen receptor plants (both from Queens Zoo) died or sustained damage to a degree that would impede pollen receipt, so they were removed from the array and analyses. At the end of the study, A. tuberculatus seeds were collected en masse and stored in 5.72 x 8.89 cm coin envelopes at room temperature (~ 24°C) until germination. S. lycopersium fruits were collected as they ripened, to minimize the effects of frugivory on estimates of pollination. S. lycopersicum seeds were then removed from fruits and placed on 125 mm diameter, grade 1 Whatman cards (GE Healthcare Life Sciences, Pittsburgh, Pennsylvania, USA), and dried and stored at room temperature (~ 24°C) until germination. Parthenocarpic S. lycopersicum fruits were excluded from the total fruit count as these fruits were not produced as the result of a fertilization event. The total number of seeds (A. tuberculatus) or fruits (S. lycopersicum) produced per pollen receptor plant served as the proxy of effective pollen dispersal (i.e. successful pollination and gene flow). This proxy for effective pollen dispersal has been used in several studies (e.g. Sahley, 2001; Albrecht et al., 2009; Collevatti et al., 2010; Baldoni et al., 2017; Young et al., 2018). Additionally, although pollen dispersal has not been investigated in S. lycopersicum, seed production as a proxy for effective pollen dispersal has been used in studies on A. tuberculatus pollen dispersal (Liu et al., 2012; Sarangi et al., 2017). Paternity assignment – Molecular techniques used for paternity assignment followed Butcher et al. (2020), except that A. tuberculatus hypocotyl and cotyledon tissue was incubated with AP1 buffer and RNase A (Qiagen, Hilden, Germany) for 2 hrs after disruption. In brief, we extracted DNA from maternal and paternal leaf tissue, as well as from seeds produced during the experiment. We then performed PCR of two markers for A. tuberculatus: an external transcribed spacer region marker (ETS) and a putative DEAD box ATP-dependent RNA helicase gene marker (PutDead). We also performed PCR of two markers for S. lycopersicum: a spotted wilt virus disease resistance gene marker (Sw-5) and a Fusarium wilt race 3 virus resistance gene marker (I-3). For A. tuberculatus, PCR products were Sanger sequenced by Genewiz, Inc (South Plainfield, New Jersey, USA) or Macrogen, Inc (Rockville, Maryland, USA). S. lycopersicum resistant and susceptible alleles could be differentiated based on allele size by running the PCR product on a 3% agarose gel, thus determining paternity without the need for Sanger sequencing. For the A. tuberculatus pollen receptors that produced at least one seed but fewer than 20 seeds (Rose Hill – 1 pollen receptor, Calder – 9 pollen receptors, Queens Zoo – 2 pollen receptors, Javits – 1 pollen receptor), we genotyped all of the seeds that germinated. For pollen receptors that produced more than 20 seeds, we randomly selected 20 seeds to genotype. Seeds were germinated on damp filter paper at 37C, and DNA was extracted from about 4 cm of hypocotyl and cotyledon tissue. Paternity could then be assigned using simple exclusion to confirm that pollen dispersed from the pollen donor group and not a local pollen donor (similar to Ellstrand, 1984). This process involved subtracting the known maternal genotype from the seed genotype, then comparing the seed’s paternal genotype to the pollen donor group genotype(s). If this genotype matched one of the pollen donor genotypes, we concluded that the pollen donor was from the pollen donor group. For the S. lycopersicum pollen receptors, we genotyped two randomly chosen seeds from each fruit produced by the receptors at the two distances furthest from the donor group, and two randomly chosen seeds from two randomly chosen fruits produced by pollen receptors at two additional distances chosen at random. All plants in the arrays carried the two resistant alleles, while plants in local populations should carry only the susceptible alleles (R. Gardner, North Carolina State University, pers. comm.). As a result, we could identify S. lycopersicum seeds fertilized by a local pollen donor, as these seeds would exhibit both the resistant and susceptible alleles in the PCR product. We genotyped more than one seed per fruit because, although multiple paternity within a single fruit has not been observed in Solanum, it has been observed in other species (e.g. Ipomopsis aggregata – Campbell, 1998; Silene latifolia – Teixeira and Bernasconi, 2007). For both sites where we found evidence of fertilization by a local pollen donor, we also assigned paternity to two randomly chosen seeds from two randomly chosen fruits from every pollen receptor at that site to assess the degree of pollen dispersal from local pollen donors. All tested seeds were germinated and grown in the greenhouse in 10-cm diameter (volume: 0.57 L) pots with Sungro Sunshine Soil-mix 4 and bottom watered twice per week with well water until the appearance of the first true leaf, which was harvested for DNA extraction. Of these 803 seeds genotyped for the A. tuberculatus paternity assignment, we were unable to produce readable sequences at the ETS marker for 63 seeds (Rose Hill – 43 seeds, Calder – 2 seeds, Queens Zoo – 9 seeds, Javits – 9 seeds) and at the PutDead marker for five seeds (three from Javits, one from Rose Hill, and one from Calder). Therefore, we cloned a subset of these samples to obtain readable sequences. First, we purified PCR products using a QIAquick PCR Purification Kit (Qiagen). We then ligated the purified products (i.e. insert) into a pGEM-T vector and transformed the vector containing the insert into JM109 cells using a pGEM®-T Easy Vector II Kit (Promega, Madison, Wisconsin, USA). An insert disrupts the lacZ α-peptide coding sequence in the vector, so no functional -galactosidase is produced; therefore, insert presence was determined by the appearance of white colonies in blue/white bacterial colony screening. Several white colonies per sample (minimum: 7, maximum: 18) were sequenced by Genewiz, Inc (South Plainfield, New Jersey, USA). All cloned samples matched an expected parent genotype. Consequently, all 68 seeds that initially produced unreadable sequences were included in statistical analyses. Statistical analyses – Statistical analyses to test our first hypothesis were performed using SYSTAT version 13 (Systat Software, San Jose, California, USA), while the analyses related to our second hypothesis were conducted using the lm function in R version 3.6.1 (R Core Team 2019). All results were compared to an alpha value of 0.05. Previous research suggests the relationship between A. tuberculatus pollen dispersal and distance follows an exponential decay pattern (Liu et al., 2012; Sarangi et al., 2017), and this pattern also has been observed in other wind-pollinated herbaceous species (Walsh et al., 2015; Dong et al., 2016; Chang et al., 2018). Although pollen dispersal has not been examined in S. lycopersicum, Beckie and Hall (2008) observed that studies modeling the effect of distance from the pollen donor on pollen dispersal in crop plants (e.g. maize, wheat) generally fit exponential decay or inverse power curves to the relationship. Therefore, assuming an exponential decay relationship, we log-transformed the number of seeds or fruits produced per plant and distance prior to performing the analyses (Motulsky and Ransnas, 1987; Kutner et al., 2004). By linearizing these data, we were able to use simple linear regression to test our first hypothesis and ANCOVA to test our second hypothesis (Barbour et al., 2005). Residual plots were checked for normality and homogeneity of variances. Log-transformation of the data improved normality and homogeneity of variances, compared to raw data. After log-transformation of the data, the A. tuberculatus analyses met the assumptions of normality and homogeneity of variances at all sites. However, the S. lycopersicum regression and ANCOVA analyses did not meet the assumptions of normality and homogeneity of variances for the Rose Hill, Queens Zoo, and Javits sites, so the results from these sites should be interpreted with caution. To test our first hypothesis, that effective pollen dispersal decreased with increasing distance for a given species at a given site we ran linear regression analyses between number of seeds (A. tuberculatus) or fruits (S. lycopersicum) produced, which served as proxies of effective number of flowers pollinated, and effective pollen dispersal distance. These regressions were performed for each species at each site, as our first hypothesis addresses the relationship between pollen dispersal and distance to the pollen donor for individual species at individual sites. To test our second hypothesis, that the effect of distance on effective pollen dispersal differed among sites, we used ANCOVA to compare the regression relationships among sites. Specifically, for each species, we used the ANCOVA interaction term (i.e. site*distance) to test whether the slopes differed among sites. If the slopes were not different at = 0.05, we ran an ANCOVA without the site*distance term to test whether dispersal at a given distance differed among sites (Kutner et al., 2004). In addition, because the degree of urbanization differed among sites, we ran a posteriori analyses that included two common measures of urbanization, % green space and % impervious surfaces, to assess the potential roles of these factors in observed differences among sites. Like the analyses for our hypotheses, these a posteriori ANCOVA analyses were also run using log-transformed data. Additionally, because the % green space and % impervious surface variables were correlated, and to avoid overfitting the models, we ran ANCOVAs including distance, site, and either % green space or % impervious surface.