Coflowering invasive plants and a congener have neutral effects on fitness components of a rare endemic plant

Abstract Network analyses rarely include fitness components, such as germination, to tie invasive plants to population‐level effects on the natives. We address this limitation in a previously studied network of flower visitors around a suite of native and invasive plants that includes an endemic plant at Badlands National Park, South Dakota, USA. Eriogonum visheri coflowers with two abundant invasive plants, Salsola tragus and Melilotus officinalis, as well as a common congener, E. pauciflorum. Network analyses had suggested strong linkages between E. visheri and S. tragus and E. pauciflorum, with a weaker link to M. officinalis. We measured visitation, pollen deposited on stigmas, achene weight and germination over three field seasons (two for germination) in four populations (two in the final season) of E. visheri and applied in situ pollen treatments to E. visheri, adding pollen from other flowers on the same plant; flowers on other E. visheri plants; S. tragus, M. officinalis, or E. pauciflorum; open pollination; or excluding pollinators. Insect visitation to E. visheri was not affected by floral abundance of any of the focal species. Most visitors were halictid bees; one of these (Lasioglossum packeri) was the only identified species to visit E. visheri all three years. Ninety‐seven percent of pollen on collected E. visheri stigmas was conspecific, but 22% of flowers had >1 grain of E. pauciflorum pollen on stigmas and 7% had >1 grain of S. tragus pollen; <1% of flowers had M. officinalis pollen on stigmas. None of the pollen treatments produced significant differences in weight or germination of E. visheri achenes. We conclude that, in contrast to the results of the network analysis, neither of the invasive species poses a threat, via heterospecific pollen deposition, to pollination of the endemic E. visheri, and that its congener provides alternative pollen resources to its pollinators.


| INTRODUC TI ON
Competition for pollinator services, recognized by Darwin, has been evaluated with respect to visitation frequency, quality of pollen received, and fitness outcomes (Spellman et al., 2015;Thijs et al., 2012) for the competing plant species. With the realization that invasive plants can have a greater impact than habitat fragmentation on flower visitation networks (Hansen et al., 2018), understanding effects of such disruption for fitness of native plants has become imperative. Detrimental effects of non-native plant invasion on native plant fitness is not a foregone conclusion, however. Studies show a range of effects from facilitative to neutral to competitive, often depending on the density, taxonomy, or floral morphology of the invader (Bruckman & Campbell, 2016;Iler & Goodell, 2014;Molano-Flores, 2014;Sun et al., 2013).
Locally rare or endemic plant species may be especially vulnerable to effects of non-native invasive species if the invaders reduce opportunities for outcrossing or result in high levels of interspecific pollen transfer. On the other hand, invading plants with abundant flowers may draw additional pollinating insects to the area that augment pollen movement among the endemic flowers (Jakobsson et al., 2009). Effects on endemic plant fitness would depend on the fidelity of individual insects to their flowers and impacts of nonconspecific pollen that may be deposited on the endemic plants' stigmas. Nonconspecific pollen may clog the stigmatic surface (Carvallo & Medel, 2016;Tscheulin & Petanidou, 2013) or may be allelopathic to conspecific pollen (Ashman & Arceo-Gomez, 2013).
Endemism is uncommon among plants in the Great Plains, with the exception of those specific to habitat "islands" with unusual edaphic characteristics, such as those preserved in Badlands National Park (BNP), South Dakota, USA. The clay outwash from the largely unvegetated erosional features, known as "badlands sparse vegetation complex" (Von Loh et al., 1999), harbors a handful of rare and/or endemic plant species, including Eriogonum visheri A. Nelson (Visher's buckwheat), the focus of the current study. Eriogonum visheri is a species of concern to management agencies in the northern Great Plains and has been assigned the rank of G3, which indicates globally vulnerable species, by NatureServe (Ladyman, 2006). Primary threats to the species are thought to be habitat destruction and invasion by exotic plants; little is known about the specifics of its habitat requirements, as populations often occur in isolated patches unevenly distributed across expanses of seemingly appropriate habitat (Ladyman, 2006).
Two invasive plant species with the potential to interfere with pollination of the buckwheat, Salsola tragus L. (Russian thistle) and Melilotus officinalis (L.) Lam. (yellow sweetclover), were identified in a study of pollination networks around E. visheri at BNP (Larson et al., 2014). Salsola tragus was determined to be more likely to have an effect on pollination of the endemic buckwheat based on its occurrence within the same module (i.e., groups of plants and flower visitors that interact more with each other than with those outside the module (Olesen et al., 2007)) as E. visheri. Melilotus officinalis shared very few flower visitors with E. visheri and occupied a different module in each of the two years of the study, so was deemed a lesser threat (Larson et al., 2014). Left unresolved by that study were fitness-level effects of these two species on E. visheri, for which, as an annual, production of viable seed is crucial. We undertook the present study to examine if the presence of either of the invasive species or a common, co-occurring congener, Eriogonum pauciflorum Pursh (fewflower buckwheat), that shared a module with E. visheri in one of the years of the earlier study resulted in changes to seed viability via direct effects of pollen on E. visheri stigmas, or indirect effects via visitation.
To determine this, we asked the following questions: (a) Are insect species that visit E. visheri consistent among years and equally likely to carry pollen from E. visheri and other species? Is visitation related to flower abundance at a site? (b) How common is the presence of nonconspecific pollen on E. visheri stigmas? (c) Does presence of E. pauciflorum, S. tragus, or M. officinalis pollen on stigmas of E. visheri influence achene weight or germination likelihood? Our overall goal was to determine whether results of this study are consistent with those of the network analysis, especially with respect to management considerations.

| Study species and study sites
Eriogonum visheri ( Figure 1) is a summer annual species in the Polygonaceae restricted to sparsely vegetated clay outwash soils found in badlands habitats in South Dakota, North Dakota, and Montana (Ladyman, 2006). Rosettes form in June, and flowering begins in July and may continue into September at BNP if moisture is adequate. The tiny (~2 mm diameter) yellow flowers are protandrous and may close in the evening but open again the following day. The total duration of a single flower is unclear but can exceed two days at our study sites. Closure brings the dehiscent anthers in close contact with the stigmas, suggesting that the species is likely autogamous. Flowers may occur singly or in groups of up to 4 per involucre.
Flower visitors to E. visheri include small halictid bees, beeflies, ants, and wasps (see Tables S1 and S2 in Larson et al., 2014), from which we infer that floral rewards include both nectar and pollen.
Eriogonum pauciflorum, a mat-forming perennial, is common throughout the badlands sparse vegetation complex at BNP. The white-to-pinkish flowers occur in a single cluster at the top of relatively long stems. Flowering begins in June and may continue through August (Great Plains Flora Association, 1986). Eriogonum pauciflorum commonly occurs interspersed with E. visheri at BNP and attracts many of the same flower visitors (Larson et al., 2014).
Salsola tragus is an annual species in the Chenopodiaceae introduced to South Dakota from Eurasia as a contaminant of crop seed in 1873 (Beckie & Francis, 2009;Great Plains Flora Association, 1986).
It typically is much taller (up to a meter in height) and produces more flowers than either of the native Eriogonums with which it grows intermingled. Flowering typically occurs in August-October (Great Plains Flora Association, 1986). Though often thought to be wind-pollinated, the bright yellow anthers were found to be visited by a variety of colletid and halictid bees in the southwestern US (Blackwell & Powell, 1981) as well as in BNP (Larson et al., 2014).
Melilotus officinalis is a biennial species in the Fabaceae and may grow even taller than S. tragus (0.5-2 m). It was first recorded at BNP in 1959 (Lindstrom, 1959) and occurs throughout the park, though with large inter-annual variation in abundance. The yellow flowers are attractive to honeybees (Apis mellifera; Otto et al., 2017); flowering typically occurs during June and July at BNP. This study occurred in 2014, 2015, and 2017. We used two of the four sites from the previous study (Larson et al., 2014); one no longer had E. visheri and one was inaccessible at the beginning of the current study. To find two replacement sites, we consulted BNP records of prior locations of E. visheri and prioritized those that were within an hour's walk of the access road through the park. All four sites we used in this study were sufficiently distant from each other to preclude sharing of insect pollinators ( Figure 2). All study sites were

| Flower abundance
We spaced ten 75-m long, variable-width transects equidistantly within each 1-ha site. We counted the number of flowering forbs by species (all species in flower were included) on each transect approximately every two weeks. Transect width varied between 0.25 and 2.0 m; very abundant flowers were counted within narrower transects than were uncommon flowers. All counts were standardized to 2-m width for analysis. Methods followed Larson et al., (2014).

| Insect visitation
We delineated insect visitation plots with pieces of PVC tubing.
Depending on the configuration of E. visheri flowering plants, the plot could be 0.5 m × 2 m or 1 m × 1 m. Because our aim was to include as many flowering plants as possible, plot location varied from survey to survey within a day and from day to day at any study site. No more than three visitation plots per site were conducted in a single day. We observed plots for 20 min, plus added handling time for captured insects. Any insect observed to visit an E. visheri flower (i.e., it was in contact with the reproductive parts of a flower) was recorded on a data sheet; insects were only captured if it was possible to do so without harming the E. visheri plant. We placed captured insects individually in vials charged with ethyl acetate. When quiet, we transferred the insect to a glassine envelope labeled with a unique identifier and information on the location and time of capture, F I G U R E 1 Counterclockwise from upper left, Eriogonum visheri flower with a U.S. penny for scale (photograph by Mary Behlke); E. visheri rosette just beginning to bolt (photograph by Diane Larson); E. visheri with Salsola tragus (lower right) on cracked clay soils typical of E. visheri habitat at Badlands National Park (photograph by Diane Larson) then placed the envelope in a jar also charged with ethyl acetate.
Although we recorded type of insect on the data sheet, comparison with those identified after capture suggests that small bees and flies were often confused. We therefore only analyzed the observational data collectively as insect visits, rather than within taxa.
To learn which species of pollen insects carried, we removed pollen from the bodies, including the surface of the scopae (Parker et al., 2015), of captured insects using small cubes of fuchsin jelly (Kearns & Inouye, 1993) as described in Larson et al., (2014). Cubes were gently melted on slides, covered with a cover slip, and the border of the slip then painted with latex paint to protect the slide contents. Pollen was identified to species and counted at 10 -100x with a light microscope; fewer than 10 grains of a plant species from an individual insect was considered contamination because small amounts of pollen could have been picked up from the net.
Specimens are deposited in the BNP museum collection. Other insects were identified only to order, with the exception of the fly, Paragus haemorrhous, which was distinctive and quite abundant.

| Achenes from pollen treatments
At each site, we marked 90-100 E. visheri rosettes in June, prior to bolt, each year, then randomly assigned them to pollination treatments, one treatment per plant (Table 1). Because plants were assigned treatments before flowers had formed, one treatment (MOP in 2015) did not have enough surviving treatment plants for some analyses. We collected pollen on wooden toothpicks per the assigned treatment and applied it to the E. visheri stigma. To be sure we were collecting enough pollen, we examined a sample of toothpicks for the targeted pollen under a microscope. In 2014, to identify which flowers were treated on each plant, we used varying numbers of knots tied into thread fixed to the branch proximal to the flower(s) F I G U R E 2 Map of Eriogonum visheri study sites at Badlands National Park, SD, USA. The two closest study sites were separated by 2.6 km. Note that the extent frame shows only the NE section of the park to be treated. We secured small mesh bags around the treated branch. Because flowers were borne in groups of up to 4 and could not be individually marked, we could not be certain which flower ( As achenes formed, we checked for ripeness by touching the achene with the tip of a pencil; ripe achenes easily dehisced. We did not forcefully remove achenes. Most achenes that dehisced were caught in the mesh bag, which we carefully opened at each visit after placing a white cloth around the base of the plant to catch the achenes that fell out of the bag. We bagged open-pollinated plants when we saw no new flowers on the plant, which corresponded with the beginning of achene ripening. When all achenes on a plant had dehisced and the plant appeared senescent, the plant was collected, dried, and weighed to document any systematic differences in plant growth among treatments. Although we did track and collect aborted flowers and achenes (Table S1) 12 hr dark at ~ 55˚F. We counted and removed achenes with a >1 mm radical (indicating germination) weekly and excluded those that developed mold prior to radical growth from the analysis.

| Pollen on stigmas
We collected no more than four (typically 1-2) flowers from an E.
visheri plant on any given day. Collection was haphazard, but an attempt was made to sample all plants at each site. In 2014, most of the collections happened within a single week, after which flower availability declined; collections in 2015 were more evenly distributed, but the season was again short in 2017 (Table S2). Flowers were removed from the plant with forceps and placed individually into glassine envelopes labeled with site, date, collector, and plant ID.
Individual flowers were kept in their own glassine envelopes prior to processing. Each flower was dissected on a cube of fuchsin jelly on a glass slide using a Meiji EMZ dissecting scope; each of 3 stigmas was separated onto the jelly cube. We gently heated the fuchsin jelly until it melted, then placed a cover slip on top. We scanned the entire area of melted jelly for pollen grains using a Leica DMLS microscope at 10 -100x, and recorded pollen to the specieslevel when possible; pollen grains were compared to a pollen ref- erence collection made at the study sites. It is important to note that it was not possible to distinguish E. visheri pollen that originated from the flower's own anthers from that brought to the stigma by an insect pollinator. The minute flowers were impossible to emasculate in the field. Data were summarized as number of grains by species per stigma. E. visheri pollen treatment, and P e refers to germination rate per achene when pollinators were excluded; P max is the larger of P s or P e .

| Statistical analyses
As noted above, we lack reliable estimates of aborted flowers and achenes, so here we evaluate effects of autogamy only on achenes that dehisced per the protocol described above.

| Characterize pollinator visitation to E. visheri
In 2014, we observed 67 plots between 2 and 30 July and counted 106 individual insects visiting E. visheri flowers (mean 1.58 per plot). Based on captured insects, for which we have positive identification, halictid bees were the most common visitors to E. visheri ( Figure 5); Lasioglossum packeri, a small sweat bee, was the only iden-

| How common is the presence of nonconspecific pollen on E. visheri stigmas?
Across all years, 97% of pollen counted on all slides of crushed E.

| Does presence of E. pauciflorum, S. tragus, or M. officinalis pollen on stigmas influence achene weight or germination likelihood?
We  (Table S4). Achene weight declined over the course of the season in 2015 (i.e., collection date was a significant covariate) but was unrelated to pollen treatment (Figure 7a; Table S4). In 2017, we saw a significant interaction between pollen treatment and achene collection date (Figure 7b, Table S4); however, at any given date (within the range of our collection dates), achene weight did not differ significantly among pollen treatments.
We expected germination rate to be related to achene weight and collection date, so we used both as covariates to explain effects of pollination treatments on germination in 2015 and 2017.
We did not assess germination rate in 2014, when we used collected achenes to develop a protocol to promote germination of ripe achenes. In 2015 and 2017, germination rate increased with achene weight, reaching an asymptote of near total germination in achenes weighing between 0.7 and 0.8 mg (Figure 8). In the 2015 and 2017 models, collection date was not significant and pollen treatment did not significantly affect germination rate in either year (Table S5).
Data from the MOP treatment were too sparse (as described above), so were not evaluated in the model.
We found no evidence for pollen limitation due to autogamy in filled dehiscent achenes. Germination rate was slightly higher in PE than OP achenes: PL = 0.004 for germination of PE versus EVP achenes, and PL = −0.015 for germination of PE versus OP achenes.
Based on the Baskin and Baskin (2018) standard, if PL falls between −0.01 and 0.01, germination is not pollen limited.
Plant weight did not vary among these treatments (Table S6), so effects we observed were not due to overall changes in plant vigor that depended on treatment. Plant biomass increased over time in 2014 and 2015 (Figure 9), but this effect was not significant in 2017 (Table S6)

| D ISCUSS I ON
Many of the observations we made in this study were consistent with those in the network study (Larson et al., 2014). The striking lack of M. officinalis pollen on both insects and stigmas, despite superabundance in 2014, makes it clear that this invasive species has little chance of interfering with E. visheri pollination. Our data cannot conclusively exclude effects on total visits from potential pollinators, but the lack of differences in visitation to E. visheri among years, despite orders of magnitude more M. officinalis flowers in 2014, suggests neutrality. Flower morphology has been implicated in likelihood of pollinator (and pollen) sharing among coflowering species (Ashman et al., 2020)  Abbreviations for treatments in the legends follow Table 1. Achene weight did not vary among pollen treatments in either year. Achene weight tended to increase over time in the pollinator exclusion treatments in 2017, but even at the end of the growing season, achene weights were not statistically different among treatments, suggesting that biological significance is unlikely E. visheri is not likely to suffer reduction in this fitness component due to heterospecific pollen receipt.
As in the case of Larson et al., (2014), L. packeri was the most common and consistent visitor to E. visheri, though other halictids sometimes carried more E. visheri pollen, especially in the long flowering season of 2015. The association between E. visheri and L. packeri is potentially an artifact of it being an abundant, generalist forager.
However, L. packeri is the smallest bee caught in this study, (female body length 4.0-4.4 mm; (Gibbs, 2010)), so it may be a better size match with the small E. visheri flowers than other co-occurring bees.
Carrying more pollen likely indicates visits to more flowers and an increased likelihood of cross-pollination. Based on germinability, E. visheri clearly is autogamous, but it lacks long-distance seed dispersal mechanisms, a common adaptation in edaphically restricted plant species (Schenk, 2013), so may rely on pollinators for genetic diversity.
Taking advantage of occasional advantageous conditions to extend the flowering season and attract a wider variety of pollinators may improve the chances of increasing genetic diversity via receiving pollen from a greater number of plants (Waananen et al., 2018). Studies of other habitat-restricted plant populations, including a rare Eriogonum subspecies (Archibald et al., 2001) and several species restricted to gypsum outcrops (Matesanz et al., 2018), have not detected reduced genetic variation, which suggests the potential for adaptation to naturally fragmented habitats (Corlett & Tomlinson, 2020).  Table 1 F I G U R E 9 Air-dried Eriogonum visheri plant weight in (a) 2014 and (b) 2015 as a function of the day-of-year that the plant was collected and treatment. Abbreviations for treatments in the legends follow Table 1 counted in 2015 was unexpected but may relate to the coflowering community. The number of E. visheri flowers was positively correlated with total flower counts, which may have reduced pollinator fidelity to E. visheri (Ashman et al., 2020). Indeed, nearly all of the recorded bee visitors are relatively small, generalist foragers (with the exception of one Melissodes, captured on E. visheri outside of a plot, which is an Asteraceae specialist and did not carry E. visheri pollen) that would not be expected to show any particular fidelity to E. visheri. Similarly diverse and unspecialized flower-visitor assemblages were found for Eriogonum pelinophilum, a rare perennial buckwheat found in restricted habitats in western Colorado (Tepedino et al., 2011). Further, the numerous males collected on E. visheri would have primarily been visiting the flowers for nectar and would be expected to carry less pollen than females (Larson et al., 2018).
Such interactions involving changes to the surrounding flowering plant community can be complex and dependent on pollinator community composition as well as changes in disturbance pattern and intensity (Portman et al., 2019).  (2017), using fluorescent dyes as pollen analogues, found that pollen movement from an invasive species to a native was greatest in the immediate vicinity of the invasive.
Given the patchy nature of S. tragus at our study sites, pollen may be lost passively or via grooming before the insect reaches many E. visheri flowers. Disentangling these hypotheses would require targeted experiments that we did not perform.
Mean achene weight declined over the collection period for all pollen treatments in 2015 but was constant in the shorter 2014 season and inconsistent but did not vary among pollen treatments in the still shorter 2017 season. The decline in 2015 seems unlikely to be related to pollination, as we saw no variation in the amount of pollen carried by insects over the flowering season. A more likely explanation is reduced nutrients and/or water as the season progressed, as precipitation in July-September was only 21% of that in April-June. Results of our germination studies suggest that these smaller achenes that ripened at the end of the season may be less likely to germinate, though this could be due to reduced viability or increased dormancy (Baskin, Chesson & Baskin 1993;Baskin & Baskin, 2014), which we did not test.
Although this study was not designed to test theories of germination as observed in E. visheri, we can add to the accumulating literature on conditions that favor Eriogonum germination (e.g., Young, 1989;Baskin, Chesson & Baskin 1993;Meyer & Paulsen, 2000). In particular, chilling alone was not effective at stimulating germination, at least within the refrigerator and greenhouse conditions our achenes experienced. However, scarification of the tip of the achene after chilling during the year after collection did result in high germination rates, given the temperature and light conditions of our greenhouse.
As Baskin, Chesson and Baskin (1993) pointed out for the winter annual Eriogonum abertianum, the predictability of the environment will likely influence the degree and kind of dormancy annuals will display.
In the case of E. visheri, environmental events such as freeze-thaw cycles may result in abrasion, which, combined with imbibing, may signal suitable conditions for germination.
More than a decade ago Mitchell et al., (2009)  While discussions of species rarity often focus on causes of rarity (e.g., Kunin & Gaston, 1993;Walck, Baskin & Baskin 2001;Combs et al., 2013), E. visheri may rather provide an example of a species well adapted to rarity. results, having found that when they excluded visits that did not lead to seed set, network metrics changed dramatically. Our results suggest that, from the plant's perspective, details such as autogamy and effects of heterospecific pollen are key factors not accounted for in network analysis. Nonetheless, the network analysis upon which this study was based did provide knowledge of the pertinent community of plants and insects, if not the specifics and direction of each interaction. Larson et al., (2014) contrasted the broad, community-level understanding gained with a network approach with more limited, but more specific, understanding gained in a study directed at the focal plant's visitors. In this case, we gained more confidence in the effects of the invasive species by directly assessing effects of their pollen on achene size and germination. Although it was still necessary to collect insect visitors for identification, fewer individuals were involved. In addition, the current study added new insight into germination requirements for a rare, edaphically specialized species of Eriogonum.

ACK N OWLED G M ENTS
We Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government.

CO N FLI C T S O F I NTE R E S T
None declared.