Newly introduced trait diversity can spur rapid evolution and facilitate local adaptation in the introduced plant Lythrum salicaria. The horticultural plant L. virgatum might further introduce meaningful trait variation by escaping into established L. salicaria populations or by hybridizing with L. salicaria. Although many experiments have focused on L. salicaria genotypes, relatively little is known about L. virgatum ecology. We used a greenhouse common garden to compare traits and flood response of L. salicaria and L. virgatum collected from two sources each in their native range. We tested the hypotheses that these two wetland taxa have comparable responses to flooding (inundation), and that flood tolerance correlated to higher fitness. Flooding produced stronger stress responses in L. virgatum. Compared to L. salicaria, L. virgatum shifted more aboveground allocation away from reproduction, decreased inflorescence biomass by 40% more, and produced 7% more stem aerenchymatous phellum, a specialized tissue that maintains aeration. Despite these more pronounced responses to flooding stress, L. virgatum had higher fitness (inflorescence biomass and reproductive allocation) than L. salicaria. Overall, L. virgatum differed from L. salicaria in functionally important ways. Lythrum virgatum persisted under flooding and produced more reproductive biomass than L. salicaria under both flooded and non-flooded conditions. However, inundation stressed L. virgatum more than L. salicaria. Lythrum virgatum is likely able to establish in the wetland habitats in which L. salicaria prevails but may possess broader habitat tolerances.

Plant material—We obtained seeds of two different accessions for each species from Europe. All four accessions originally hailed from local natural collections rather than horticultural sources, an important requirement given the unknown origins of many cultivars (Strefeler et al., 1996b). Two accessions of L. salicaria, one each from Cornwall and Surrey, UK; were provided by the Millennium Seed Bank Partnership, Kew Gardens. We obtained two L. virgatum accessions, one from Rostock Botanical Garden, Rostock, Germany, and one from Anastasie Fatu Botanic Garden, Iaşi, Romania. Purported species identities of each accession matched identifying traits described by Haining and Graham (2007). Compared to the L. salicaria accessions, L. virgatum plants were overall smaller with smaller flowers (note flower size differences in Fig. 2A,B) and were glabrous rather than hairy (Fig. 2A,B). We also measured calyx appendages and calyx lobes; L. salicaria has long appendages, while those of L. virgatum are much shorter (Haining and Graham, 2007). Ratios of calyx appendage length to calyx lobe length (Fig. 2A,B) aligned with expectations: L. salicaria Cornwall, UK = 1.988 [1.373—2.529], L. salicaria Surrey, UK = 1.891 [1.165—3.274], L. virgatum Germany = 0.662 [0.320—1.206], L. virgatum Romania = 0.914 [0.530—1.211] (means and ranges across 3—6 flowers from 4—6 individuals of each accession).

Experimental design—Seeds from all accessions had high survivability (~90%) when germinated in Petri dishes on moist filter paper. We transplanted germinants into fertilizer-free propagation mix (LM-AP All Purpose Mix, Lambert, Rivère-Ouelle, Quebec, Canada). We retained four robust genets of each of the four accessions to serve as maternal plants for our experiment. Within accessions, these genets were presumably distinct genotypes, although no information was provided regarding how our source seeds were originally collected. We grew and acclimated genets for about one year in a greenhouse at The Ohio State University (Columbus, Ohio, USA) under 16 h/d of light. One L. salicaria (from Cornwall, UK) died during this time, bringing the total to 15 genets. We cloned genets by planting stem cuttings dipped in rooting hormone (Garden Safe TakeRoot, Spectrum Brands, Middleton, Wisconsin, USA) in germination trays. We allowed cuttings at least 20 d to establish, then selected robust individuals to be transplanted into 20 cm diameter × 13 cm tall pots with drainage holes, filled with propagation mix (LM-ORG, Lambert, Rivière-Ouelle, Québec, Canada) mixed with ~1% solution of dish detergent (Ajax, Colgate-Palmolive, New York, New York, USA) to aid rewetting. Prior to initiation of treatments, these pots were kept in trays and bottom-watered every other day by filling the tray to a depth of ~5 cm (8 cm below the top of the pot).

We used a randomized block design with 64 individuals total: two greenhouse rooms × two species × two seed accessions × four genets (with the contribution of one salicaria Cornwall mother doubled) × two experimental treatments. This procedure resulted in eight replicates per species (two accessions × four clones) per treatment in each of two blocks (i.e., greenhouse room). We randomized placement of individuals within replicates.

Treatments—After a period of 60 d post-transplant, we began preparing plants for the start of our two treatments: flooded and non-flooded conditions. We first spread a ~3 cm layer of coarse sand (Paver Base, Quikrete, Atlanta, Georgia, USA) across the top of all pots to prevent flooded individuals from floating. All individuals, flooded and non-flooded, received sand to control for any effect this media addition might introduce. We trimmed aboveground biomass to standardize individuals to a similar size prior to the start of treatments. We retained the single largest stem, trimming any additional stems flush to the sand. We trimmed the single retained stem to a height of 30 cm above the sand and removed any lateral branches > 1 cm long. We collected the removed plant material, separated it into vegetative and inflorescence fractions, measured the dry mass (after drying at 60°C), and used these data as model covariates to control for pre-experiment size differences among individuals. We also noted at this time which individuals were flowering prior to the start of the experiment. We then nested all 13 cm tall pots within larger ~19 L pots. For the flooded treatment, the outer pots were  28 cm in height and did not contain drainage holes. For the non-flooded controls, the outer pots were holed pots with heights of 24—31 cm. These heights varied due to product availability, but we spread pots of different heights evenly across different replicates and accessions. Nine days after these preparations were completed, we initiated the flooding treatment. For flooded individuals, we lined the unholed outside pots with 2 mm contractor bags (Up & Up, Target, Minneapolis, Minnesota, USA) to ensure they were watertight and filled them to a water depth of ~27 cm (14 cm above the top of the inner pot containing the plant). An inundation depth of 14 cm is near the upper limit reported for L. salicaria in the wild, and the species is seen across a wide range of moisture regimes, from irregular soil saturation to permanent inundation (NCHRP, 1996). For non-flooded individuals, bottom-watering proceeded as previously described. We topped off water every other day and replaced flooded individuals’ water every 7 d to limit algae growth. We ran the experiment for a total of 53 d, at which point some flooded individuals were showing signs of senescence. Given these species’ invasive tendencies in the area we conducted this study, we did not cross-pollinate individuals, meaning they did not set seed (these heterostylous species require crossing to set seed [Colautti et al., 2010b]).

Data collection—To quantify each species’ response to flooding, we collected aboveground biomass and sampled stem aerenchymatous phellem at the end of the experiment. We separated the aboveground material into vegetative and inflorescence biomass and measured dry masses. Fractioning the biomass allowed us to examine potential differences in allocation strategies, which have been shown to be genetically controlled in L. salicaria (Olsson, 2004; Colautti and Barrett, 2011). We investigated total aboveground biomass as a metric of overall plant size and vigor, as well as two indicators of fitness. The first was inflorescence biomass, to gauge total reproductive output. The other fitness metric was the ratio of inflorescence biomass to total aboveground biomass, or an individual’s reproductive allocation. We also quantified the aerenchymatous phellum that formed around the interior vascular cylinder of the stem (Fig. 2C)—a flood tolerance response well-characterized in L. salicaria (Schenck, 1889; Stevens et al., 1997; Lempe et al., 2001; Stevens et al., 2002) and also observed for L. virgatum (Schenck, 1889). From the largest stem of each individual, we cut a 4 cm portion starting 1 cm from the top of the soil and submerged it in FAA (Formalin-Alcohol-Acetic Acid, 10% : 50% : 5% + 35% water). Then, we hardened the tissue by storing it in 70% ethanol. From each stem section, we used a razor blade to take a cross-section from the middle (~3 cm from the top of the soil). We photographed cross sections (Olympus SZX7, SC100, and cellSens, Olympus, Tokyo, Japan) and used winFOLIA (Regent Instruments, Sainte-Foy, Quebec, Canada) to draw measurement polygons around the inner vascular cylinder and the outer aerenchyma. The aerenchyma area equaled the area inside the outer polygon minus that inside the inner polygon. We calculated aerenchyma production: the ratio of aerenchyma area to total stem area. Aerenchyma measurements excluded data from one flooded German L. virgatum individual that died a few days before harvest and for which the aerenchyma had become visibly deflated.

Data analysis—We tested Hypothesis 1 by comparing all traits across species by treatments and Hypothesis 2 by testing whether our two fitness indicators (inflorescence biomass and reproductive allocation) were predicted by our flood tolerance metric (aerenchyma production).  To compare how many individuals of each species were flowering prior to the start of the experiment, we used a Pearson’s Χ² test (stats::chisq.test; R v. 4.0.3, R Development Core Team, 2020). For all other analyses, we used linear mixed models with general purpose optimization (Pinheiro et al., 2020). For Hyp. 1 models, we tested whether trait values were predicted by species × treatment interactions. Hypothesis 2 models had a fitness metric as a response variable and aerenchyma production as a predictor. Species was also included as a predictor in Hyp. 2 models because the relationship between aerenchyma production and fitness might be moderated by species. Hypothesis 2 models included only individuals that produced aerenchyma (flooded treatment). All Hyp. 1 and 2 linear mixed models included random effects for block (i.e., replicate) and accession, and we weighted variance by accession to address variance heterogeneity. Analyses of the three biomass-based responses included as covariates their respective pretreatment measurements. These covariates accounted for the fact that some plants had grown larger than others prior to being trimmed to a standard size at the start of the experiment, and some had initiated flowering while others had not. Analyses of traits other than total aboveground biomass also included that trait as a covariate, so that model results could be interpreted independent of potential size-based trait differences. The statistics we present for each predictor from the linear mixed models are P-values for significance, calculated from type III sums of squares (Fox and Weisberg, 2019), and β coefficients to describe the strength of relationships. For continuous predictors, β coefficients are interpreted as regression coefficients or slopes of the relationship of that predictor to the response (e.g., for Hyp. 2, the relationship of aerenchyma production to a fitness metric). For categorical predictors, β coefficients contrast means of each category (e.g., for Hyp. 1, the mean trait value for L. virgatum is compared to the mean for L. salicaria).

R: a language and environment for statistical computing, https://cran.r-project.org