Herbivores play a critical role in plant invasions either by facilitating or inhibiting species establishment and spread. However, relatively few studies with invasive plant species have focused on the role of plant tolerance and how it varies geographically to influence invasion success. We conducted a common garden study using two lineages (native and invasive) of the grass Phragmites australis that are prevalent in North American wetlands. Using 31 populations collected across a broad geographic range, we tested five predictions: 1) the invasive lineage is more tolerant to simulated folivory than the native lineage, 2) tolerance to herbivory decreases with increasing latitude of origin of the populations, 3) estimates of tolerance are correlated with putative tolerance traits and plasticity in those traits, 4) a tradeoff exists between tolerance and resistance to herbivory, and 5) tolerance has a fitness cost. Response to folivory varied substantially among populations of P. australis, ranging from intolerance to overcompensation. Our model selection procedure deemed lineage to be an important predictor of tolerance but, contrary to our prediction, the native lineage was 19% more tolerant to folivory than the invasive lineage. Tolerance for both lineages exhibited a u-shaped relationship with latitude. A tolerance-resistance tradeoff was evident within the invasive but not the native lineage. Also, tolerance was positively correlated with belowground biomass allocation, leaf silica concentrations, specific leaf area and plasticity in stem density, and negatively correlated with the relative growth rate of the population and plasticity in putative resistance traits. Lastly, although we did not detect costs of tolerance, our results highlight that fast growth rates can maintain high fitness in the presence of herbivory. Herbivory and plant defense strategies for P. australis lineages in North America exhibit complex biogeographic patterns that cause substantial heterogeneity in enemy release and biotic resistance and, consequently, invasion success.

From Croy et. al. (2020):

Study system

Phragmites australis is a 2-5 m tall perennial grass commonly found in wetlands, estuaries, salt marshes, ponds, and rivers on every continent except for Antarctica (Clevering and Lissner 1999). Although present in North American wetlands for millennia (Hansen 1978, Orson 1999), P. australis began spreading aggressively, dominating wetlands and negatively impacting native plant species, hydrologic regimes, nutrient cycles, and ecosystem function (Chambers et al. 1999, Meyerson et al. 2009, 2010). The rapid spread is attributed to the introduction of an invasive Eurasian lineage (Haplotype M; P. australis australis) that first appeared in the herbarium record about 150 years ago (Chambers et al. 1999, Saltonstall 2002). Populations of the Eurasian lineage in North America are genotypically diverse (Saltonstall 2003) and despite being clonal, genotypic variation has been identified within patches (McCormick et al. 2010). Additional haplotypes have been introduced from Europe, North Africa (e.g., Lambertini et al., 2012; Meyerson and Cronin, 2013) and Asia (Lambert et al. 2016), but their known distributions are localized and none were included in this study. Throughout North America, at least 14 closely related native endemic haplotypes have been identified (Saltonstall 2002, Meadows and Saltonstall 2007, Vachon and Freeland 2011). These native haplotypes have been given subspecies status (P. australis americanus) but are collectively referred to as the native lineage.

In Europe where it is considered native, P. australis is host to a diverse assemblage of arthropod herbivores with over 170 species identified. In contrast, only 26 herbivore species have been reported in North America (Tewksbury et al. 2002). Within North America, herbivory on P. australis is primarily attributed to herbivores introduced from Europe, most prominently stem-galling flies in the genus Lipara (Lambert et al. 2007, Cronin et al. 2015, Allen et al. 2015) and the mealy plum aphid Hyalopterus pruni (Lozier et al. 2009, Cronin et al. 2015). Based on previous research, there is strong support for the enemy release hypothesis: the invasive lineage suffers substantially less herbivory from each of three major feeding guilds (aphids, stem-gallers and leaf chewers) in its invasive than native range, and in comparison to co-occurring plants of the native lineage in North America (Cronin et al. 2015, Allen et al. 2015, Bhattarai et al. 2017a, b). Because these differences in herbivore abundance or damage between the native and invasive lineages are also manifested under common garden conditions, it suggests greater resistance to herbivory by the invasive than native lineage of P. australis and that resistance is genetically based (Lambert and Casagrande 2007, Allen et al. 2017, Bhattarai et al. 2017b). Interestingly, Cronin et al. (2015) found that leaf damage and stem-galler incidence decreased with increasing latitude for the native lineage but not invasive lineage, and these non-parallel latitudinal gradients in herbivory resulted in stronger enemy release at southern than northern latitudes (Cronin et al. 2015). This genotypic and latitudinal variation in herbivory lends credence to a biogeographic investigation into the role of plant tolerance to herbivory in invasion success.

Common garden

The common garden design used in this experiment is detailed in Bhattarai et al. (2017b). Briefly, rhizomes from P. australis patches were collected from field sites throughout North America spanning 19.5° of latitude and 55.9° of longitude and planted in a common garden established at Louisiana State University, Baton Rouge, LA (30.35° N, 91.14° W) in 2009 (Figure 1; Appendix A: Table S1). Plants were grown for at least one year prior to the start of the experiment to minimize the influence of maternal effects. In early March 2014, we potted 10 replicates for each of 31 source populations of P. australis (13 native, 18 invasive) in 7.6-liter pots. Each replicate consisted of a single rhizome cutting (10-15 g) potted in sand to standardize nutrients and initial starting conditions. Due to lower than expected propagation success, we added five additional replicate pots for each population in early May 2014. We fertilized each pot with Osmocote^® (58 grams/pot of 3 month, slow-release in March followed by 58 grams/pot of 9 month, slow-release in June of 15-9-12 NPK, The Scotts Miracle-Gro Company^®, Marysville, Ohio) and Ironite^® (1.7 grams/pot; Pennington^®, Madison, Georgia) to ensure that resources were standardized. We repeatedly sprayed the plants with a non-systemic insecticide (Ortho^® Malathion; The Scotts Miracle-Gro Company^®, Marysville, Ohio) to prevent herbivore damage. Finally, we placed potted plants from each population in the same plastic pool (1.2 m diameter) filled with tap water. Plants subjected to different levels of folivory within a population were randomly distributed within pools, and populations were randomly distributed within the common garden.

Folivory treatment and tolerance

We implemented an artificial folivory treatment to simulate the effects of heavy folivory, a common approach in the study of tolerance to herbivory (e.g., Marquis 1988, Tiffin and Inouye 2000, Vergés et al. 2008, Ashton and Lerdau 2008, Lurie et al. 2017). Phragmites australis lineages vary widely in resistance to folivory (Cronin et al. 2015, Bhattarai et al. 2017b) and, consequently, it would have been difficult to achieve a standardized level of herbivory among source populations without varying herbivore density and/or exposure time. We assigned plants within a population at random to either a folivory or no folivory treatment. Starting in late May 2014, we clipped 40% of the leaf area, followed by a monthly removal of 40% of the new growth until late August. We clipped plants from the top-down to reflect herbivore behavior, because they generally remove the newest, most palatable leaves first. Although 40% folivory is a severe damage treatment, this level of herbivory has been observed in the field (Cronin et al. 2015).

At the end of the growing season (early November in Louisiana), we harvested above- and below-ground plant material, which was then air-dried on benchtops in the greenhouse until completely dried (2 mo.), and measured with a hanging scale (Pesola^©,Schindellegi, Switzerland; precision of ± 0.3%). Because flowering frequencies for P. australis in the first year following propagation from small rhizome cuttings are quite low, and biomass encapsulates all aspects of asexual reproduction, we used the total dry biomass at the end of the season as our proxy for fitness. Moreover, we had two planting dates owing to the need for supplemental plants, and it was our intention to include a blocking effect for planting date in our statistical models. However, we did not have sufficient replication within the second block to calculate tolerance for each source population. To account for possible differences in final plant size between planting dates, we used least-squares means for plant biomass (above-, belowground, and total biomass) computed from a general linear model that included the main and interactive effects of clipping treatment and population, as well the main effect of planting date. From this, we obtained a planting date-independent estimate of plant biomass for each population in both the clipped and unclipped treatments. With these estimates, we calculated tolerance for each population using log-response ratios, where Tolerance_total = ln[mean end-of-season total biomass of clipped plants/mean end-of-season total biomass of unclipped plants] (Hedges et al. 1999). A Tolerance_total value of zero would indicate no effect of folivory on end-of-season biomass whereas a Tolerance_total > 0 would indicate overcompensation. Overall, the larger the value of Tolerance_total, the greater the tolerance of that population. We also calculated log-response ratios for aboveground (Tolerance_above) and belowground (Tolerance_below) biomass separately to compare the relative impacts of folivory on the shoot and root biomass, respectively. Lastly, we note here that the above method for estimating tolerance necessitated that the population, not the individual pot, was our unit of replication in this experiment, resulting in a single estimate of tolerance for each population. We averaged the estimates for dependent variables at the population-level to avoid pseudo-replication (rhizomes were collected from same pool of plants for each population and thus are not completely independent samples).

Putative tolerance and resistance traits

In addition to population-level measures of tolerance, we also measured a suite of plant functional traits that are widely recognized as correlates of plant tolerance and resistance to herbivore damage. How these traits change in response to a folivore treatment can provide clues as to the mechanisms underlying plant defenses. Putative tolerance traits like relative growth rate (RGR), root-shoot ratio, stem density, and photosynthetic rate are closely associated with compensatory ability (e.g., Meyer, 1998; Strauss and Agrawal, 1999; Tiffin, 2000). To estimate plant RGR, we divided total end-of-season biomass by the number of days in the growing season. With the exception of our blocked design, all plants were propagated and harvested at the same time, and so we believe that a linear estimate of RGR captures variation in RGR across populations. To account for the fact that some plants grew for a longer period of time than others due to the blocked design, we accounted for the blocking effect RGR statistically as was done for biomass (see above). In addition to RGR, we measured stem density at the last census of the growing season (late August 2014) and calculated root mass fraction (= belowground biomass/total biomass; Pérez-Harguindeguy et al., 2016).

High rates of photosynthesis are thought to be a potential mechanism of tolerance, and specific leaf area is a frequently used correlate for photosynthetic ability (Pérez-Harguindeguy et al. 2016). In August 2014, we photographed and subsequently collected the uppermost fully open leaf of three stems per pot. The leaves were dried at 50°C for 72 hours before being weighed. We measured leaf area using ImageJ software (Schneider et al. 2012), and calculated specific leaf area as the ratio of area to dry biomass for a leaf (mm² mg^-1). Using a porcelain pestle and mortar, we ground the dry leaf material into fine powder for leaf chemical analyses (see below).

Foliar nitrogen and carbon serve important physiological and ecological functions, with nitrogen content linked to photosynthetic ability (Evans 1989) and both carbon and nitrogen content shown to influence herbivore performance (Agrawal 2004, Imaji and Seiwa 2010, Cronin et al. 2015). In particular, because nitrogen is often positively correlated with plant RGR and RGR is a putative tolerance trait, plants with high foliar nitrogen content are expected to better compensate for herbivore damage than plants with low nitrogen content (e.g., Vergés et al., 2008; Bagchi and Ritchie, 2011; Mundim et al., 2012). Therefore, we assayed dried leaf tissues for percent carbon (%C), percent nitrogen (%N), and C:N ratio using an elemental analyzer at Brown University Environmental Chemistry Facilities (http://www.brown.edu/Research/Envchem/facilities/).

            In addition to foliar nutrients, we measured leaf toughness and total phenolics, two putative resistance traits known to correlate with P. australis herbivory (Cronin et al. 2015, Bhattarai et al. 2017b). Using a penetrometer (Itin Scale Co., Inc., Brooklyn, NY), leaf toughness (force [kg] required to push a blunt steel rod [4.8 mm in diameter] through the leaf) was measured on the uppermost fully open leaf on three randomly selected stems per plant. Total phenolics (nM/g of dried leaf tissue) were estimated using a modified version of the Folin-Ciocalteu method (Waterman and Mole 1994, Cronin et al. 2015) at the University of Rhode Island, Kingston, RI. Although never before examined in P. australis in the context of plant defense, we also analyzed plants for silica content (g kg^-1) at the Louisiana State University AgCenter Soil Testing and Plant Analysis Lab, using methods adapted from Kraska and Breitenbeck (2010). Silica is a known herbivore defense in grasses (McNaughton et al. 1985, Massey et al. 2006, Reynolds et al. 2012).

            Putative resistance traits sometimes fail to predict, or are only weakly correlated with, true resistance measured in terms of herbivore preference or performance on different plants (e.g., Fritz and Simms, 1992; Bhattarai et al., 2017b). The best test of tolerance-resistance tradeoffs is with true measures, not putative correlates of each trait. Although we did not measure true resistance in this study, we used data from Bhattarai et al. (2017b) that quantified P. australis resistance to a generalist leaf-chewing herbivore, Spodoptera frugiperda (fall armyworm; Lepidoptera: Noctuidae) on the same clonal populations used for this study and from within the same common garden. Briefly, in this bioassay, Bhattarai et al. (2017b) reared individual pre-weighed S. frugiperda larva on replicate plants of each source population for eight days. Afterward, larval biomass and leaf area consumed were measured and plant palatability to the herbivore (biomass conversion efficiency) was calculated (proportional change in larval fresh mass per unit area of leaf consumed; all measurements ln-transformed). The inverse of this metric was used as a measure of resistance. Between the two experiments, there were 20 overlapping populations that used identical genetic material for plant propagation. Finally, we note that although resistance and tolerance were measured one year apart, our work suggests that population-level resistance is strongly genetically based and relatively fixed over time (Allen et al. 2017, Bhattarai et al. 2017b). Moreover, clonal integrity was maintained in the garden over the years by preventing sexual reproduction (via removal of all panicles prior to floral maturation) and external contamination of our populations from naturally occurring seed sources was exceedingly unlikely given that the nearest source of P. australis was > 100 km away. Thus, we feel confident that these two data sets allow for a valid assessment of a tolerance-resistance tradeoff among P. australis populations.

Also, see the following manuscript for description on the common garden design.