Asclepias syriaca is a weedy perennial plant native to the Great Plains and northeast regions of the United States and into Canada (Woodson 1954; Bhowmik 1994), although its range has been extending farther south, perhaps following human disturbance (Wyatt 1996). It reproduces asexually using underground rhizomes, as well as sexually and primarily through outcrossing (Kephart 1981; Bhowmik 1994). Milkweed pollination occurs via transfer of pollinia, meaning that each wind-dispersed seed pod contains seeds that are full siblings. Milkweeds, including A. syriaca, are known for a suite of defensive traits, including sticky latex, toxic cardenolides, and trichomes; nevertheless, they harbour a variety of specialist insect herbivores that consume all parts of the plant (Agrawal 2005) and that could select for high tolerance. Plausible mechanisms of compensation have been reported for A. syriaca. These include responding to damage by reducing photosynthesis in damaged leaves, but increasing photosynthesis slightly in neighbouring undamaged leaves (Delaney et al. 2008), and by shifting allocation of nitrogen to stems after damage to leaves and roots (Tao and Hunter 2013). Previous studies have shown latitude-based patterns in phenology, growth, defence, and reproductive traits (Woods et al. 2012). However, we are not aware of any studies reporting geographic variation in compensatory growth responses to damage in this species.

We grew seeds from 16 genetic families collected from 14 different sites across the range of A. syriaca (Fig. 1; Table 1). Seedpods from naturally-occurring plants were collected in the fall of 2016 and sent to St. Olaf College for storage until planting. We planted seeds from a single seed pod per plant. This design maximizes our ability to measure compensation by comparing damaged and undamaged plants from the same genetic family. However, 12 sites were represented by just a single genetic family, and two additional sites were represented by two genetic families (Fig. 1; Table 1). This decision limits our ability to detect fine-scale (site-by-site) patterns in geographic variation. For analyses, we divided the range of common milkweed (60–105°W and 35–50°N; Bhowmik 1994) into two regions which allows us to run ANCOVAs with higher levels of replication than were available at the site level (see Data analysis section below). First, we split the sites into regions of high and low latitude, split at 42.5°N, and we also split the sites into groups of high and low temperature, based on the 30-year mean annual temperature from the nearest weather station to each site (1981–2010 US Normals Data, Arguez et al. 2010; Fig. 1; Table 1). Average temperatures at sites ranged from 3.72–12.67°C, and we divided sites into temperature groups above and below 8°C. We acknowledge geographic variation correlates with variation in a variety of other environmental factors; we also investigated grouping sites by 30-year mean annual precipitation (1981–2010 US Normals Data, Arguez et al. 2010), but we did not find significant patterns, so we do not report these analyses.

Figure 1. 

The distribution of milkweed source populations (Asclepias syriaca), bisected at 42.5°N. There are seven populations in the high-latitude regional group, above the line, and seven populations in the low-latitude regional group, below the line. There are also seven populations in the high-temperature group, with a mean annual temperature above 8°C, and seven populations in the low-temperature group, with a mean annual temperature below 8°C. To control for genetic variation, a single genetic family was studied from most source populations; however, two genetic families were studied from the highest and lowest latitude sites: BMS and LYN.

Seeds from each pod were massed, and we cold-stratified 50 seeds per family in moist sand in the refrigerator for three weeks before sowing seeds for germination in M1 potting mix (Grower Select; BFG Supply, Burton, OH). Greenhouse conditions were set to 25/18°C on a 16:8 day/night cycle and flats were watered as necessary. Three weeks after germination, we transplanted seedlings into their experimental 3”-square pots and added ~2.5 g 14-14-14 NPK slow-release fertilizer (Osmocote).

The experiment was conducted in eight spatial blocks of approximately 48 plants each, spread across two bays of the St. Olaf College greenhouse. The placement of each plant was randomized within the block. Each family was represented by 24 plants divided such that there were twice as many plants in the control group as in the damage treatment group for each family. Not all plants survived, so replication was slightly uneven across family groups (Table 1). Damage was applied manually seven weeks after sowing by tearing leaves along the midrib (following Kula et al. 2020) to remove half of each of the top 50% of the leaves on the plant, with the goal of removing approximately 25% of the total leaf area (Fig. 2A).

Figure 2. 

Experimental design and analysis plan. A. We tested for variation in the effect of 25% mechanical leaf area removal on total mass. Genetic families were divided between high- and low-latitude regional groups and high- and low-temperature groups, and separate models tested for a genetic family × treatment interaction and a group × treatment interaction. B. We interpret results as follows: groups exhibit overcompensation if they grow more in the damage treatment than in the control treatment (a), exact compensation if they grow similarly in the damage and control treatments (b), or undercompensation if they grow less in the damage than in the control treatments (c).

The end of the experiment occurred 10 weeks post-germination. We assessed plant compensation to damage by measuring total plant biomass (roots, shoots, leaves) after the final harvest. Groups exhibit overcompensation if they grow more in the damage treatment than in the control treatment, and undercompensation if they grow less in the damage treatment than in the control treatments (Fig. 2B). We also evaluated possible mechanisms for variation in compensation responses, including phenology (early stem height, following Woods et al. 2012), resource allocation (stem investment, root:shoot ratio), and relative growth rate (RGR).

Early stem height, our metric of phenology, was measured at 5 weeks from the soil to the top of the stem. Prior to damage at week 7, we measured stem height and the basal diameter of the stem 1 cm above the soil, a metric which was used as a covariate in many analyses to account for initial plant size. While the plants were early in their development, seven weeks represents a period in the plant lifespan when exposure to insect herbivores in a natural field setting is high. We harvested plants at 10 weeks to prevent loss of leaf tissue due to senescence. Our experiment was slightly shorter than similar experiments with milkweed (Hochwender et al. 2000 applied damage at 70 days and harvested at 100 days; Woods et al. 2012 destructively harvested at 85 days).

At the end of the experiment, resource allocation was measured both as root:shoot ratio, a predictor of tolerance under some conditions (Hochwender et al. 2000), and as “stem investment,” because milkweeds shift some resources to stems after damage (Tao and Hunter 2013). We also measured the basal diameter and height of each plant. Then, plants were harvested by cutting the roots from the shoots (stems and leaves), washing soil from the roots over a sieve, and freeze-drying (Labconco FreeZone 4.5). Once freeze-dried, the roots and shoots were separately massed (Mettler MS1602TS) to calculate root:shoot ratios. We estimated the final volume of the stem, using the basal diameter and final stem height measurements, and assuming the stem is roughly cylinder shaped. Stem investment was calculated as the ratio of estimated final stem volume to final shoot biomass. We calculated the relative growth rate (RGR) over the treatment period as the log of the ratio of the final stem height to the pre-damage stem height, divided by the days between measurements.

We log-transformed root:shoot ratios to improve normality of the residuals for analyses, and we log-transformed total biomass to facilitate proper interpretation because damage was applied proportionally (Wise and Carr 2008).

For our measure of phenology, high-latitude plants were not significantly taller on average than low-latitude plants during the early season at week 5 (Χ² = 2.3779, d.f. = 1, p = 0.1231), nor were there differences among temperature groups (Χ² = 0.5439, d.f. = 1, p = 0.4608). There were also no significant differences between groups in pre-damage basal diameter, the covariate used in models evaluating response to damage (Region: Χ² = 0.4019, d.f. = 1, p = 0.5261; Temperature Group: Χ² = 0.3715, d.f. = 1, p = 0.5422).

Damage (25% leaf area removal) reduced final plant biomass by 18% on average, but its effects (assessed three weeks after damage treatment) varied with genetic family, ranging from 30% average decrease to 10% average increase in mass following damage (Fig. 3A; Table 2). Damaged plants also shifted their biomass allocation toward shoots, reducing root:shoot ratios by an average of 12%, a response that did not vary significantly among families (Fig. 3B; Table 2). There were no overall effects of damage on stem investment (Fig. 3C) or relative growth rate (Fig. 3D), but the effects of damage on these traits varied significantly among families (Table 2).

Figure 3. 

Variation among temperature groups and genetic families (background) in the effects of damage on milkweed traits. Compared to plants from warmer regions (average temperature > 8°C; red solid lines), plants from colder regions (dashed black lines, average temperature < 8°C) (A) had constitutively higher mass but experienced greater reductions in total mass in response to damage. (B) Plants from both regions responded similarly to damage by reducing root:shoot ratios. Damaged plants from warmer regions tended to (C) increase stem investment (interaction p = 0.088), and (D) increase relative growth rate (based on height) compared to plants from colder regions (interaction p = 0.095). Variation among genetic families in the slope of damage responses (background lines) was significant for all traits shown except for root:shoot ratios.

At week 10, we found that plants from high latitudes and colder sites were larger in the control treatment but less able to compensate for damage. Damage reduced the total mass of plants from low-temperature populations by 24% (21% in high-latitude sites), compared to a 12% reduction in the total mass of plants from high-temperature populations (14% in low-latitude sites; Table 3; Fig. 3A). High-latitude and low-temperature plants tended to invest less in stems and roots, though the main effect is only significant for the effect of temperature group on stem investment (Fig. 3; Table 3). We found no significant regional differences in plant responses to damage via root investment, stem investment, or relative growth rate. However, there was a marginally significant trend for plants from the cold-temperature group to reduce stem investment and relative growth rate in the damage treatment while those from the high-temperature group tended to increase stem investment and relative growth rate (Fig. 3C, D; Table 3).

Growing season length varies dramatically with latitude and could explain the regional variation in compensation we observed by one of three different pathways. First, genetic differences in germination timing and phenology could cause plants to be at different developmental stages when damaged, potentially affecting both the resources and time available for plants to recover from damage (e.g. Lehndal and Ågren 2015).

Second, traits associated with adaptation to short growing seasons or stress, such as root investment and fast growth rates, often facilitate compensation (Więski and Pennings 2014; Hahn and Maron 2016; Croy et al. 2020). However, we observed a trend for faster relative growth rates in the damage treatment in plants from warmer sites, where fast growth is unlikely to be an adaptation to a short growing season, but may be associated with phenological differences in resource allocation (see below). Compensatory ability following damage is associated with storage of resources in roots and stems (Strauss and Agrawal 1999; Boege et al. 2007; Stevens et al. 2008; Croy et al. 2020), and specifically root storage in A. syriaca (Hochwender et al. 2000). Plants from warmer sites exhibited greater stem investment, and a trend toward greater root investment, in our study. As a result, they may have been better able to compensate for damage. While a previous study led us to predict higher storage in roots (Woods et al. 2012) and stems in plants from colder high-latitude sites, stem investment can also be favoured in productive sites where height is advantageous (Poorter et al. 2015). Milkweeds grow taller in low-latitude sites (Woods et al. 2012), so increased stem investment could be necessary in warmer regions with longer growing seasons.

Third, divergent geographic patterns in the phenology of resource allocation could also contribute to differences in compensatory responses to damage, due to the relative timing of plant damage compared to plant development (Fig. 4). For example, early flowering tarweed phenotypes (Madia elegans D.Don) are less able to recover from damage than late flowering phenotypes, presumably because they do not store as many resources below ground (Krimmel and Pearse 2016). The same mechanism may have occurred in our study, where high-latitude plants from colder regions with short growing seasons stored fewer resources in stems and roots and were less able to compensate for damage in terms of overall biomass.

Figure 4. 

Hypothesized relationship between phenology, timing of measurement, and latitudinal patterns. If high-latitude plants (solid lines) have shorter growing seasons, they may grow more quickly early in the season, accruing more biomass and smaller root:shoot ratios. However, if they start to senesce earlier, there will be a window when they have higher root:shoot ratios and lower biomass than low-latitude plants, thus different geographic patterns will be observed depending upon the relative timing of measurements, as indicated by the vertical dotted lines. We suspect the timing of our final measurements corresponded with the left vertical line, and the right vertical line corresponds more with the measurements in Woods et al. (2012).

In contrast to our findings, two previous studies report greater compensation in high-latitude populations (Lehndal and Ågren 2015; Croy et al. 2020). We speculate that there may have been different geographic patterns in resource allocation for plants observed by these three studies, especially because the relative timing of damage varied dramatically. We applied damage once, seven weeks past germination, and harvested plants two weeks later. Furthermore, our study took place in a greenhouse with controlled temperature and photoperiod, which may have eliminated environmental cues that would signal shifts in growth/allocation patterns. In contrast, Croy et al. (2020) applied 40% leaf area removal at monthly intervals four times before allowing regrowth for two months prior to harvest, and the study took place in an outdoor experimental garden. Lehndal and Ågren (2015) allowed natural insect damage on three-year old plants in an outdoor common garden. We agree with Lehndal and Ågren (2015) that explicit manipulation of the timing of damage applied to plants from different regions would inform interpretations about the relationship between phenology and compensation.

We observed patterns relevant to early-life stages of plants; however, it is possible we would have observed greater compensation in high-latitude/low-temperature plants if we had harvested later in the growing season. Plants from high-latitude populations (Woods et al. 2012; Więski and Pennings 2014; Lehndal and Ågren 2015) or higher altitudes (Dostálek et al. 2016) are almost universally smaller (shorter, less massive, fewer leaves) and have more root investment than plants from lower elevations or latitudes. In contrast, by week 10, we observed a regional pattern of constitutively larger plants from high latitudes and colder temperatures, and we saw more stem and root investment in plants from low latitudes and warmer temperatures. Previous research demonstrated that common milkweed plants from high-latitude sites are larger earlier in the season, indicating earlier phenologies, but by the end of season high-latitude plants are smaller than low-latitude plants (Woods et al. 2012). We suspect that high-latitude/low-temperature plants in our study initially grew rapidly, but given more time and the appropriate seasonal cues, we expect that they would have shifted to greater resource storage in roots and stems. In contrast, low-latitude/high-temperature plants may be adapted to longer growing seasons, allocating more resources to stems and roots early in the season to facilitate greater growth over time. This potential pattern of seasonal resource allocation suggests that the regional patterns of compensatory growth observed could depend on the timing, and even the location, of measurements. We posit that the timing of measurement could reverse the regional patterns of compensation observed if high-latitude plants initially grow faster and store fewer resources than low-latitude plants but also if high-latitude plants shift to senescence earlier (Fig. 4). Earlier measurement would generate patterns we observed (larger plants with less storage and less compensation from cold temperatures/high latitudes); whereas later measurement would generate patterns similar to Woods et al. (2012) (smaller plants with more storage from cold temperatures/high latitudes).

Of course, our data are limited, and these hypotheses require more rigorous tests. For example, our evidence demonstrates that grouping by temperature and latitude show similar patterns of compensation, while grouping by precipitation does not. However, other environmental factors, such as nutrient availability, growing season length, and climate variability cannot be ruled out as potential drivers of the patterns we observed.