Intensity and mode of Lindera melissifolia reproduction are affected by flooding and light availability

Abstract We studied the impact of flooding and light availability gradients on sexual and asexual reproduction in Lindera melissifolia (Walt.) Blume, an endangered shrub found in floodplain forests of the Mississippi Alluvial Valley (MAV), USA. A water impoundment facility was used to control the duration of soil flooding (0, 45, or 90 days), and shade houses were used to control light availability (high = 72%, intermediate = 33%, or low = 2% of ambient light) received by L. melissifolia established on native soil of the MAV. Sexual reproductive intensity, as measured by inflorescence bud count, fruit set, and drupe production, was greatest in the absence of soil flooding. Ninety days of soil flooding in the year prior to anthesis decreased inflorescence bud counts, and 45 days of soil flooding in the year of anthesis lessened fruit set and drupe production. Inflorescence bud development was the greatest in environments of intermediate light, decreased in high‐light environments, and was absent in low light environments. But low fruit set diminished drupe production in intermediate light environments as compared to high light environments. Asexual reproduction, as measured by development of new ramets, was greatest in the absence of soil flooding and where plants were grown in high or intermediate light. Plants exhibited plasticity in reproductive mode such that soil flooding increased the relative importance of asexual reproduction. The high light environment was most favorable to sexual reproduction, and reproductive mode transitioned to exclusively asexual in the low light environment. Our results raise several implications important to active management for the conservation of this imperiled plant.


| INTRODUC TI ON
Reproduction by clonal plants has two basic modes-asexual reproduction is accomplished through vegetative formation of a clone, while sexual reproduction is accomplished through seed formation. Asexual reproduction in clonal woody plants is commonly initiated through the development of rhizomes, root suckers, layered branches, or lignotubers (Jeník, 1994). This reproductive mode increases genet size, facilitates resource capture, and maintains genet longevity often under suboptimal or heterogeneous environmental conditions (Hutchings & Wijesinghe, 2008;de Witte & Stöcklin, 2010). Sexual reproduction in clonal woody plants enables dispersal of new genets thereby promoting genetic variation in existing populations and establishment of new populations (Eriksson, 1992;Kanno & Seiwa, 2004;Stöcklin & Winkler, 2004).
While benefits of each reproductive mode are tempered by a range of ecological and energetic costs, the ability to reproduce via two different modes provides clonal plants with reproductive plasticity (Gardner & Mangel, 1999;Herben et al., 2015;Lei, 2010).
Investigations in temperate forests demonstrate that plasticity in expression of reproductive mode by clonal woody plants is driven by changes in components of the forest environment across space and time (Bunnell, 1990;Hewitt, 2020;Hosaka et al., 2008;Kanno & Seiwa, 2004;Moola & Vasseur, 2009;Salter et al., 2010). Variations in the availability of light, soil moisture, and other biotic and abiotic resources can affect resource acquisition, which then influences resource allocation patterns of a plant. Photosynthate allocation to components of vegetative or sexual reproduction supports the primary reproductive mode expressed by the plant in response to its environment. In an old-growth Japanese beech (Fagus crenata Blume) forest, for example, flowering by the understory shrub Hydrangea paniculata Sieb. was limited to disturbed habitats of forest gaps (Kanno & Seiwa, 2004). However, this shrub reproduced almost exclusively through layering where canopy disturbance was lacking (Kanno & Seiwa, 2004). Other studies of temperate forest species found that expression of the two reproductive modes is affected by resource gradients, sexual reproduction especially is affected by light availability (Bunnell, 1990;Eckerter et al., 2019;Hosaka et al., 2008).
Understory environments of floodplain forests throughout the temperate zone are characteristically heterogeneous (Hall & Harcombe, 1998;Küẞner, 2003;Suzuki et al., 2002). Stand development and canopy disturbance dynamics control understory light regimes while alluvial processes active on the floodplain determine variability in edaphic and hydrologic components of the environment. Soil flooding, sediment accretion, and substrate erosion interplay with stand development and canopy disturbance agents, such as windstorms, ice storms, or insect and pathogen outbreaks, resulting in complex and often interacting gradients of resource availability. Several authors have linked species occurrence and growth in floodplain forest understories to plant stress tolerance and resource availability along gradients of flooding and light availability (Battaglia & Sharitz, 2006;Küßner, 2003;Lin et al., 2004;Sakai et al., 1999).
Less is known about how these disturbance-mediated environmental gradients in floodplain forests regulate expression of reproductive mode by understory woody plants capable of reproductive plasticity. However, evidence suggests that the relative expression of sexual and asexual reproduction is not controlled solely by one factor, such as light availability (Hosaka et al., 2008).
Lindera melissifolia (Walt.) Blume, commonly known as pondberry, is a dioecious, rhizomatous, and deciduous shrub in the Lauraceae (Devall et al., 2001). It is endangered but found in wet forests of the southeastern USA, namely, in Alabama, Arkansas, Georgia, Mississippi, Missouri, North Carolina, and South Carolina (Echt et al., 2011). In the floodplain forests of the Mississippi Alluvial Valley (MAV), L. melissifolia forms predominately single-sex colonies in the understory of mixed, deciduous broadleaves (Hawkins et al., 2009;Wright, 1994). The recovery plan developed for L. melissifolia following its listing as an endangered species identified soil moisture and light intensity as key environmental factors to consider and understand better regarding the sustainable management of the species (USFWS, 1993). We wanted to know how these two environmental factors affect plasticity in expression of the reproductive modes in L. melissifolia.
To address our question, we established an experiment to investigate the effects of soil flooding and light availability on drupe (sexual reproduction) and ramet (asexual reproduction) production observations indicate that L. melissifolia anthesis begins prior to leaf out when soil moisture typically is high, and in many instances when soils are flooded. However, we anticipated inflorescence bud formation, fruit set, and drupe production to sharply decrease as the duration of soil flooding increased into the growing season because of disruptions to physiological processes associated with anaerobiosis (Lockhart et al., 2017) (Figure 1b). We also expected ramet production to be limited by soil flooding, but to a lesser extent than drupe production. This is because rhizome and new ramet growth do not appear to occur when the soil is inundated, but likely resumes after floodwater recedes and an aerobic soil environment prevails (Lockhart et al., 2013). Thus, we hypothesized that L. melissifolia would favor sexual reproduction when grown in soil not subject to flooding but favor asexual reproduction as the duration of soil flooding increased into the growing season ( Figure 1b, dashed line).
We also predicted that inflorescence bud formation, fruit set, and drupe production by L. melissifolia would be greatest in a relatively high light environment, and these would decline as light availability decreased because of limitations to photosynthate production (Lockhart et al., 2017) (Figure 1c). Ramet production was also expected to decline with decreasing light availability, but more gradually than drupe production (Lockhart et al., 2013). Accordingly, we hypothesized that L. melissifolia would favor sexual reproduction when grown in high light environments but favor asexual reproduction in environments of relatively low light availability (Figure 1c, dashed line).

| Study species
As noted above, L. melissifolia anthesis begins prior to leaf expansion, typically in late February or early March for MAV populations (Hawkins et al., 2010). The insect-pollinated, yellow flowers that are 5-to 6-mm wide with 2-mm long tepals usually arise in clusters of 3 from umbellate, axillary inflorescences. Drupes mature in August and September to an average of 11-mm long and contain a 6-mm-long seed Hawkins et al., 2010). Field observations indicate that L. melissifolia can produce large crops of drupes, but Devall et al. (2001) noted that seedling establishment rarely has been observed. L. melissifolia reproduces asexually by generating rhizomes that give rise to ramets (Wright, 1990). Information on L. melissifolia ramet biology is sparse, but Wright (1994) suggested asexual reproduction appears to be the dominant form of L. melissifolia regeneration. Also, of consequence to L. melissifolia regeneration, populations are reported to be male biased, with male to female colony ratios in the MAV ranging from 7:1 to 19:1 Wright, 1994).

| Study site
Our study was conducted in a 6-ha impoundment network called the Flooding Research Facility (FRF) on the Theodore Roosevelt National Wildlife Refuge Complex, Sharkey County, Mississippi, USA (32°58′N, 90°44′W, 30 m elevation) (Lockhart et al., 2006). This experimental site is within 5 km of natural L. melissifolia colonies growing on the USDA Forest Service's Delta National Forest. The site lies in a humid, subtropical region of the temperate zone-average daily temperature at the FRF is 17.3°C with a range from 27.3°C in July to 5.6°C in January, and precipitation averages 1,350 mm annually (WorldClimate, 2008). Soil within the FRF is Sharkey clay (very-fine, smectitic, thermic Chromic Epiaquerts), it is alluvial in origin and a predominant soil series in the MAV. The FRF consists of 12, 0.4ha, rectangular impoundments that can be independently flooded or drained to create and control replicates of experimental hydroperiods. Lockhart et al. (2006) present more detailed information regarding design and operation of the FRF.

| Plant material and establishment
Planting stock for this experiment consisted of 20 L. melissifolia genotypes that we collected in the MAV and replicated with tissue culture techniques as described in Hawkins et al. (2007). Rooted cuttings of each genotype were container-grown for about 11 months in greenhouses after which stem length averaged 21.6 ± 0.3 cm (mean ± one standard error) and basal diameter averaged

| Experimental factors
Our experimental design included three levels each of two factors, soil flooding and light availability, used to provide gradients of environmental conditions that could result from two different ily rainfall captured and stored in an adjacent reservoir, but some ground water was used to supplement stored rainfall, as needed.
Flood-water depth was maintained near 12 cm above the soil surface in 2006 and 19 cm above the soil surface in 2007 when experimental plants were taller. Impoundments were drained at the end of each scheduled flood, and ambient rainfall was the only source of soil moisture during nonflooded periods.
We constructed three shade houses in each impoundment to control light availability in a fashion representative of a range of forest canopy cover. A shade house consisted of a 25.6-m long by 7.3-m wide by 2.4-m tall wooden frame covered with neutral density shade cloth (PAK Unlimited, Inc., Cornelia, Georgia, USA). Shade houses were built over areas in each impoundment noted above as plots. Each shade house in an impoundment was randomly assigned a relatively "high," "intermediate," or "low" level of light availability.
We used 30% shade cloth to provide high light, 63% shade cloth to provide intermediate light, and 95% shade cloth to provide low light. Actual light availability measured in shade houses (Lockhart et al., 2013) differed from shade cloth ratings such that plots received an average of 72%, 33%, or 2% of available photosynthetically active radiation for the high, intermediate, and low levels, respectively.
Shade house construction was completed prior to transplanting, so that light availability assignments were in place during field acclimation of plants in 2005.

| Measurements
We measured variables of sexual and asexual reproduction on all fe-   and pedicels (0.29 g) was multiplied by drupe counts to estimate the total dry weight of drupes and pedicels produced by each plant.

| Experimental design and analyses
Reproductive mass (g) was defined as the total weight of drupes and their pedicels, or total weight of ramets for a plant, and reproductive mass ratio (g) was calculated by dividing drupe and pedicle mass by ramet mass.

| Inflorescence bud production and fruit set
Mean inflorescence bud production by L. melissifolia ranged between 0 and 485 per plant showing substantial variation relative to soil flooding and light availability (Table 2). Plants receiving 90 days of flooding produced 33% fewer buds than plants receiving 45 or 0 days of flooding ( Table 2). The extended flood also limited inflorescence bud production per unit of stem to about 77% of that observed for plants receiving the 45-day flood (Table 2).
Inflorescence bud production by L. melissifolia was greatest when plants were grown under intermediate light (Table 2), with plants raised under this light level producing 40% more buds than those raised under high light. But, inflorescence bud counts relative to stem length were equivalent for these two light levels.
L. melissifolia grown under low light did not develop inflorescence buds (Table 2).
Between 1%    Values are means ± standard error, and letters in a column by treatment indicate differences at p ≤.05.  (Table 3).

| Reproductive intensity ratio
The reproductive intensity ratio of L. melissifolia was generally greater than 1, the exception being plants established under low light ( Figure 3). Mean values of reproductive intensity ratio relative to soil flooding levels ranged between 20:1 and 4:1 (Figure 3a).
Plants raised in the absence of flooding (0 days soil flooding) showed the highest reproductive intensity ratio, soil flooding for either 45 or 90 days reduced the ratio by 79% (Figure 3a). The reproductive intensity ratio relative to light availability was greatest  (Figure 3b).

| Reproductive mass
Soil flooding and light availability independently affected total drupe mass of L. melissifolia (Table 4)

| Reproductive mass ratio
L. melissifolia showed reproductive mass ratios less than 1 for all lev-

| Sexual reproduction
We hypothesized that soil flooding would reduce sexual reproductive intensity of L. melissifolia, and this effect would be most pro-  Values are means ± standard error. Capital letters indicate differences within a row for soil flooding means, lower case letters indicate differences in a column for light availability means at p ≤ .05. 2 Test statistics: soil flooding x light availability (F (2,9) = 1.11, p = .37); soil flooding (F (2,9) = 16.57, p = .001); light availability (F (1,9) = 8.34, p = .02).
plants having more reproductive stem length in that environment (Lockhart et al., 2013). Light availability exerts strong influence on sexual reproduction for many clonal, woody plants in temperate understories (Hosaka et al., 2008;Kanno & Seiwa, 2004;Roper et al., 1995). V. myrtillus L. produced a greater number of flower buds, flowers, and fruit in the relatively high light environment of forest gaps compared to closed-canopy locations in southern Germany (Eckerter et al., 2019).
In southeastern Pennsylvania, the proportion of Lindera benzoin (L.) Blume flowers that initiated fruit was the same on control and shaded branches but significantly fewer fruit matured on shaded branches (Niesenbaum, 1993). Flower abundance in Sambucus racemosa L. in the Pacific Northwest, USA was greater in gaps than under intact forest canopies (Wender et al., 2004). Sexual reproduction requires greater photosynthate availability than asexual reproduction (Holsinger, 2000). Kanno and Seiwa (2004)

| Asexual reproduction
We predicted soil flooding would limit asexual reproduction of L. melissifolia by inhibiting ramet production in parallel with flood duration. Likewise, we predicted an increasing limitation on asexual reproduction with decreasing light availability. In contrast to our findings for sexual reproduction, the effects of soil flooding on L. melissifolia asexual reproduction were conditioned by light availability, that is, the two effects were not independent, they interacted.
Ramet production and mass were greatest in the absence of soil flooding and where plants were grown in high or intermediate light. Also, maximal tiller production by the perennial grass, Arundinella hirta (Thunb.) Koidz., occurred at low topographic positions of highest flood duration on a central China floodplain (Zeng et al., 2006). L. melissifolia ramet development relative to light availability patterned somewhat consistently with other temperate woody plants that possess the ability to reproduce via rhizomes or root sprouting.
Most of these clonal plants can generate new ramets across a wide range of light environments, but ramet development often peaks in high-light environments more supportive of plant vigor (Kawamura & Takeda, 2002, 2008Kowarik, 1995). However, Hosaka et al. (2008), who studied A. triloba in temperate broadleaf forests in Maryland, USA, reported that ramet recruitment did not correlate with light availability beneath closed canopy versus canopy gap locations.
Also, clonal plants that rely on other forms of asexual reproduction, such as layering or fragmentation, may show higher rates of ramet development in relatively low-light environments where genet persistence is critical (Kanno & Seiwa, 2004).

| Plasticity in reproductive mode
Plants capable of sexual and asexual reproduction often exhibit reproductive plasticity by favoring one reproductive mode over the other in response to biotic or abiotic environmental factors (Loehle, 1987;Yang & Kim, 2016). We hypothesized that L. melissifolia would exhibit plasticity in expression of reproductive mode along gradients of soil flooding and light availability. No soil flooding was expected to favor sexual reproduction, but reproductive mode would transition to favor asexual reproduction with increasing duration of soil flooding. Likewise, a high light environment was expected to favor sexual reproduction, but reproductive mode would transition to favor asexual reproduction with decreasing light availability.
Our results for reproductive intensity ratio and reproductive mass ratio demonstrate impact of soil flooding on the relative expression of reproductive mode by L. melissifolia. A sharp reduction in drupe production and mass occurred with flooding regardless of light availability. Ramet production and mass also declined with flooding, but the effect was not as sharp as with drupes. As to be expected, reproductive intensity ratio and reproductive mass ratio illustrate opposing results for the relative importance of a reproductive mode.
For L. melissifolia, reproductive intensity ratio appears biased towards sexual reproduction, and reproductive mass ratio appears biased toward asexual reproduction. Nevertheless, both indices reveal that the relative importance of sexual reproduction was greatest in the absence of soil flooding, and this reproductive mode lost considerable importance with 45 or 90 days of soil flooding. Thus, it appears that soil flooding invoked a plastic response in L. melissifolia reproductive mode by increasing the importance of asexual reproduction while decreasing the importance of sexual reproduction.
Other authors reporting a similar observation associate asexual reproduction with the capacity to recover from disturbance and enhanced genet persistence in harsh floodplain environments (Chong et al., 2013;Zeng et al., 2006). Though flooding altered the relative importance of each reproductive mode, we did not find evidence to support our hypothesis that soil flooding would prompt a transition from favoring one reproductive mode to the other. These results may be the first to document the effect of soil flooding on expression of reproductive mode in an understory shrub endemic to temperate floodplain forest habitats.
A gradient of light availability also elicited a plastic response in L. melissifolia reproductive mode. Reproductive intensity ratio and reproductive mass ratio each illustrate that the relative importance of sexual reproduction was greatest in the high light environment, and this reproductive mode lost importance with decreasing light availability. Asexual reproduction increased in relative importance with decreasing light availability. In fact, we observed a transition in reproductive mode to exclusively asexual reproduction in the low light environment, and this observation, to some extent, supports our hypothesis.
A transition in reproductive mode across an environmental gradient of light availability (or surrogates of light availability) has been reported for other understory shrubs of temperate forests including Gaultheria shallon Pursh in Canada (Bunnell, 1990) and H. paniculata Sieb. in Japan (Kanno & Seiwa, 2004). Along an environmental gradient of light, sexual reproduction is often associated with relatively high light availability because processes of flowering and fruit development are energetically costly requiring a highly functional photosynthetic system (Holsinger, 2000). As we observed for L. melissifolia, intensity of sexual reproduction tends to decrease as shading increases, and asexual reproduction becomes the more prominent reproductive mode in clonal plants (Kanno & Seiwa, 2004;Kawamura & Takeda, 2002). In this research, L. melissifolia was capable only of asexual reproduction when grown in the low light environment.
Though minimal, ramet development in this environment enabled plants to increase photosynthetic surface area and capture growing space, both of which foster genet persistence in habitats of suboptimal light (Kanno & Seiwa, 2004;Yang & Kim, 2016).

| Implications to conservation
Our research is the first to examine reproductive biology of L. melissifolia relative to environmental factors prominent in its floodplain habitat. We discovered impacts to L. melissifolia sexual and asexual Fostering regeneration of this species will require vigilance in assessing floodplain inundation events, their impacts to inflorescence bud development, drupe production, and ramet development and growth, and flexibility in application of management practices to support and sustain developing cohorts, either seedlings or ramets, of reproduction.
We also demonstrate responses to sexual and asexual reproductive intensity and plasticity in reproductive mode of L. melissifolia along a gradient of light availability, that is, the light environment of a given L. melissifolia colony will differentially affect intensity and Light availability in understories of mature temperate forests varies spatially and temporally because of canopy gaps and the distribution of foliage among canopy layers (Canham et al., 1990;Runkle, 1982). In the absence of canopy gaps, light available in the understory of mature floodplain forests is minimal averaging <5% of full sunlight over the course of a day (Cunningham et al., 2011;Jenkins & Chambers, 1989). This is likely the case in floodplain forests of the MAV where L. melissifolia grows because these forests develop multi-storied canopies that significantly reduce light transmission to the understory (Hawkins et al., 2009;Oliver et al., 2005).
Forest stand structure can be managed to create light environments that promote L. melissifolia colony vigor and structures conducive to sexual and asexual reproduction.
Our summary of findings presented above highlights the ability of L. melissifolia to respond reproductively to favorable hydrologic and light environments. Still, an additional implication to conservation of this imperiled species may be drawn from our observations of its reproductive plasticity. Along with decreases in reproductive intensity, we report change in the relative importance of reproductive modes expressed by L. melissifolia along soil flooding and light availability gradients. Specifically, we observed an increase in the relative importance of asexual reproduction as soil flooding and low light availability limited sexual reproduction. This reproductive plasticity, which confers persistence in suboptimal habitats, is a trait of high importance to conservation management because it allows for flexibility in the timing and intensity of management activities. For example, management practices aimed at improving the light environment in habitats of multi-storied, closed-canopy floodplain forests could be implemented in methodical stages that sequentially reduce midstory and overstory canopy cover. Such an approach would enable the manager to monitor and react to the L. melissifolia response, as well as the response of competing vegetation or other factors that potentially impact L. melissifolia survival, growth, and reproduction.
Historical deforestation in the MAV-72% loss of forest cover (Gardiner, 2015)-has reduced greatly the availability of potential forest habitat for L. melissifolia. The current state of forest land cover in the MAV may lead managers toward active management to conserve this species. We contribute knowledge of L. melissifolia reproductive intensity and mode that is essential to the long-term conservation of this species in floodplain forests of the MAV.

ACK N OWLED G M ENTS
We thank Tom Dell and Eric Zenner for statistical advice. We also

Department of Interior, Fish and Wildlife Service permit number
Endangered-Threatened Species Sub-permit SA0142-Amendment 3.

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

DATA AVA I L A B I L I T Y S TAT E M E N T
Original data pertaining to this research are available on the Dryad