We used G. longituba as the model in the experiment. G. longituba is a perennial herb widely distributed in forests, along roadsides or creeks of tropical, subtropical, and temperate areas in China, Vietnam, South Korea, and Russian Far East (Zhang and He, 2009). The species naturally occurs under the canopy, so it experiences environment with heterogeneous UV-B distribution. In the wild, the genus Glechoma has two different growth forms. In its flowering phase, from March to June, it produces a vertical stem while it produces plagiotropic, monopodial stolons in vegetative phase. The monopodial stolons are able to creep on the ground and the stolons bear nodes at intervals of 5–10 cm. A pair of opposite, orbicular leaves is produced at each node. Adventitious roots may also develop at the nodes. There is a bud in the axil of each leaf, which may generate a higher-order stolon only when several younger nodes have been produced along the parent stolon (Birch and Hutchings, 1994; Liao et al., 2003). Since the G. longituba usually does not flower in the greenhouse, it grew only horizontal stolons in our experimental setting. The G. longituba was collected from Jiwozi in Qinling Mountains, Shaanxi, China (33° 47′N, 108°45′ E).

We selected a single genotype of G. longituba that was propagated in a greenhouse for 6 months. When we had enough plant material, we severed and replanted 64 youngest ramets (further considered as parental ramets in the study) of comparable size and planted them individually to pots (7 × 7 × 7 cm) filled with soil (25% sand, 25% organic matter, and 50% peat). We allowed the ramets to recover and root for 7 days before we started the main study. The experiment was conducted in a greenhouse from January to March of 2019.

Seven days after parental ramets were planted to pots, we randomly subjected them to two light training environments, Control light environment (further referred to as Control group, N = 32), or UV-B radiation for 15 days (further referred to as UV-B group, N = 32). The UV-B treatment is described later in the text. After 15 days, the UV-B radiation of the parental ramets was terminated. The newly emerging offspring ramets developing from the parental ramets of both parental light treatments (all parental ramets had single emerging offspring ramet at this point of the study) were allowed to grow in a narrow plastic runway (2 cm wide) for 30 days. Two plastic trays (54 cm long × 28.5 cm wide × 7 cm high) filled with soil (25% and, 25% organic matter, and 50% peat) were placed on the two sides of the runway. The tray on one side of the runway was exposed to Control light environment (Control offspring environment) and the other received additional UV-B radiation (UV-B offspring environment) (Figure 1A). Two parental ramets (one UV-B and one Control) were sharing the same UV-B source (Figure 1B) and thus represented a block. At the beginning, the growth direction of all plant individuals was parallel to the runway, and newly developed interconnected ramets (further referred to as offspring ramets) faced Control light from one side and UV-B radiation from the opposite side. As the growth of the individual continued, it turned its elongating stolon either in one or the other direction, thus selecting the UV-B or Control environment. None of the main stolons remained in the runway—so each main stolon selected one or the other side. In our study, all plants produced only the main stolon but one in which we removed the secondary stolon. To prevent the UV-B radiation penetration to the side of control offspring light environment, a UV-B baffle (0.3 mm transparent polyester film, Dongguan Linuo Plastic Insulation Material Co. LTD, Guangdong, China) was settled between the two types of the light environments (Figure 1). The bottom of the baffle was 2-cm above the plant, ensuring that the newly emerged ramets can sense different light conditions on both sides of the baffle.

During the experiment, the mean irradiance in the greenhouse was 357 ± 20 μmol m^–2 s^–1, and humidity was 40 ± 5% with a 14 h/10 h light/dark cycle and a 25/20°C day/night temperature cycle. Plants were watered every 2 days to prevent water stress.

Following the method of Liu et al. (2015), UV-B radiation was supplied by square-wave UV-B fluorescent lamps (36 W, Beijing Lighting Research Institute, Beijing, China). The maximum output of these lamps was 313 nm. The lamps were wrapped with 0.13 mm cellulose acetate film (Grafix Plastics, Cleveland, OH, transmission down to 290 nm) for the supplemental UV-B radiation group (UV-B group). The lamps were active for 8 h per day from 9:00 to 17:00. The daily radiation dose was 2.88 kJ m^–2 d^–1. Unlike the parental ramets that were exposed to UV-B for 15 days, the offspring ramets were exposed to UV-B light radiation for 45 days if they have chosen the UV-B light environment. The amount of UV-B radiation was measured using a UV radiometer (Handy, Beijing, China) every 2 days. The cellulose acetate film was replaced every 5th day.

During the experiment, offspring ramets of G. longituba grew either into the Control light conditions or into the UV-B light environment (all plants had only the main stolon, so they had only one choice for UV-B or Control environment in our experimental setup). After 45 days (the end of the study), we count the number of offspring selected UV-B or Control environment to calculate the proportion of offspring ramets selection/foraging and measured a range of plant size parameters. First, we recorded biomass, leaf area, and specific leaf area of the parental ramets of every individual, i.e., the part that did not enter the pathway. For the offspring part, we recorded total biomass, total number of offspring ramets, stolon length, total leaf area, and specific leaf area of the whole offspring part of the individual. The biomass includes only above-ground biomass. To assess leaf area, fresh leaves were scanned with a scanner (EPSON Perfection V19, EPSON, China), and leaf area was calculated with Motic software (Motic Images Plus 2.0. Ink, Motic, China). Above-ground biomass was dried for 72 h at 80°C until constant weight and weighted immediately using an electronic balance (SartoriusBT25S, Beijing, China).

Methylation alterations in cytosine modification of G. longituba were detected using methylation-sensitive amplification polymorphism (MSAP) analysis. For the analyses, we sampled leaves from parental ramet, older offspring (the third offspring ramet counted from the parental ramet on the main stolon, later referred to as older offspring), and the youngest fully developed offspring ramet (the last offspring ramet counted from the parental ramet on the main stolon, later referred to as younger offspring). For each offspring type and combination of parental and offspring environment, we had 3–6 individuals. This unbalanced design was given by the foraging decisions of the plants. The samples were scrubbed gently with 75% ethanol minimize contamination by microorganisms and then dried in silica gel for the subsequent extraction of DNA. This allowed testing whether epigenetic memory of UV-B radiation is transmitted trans-generationally among clonal offspring (ramets). Total genomic DNA was extracted from 30 mg of dry leaves using BioTeKe (Beijing, China), DNA quality was examined by electrophoresis in agarose gel 1% (w/v), and DNA concentration and purity were examined spectrophotometrically with NanoDrop2000 (Thermo Fisher Scientific, United States). The qualified DNA was diluted to the same concentration (100 ng/μl) for MSAP analysis. We used the endonuclease combination 1 μl of EcoRI + 1 μl of HpaII (E + H) (NEB, United States) and 1 μl of EcoRI + 2 μl of MspI (E + M) (NEB, United States) to double-enzyme genomic DNA, and the digested ends were ligated with 1 μl of HpaII-Msp-adapter (50 pmol/μl), 1 μl of EcoRI adapter (5 pmol/μl) (Biotech, China), and 0.5 μl of T4 DNA ligase (TAKARA, Japan). Both the digestion and ligation reactions were performed in a final volume of 20 μl. The enzyme was cut at 37°C for 5 h. Connect at 8°C for 4 h.

The 2 × Taq PCR master mix and pre-amplification and selective amplification primers used in the experiments were synthesized by Shanghai Biotech (Supplementary Appendix Table 1). Both the reactions of pre-amplification and selective amplification were in a final volume of 50 μl. A pre-amplification step was carried out with EcoRI pre-amplification primers and HpaII/MspI pre-amplification primers. The PCR mix contained 2 μl of ligated DNA, 21 μl of double-distilled water, 1 μl of H-M pre-amplification primers (10 μM), 1 μl of EcoRI pre-amplification primers (10 μM), and 25 μl of 2 × Taq PCR master mix. The pre-amplification conditions were as follows: 72°C for 2 min; 94°C for 2 min; 20 cycles at 94°C for 30 s, 56°C for 30 s, and 72°C for 1 min; and a final elongation step at 72°C for 10 min.

The pre-amplification products were diluted 10 times as a selective amplification template. A selective amplification step was carried out with 11 pairs of selective primer combinations, including the following: EcoRI-AAG/HpaII-TGA, EcoRI-AAG/ HpaII-TTA, EcoRI-AAG/ HpaII-TTG, EcoRI-ACT/ HpaII-TCC, EcoRI-ACT/ HpaII-TTG, EcoRI-AGG/ HpaII-TTC, EcoRI-AGG/ HpaII-TGA, EcoRI-AGG/ HpaII-TCC, EcoRI-AGG/ HpaII-TTG, EcoRI-ACG/ HpaII-TTG, and EcoRI-AGC/ HpaII-TCC. The PCR mix contained 1 μl of pre-amplified DNA, 22 μl of double-distilled water, 1 μl of H-M selective primer (10 μM), 1 μl of EcoRI selective primer (10 μM), and 25 μl of 2 × Taq PCR Master Mix. The selective amplification conditions were as follows: 94°C for 2 min; 10 cycles at 94°C for 30 s; 65°C for 30 s and 72°C for 1 min (each cycle is decremented by 1°C); 23 cycles at 94°C for 30 s; 56°C for 30 s and 72°C for 1 min; and a final elongation step at 72°C for 10 min.

Before polyacrylamide gel, the selective amplification product was inactivated at 70°C for 10 min, then the selective amplification samples were separated by 10% denaturing polyacrylamide gel electrophoresis and subjected to electrophoresis at 220 V for 4 h, and the gel was applied with silver staining. Following staining of the gel, rinsing, developing, and photographing were performed, and band statistical analysis was performed. Fragments from approximately 100–500 bp were scored. The amplified MSAP products were resolved using the method described in Xu et al. (2016).

A total of 105 MSAP loci were amplified from each individual using 11 primer pair combinations. When MSAP profiles of all ramet types were analyzed together (i.e., parental and older and younger offspring ramet), loci composition, epigenetic diversity (Shannon’s diversity index), and overall DNA methylation level were significantly affected by parental training environment (Control vs. UV-B, Table 2, Pa). Shannon’s diversity index (I_epi) and overall DNA methylation level (L_epi) were significantly lower for ramets of UV-B trained parents (I_epi = 3.82 ± 0.22; L_epi = 44.34 ± 0.09%) than Control trained parents (I_epi = 3.97 ± 0.14; L_epi = 50.86 ± 0.07%). Loci composition, but not diversity and methylation level, was significantly affected by ramet type. Loci composition significantly differed between parental training environment and ramet type (Table 2, Pa × Ramet). Therefore, we consequently tested the effects of parental and offspring environment on each ramet type separately.

In case of the parental ramet, the loci composition (F = 1.81; P = 0.003; df = 1.16), Shannon’s diversity index (Control = 4.00 ± 0.07, UV-B = 3.70 ± 0.25; F = 11.78; P = 0.003), and total DNA methylation level (Control = 52.28 ± 0.04%, UV-B = 39.58 ± 0.10%; F = 13.32; P = 0.002) significantly depended on the parental training environment.

When analyzing offspring ramets (older offspring ramet and younger offspring ramet), we found only loci composition to be significantly affected by parental training environment (Table 3). Offspring environment had no effect on any of the variables.

We demonstrate that the 15-day long training of parental ramets of G. longituba to increased UV-B radiation has a negative effect on their growth (biomass) as well as on the number and biomass of offspring ramets and hence on the fitness of the whole individual. This is in line with several other studies (Frohnmeyer and Staiger, 2003; Kakani et al., 2003; Vanhaelewyn et al., 2016; Dotto and Casati, 2017). However, our study provides an additional unique finding that the experience of parental ramet with UV-B radiation strongly affects consequent foraging behavior for light of the growing individual. Individuals with the parental ramet’s experience with UV-B radiation preferentially placed offspring ramets to light conditions without UV-B radiation in comparison to the genetically identical individuals without the parental experience with UV-B radiation (see Figure 3). Such behavior probably helped to mitigate negative consequences of UV-B radiation on the individual’s fitness. Avoiding patches with high UV-B levels for already weakened (offspring ramets of UV-B stressed parental ramet) individuals may reduce further negative impact of UV-B radiation on fitness of the genets. Such a behavior can be considered as an escape strategy, which has been also documented in other clonal species (de Kroon and Hutchings, 1995; Ye et al., 2006; Ikegami et al., 2007; Puijalon et al., 2008).

Our study adds to the mounting evidence that the behavior of clonal plants is not independent on their environmental interactions in the past. For example, Louapre et al. (2012) showed that the foraging behavior of clonal plants Potentilla reptans and Potentilla anserina is affected by the nutrient availability in older ramets. Latzel and Münzbergová (2018) found that clonal plant Fragaria vesca is able to store information on the light and nutrient availability of older ramets and based on this information decide where to place offspring ramets, which they consider as an exhibition of anticipatory behavior in clonal plants (Latzel and Münzbergová, 2018). A puzzling question is which mechanisms allow for the memory on the past environmental interactions and consequent change in foraging behavior of clonal plants.

It is very likely that the information passed from parental to interconnected offspring ramets is enabled by their connection via stolons. Some researchers suggest that the connection between ramets allows not only for transport of water with dissolved assimilates but also for transmission of signaling molecules like phytohormones (e.g., Alpert and Mooney, 1986; Hutchings, 1999; Stuefer et al., 2004; Gómez et al., 2008; Louapre et al., 2012; Waters and Watson, 2015). Hence, parental ramet can communicate with offspring ramets (and vice versa), which can alter overall behavior of the genet. Another theory suggests that it is shared epigenetic memory that can be involved in the behavior of clonal plants (González et al., 2016, 2018; Latzel et al., 2016). It has been reported that memories on the environmental interactions can be stored and transmitted to next generations via epigenetic change such as the change in DNA methylation (Molinier et al., 2006; Bossdorf et al., 2008; Boyko et al., 2010; Verhoeven et al., 2010; González et al., 2016; Latzel et al., 2016; Richards et al., 2017). Latzel et al. (2016) suggested that parental ramet can carry epigenetic information about its experiences with environmental interactions and pass the information to its offspring ramets. Hence, behavior of clonal plants can be strongly influenced by epigenetic memories on the past environments (Latzel et al., 2016).

In this study, we demonstrated that UV-B radiation significantly reduced DNA methylation level and Shannon’s diversity index of parental ramets. Similar reduction in DNA methylation due to increased UV radiation was reported also for Zea mays (Steward, 2002; Sokolova et al., 2014), Picea abies (Ohlsson et al., 2013), and Artemisia annua (Pandey and Pandey-Rai, 2015; Pandey et al., 2019). Moreover, the MSAP analysis revealed that the loci composition of parental ramets that experienced UV-B radiation significantly differed from parental ramets subjected to control light conditions. Both results suggest that DNA methylation change was involved in response to the UV-B stress. Our study also provides some evidence that the UV-B-induced DNA methylation variation can be, to some degree, passed to connected clonal offspring ramets and involved in the change of foraging behavior. Different loci composition triggered by the parental UV-B treatment was detected in both older and younger offspring ramets. In addition, the observed reduced level of DNA methylation of parental ramets was inherited by older but not younger offspring ramets (Table 3). A similar pattern was detected for Shannon’s diversity index. This suggests that the epigenetic memory can be passed from parental to offspring ramets through mitotic cell division, but is gradually degrading during clonal growth (i.e., after several asexual generations, in our case, the younger ramet was usually the fourth clonal generation derived from the older offspring ramet). A similar conclusion was reached by Shi and colleagues (Shi et al., 2019) on a clonal plant alligator weed (Alternanthera philoxeroides). They found that environmentally induced epigenetic variation is gradually resetting when plants of different populations (environments) are transplanted to a common garden. After 10 asexual generations and 2 years of cultivation in a common garden, plants of previously different epigenetic profiles become epigenetically comparable (Shi et al., 2019). Our findings thus suggest that the foraging behavior of clonal plants might be at least partly under epigenetic control, which supports the model of epigenetically coordinated advanced behavior of clonal plants described by Latzel et al. (2016). However, better insights into the role of epigenetic memory in the observed changes in foraging behavior require more sophisticated molecular methods such as whole genome bisulfite sequencing (Richards et al., 2017).