The endosperm, a tissue that nourishes the embryo in the seeds of flowering plants, is often disrupted in inviable hybrid seeds between closely related species. A key question is whether parental conflict is a major driver of this common form of reproductive isolation.  Here, we performed reciprocal crosses between pairs of three monkeyflower species (Mimulus caespitosa, M. tilingii, and M. guttatus). The severity of hybrid seed inviability varies among these crosses, which we inferred to be due to species divergence in effective ploidy. By performing a detailed time series of seed development, we discovered parent-of-origin phenotypes that provide strong evidence for parental conflict in shaping endosperm evolution. We found that the chalazal haustorium, a tissue within the endosperm that occurs at the maternal-filial boundary, shows pronounced differences between reciprocal hybrid seeds formed from Mimulus species that differ in effective ploidy. These parent-of-origin effects suggest that the chalazal haustorium might act as a mediator of parental conflict, potentially by controlling sucrose movement from the maternal parent into the endosperm.  Our study suggests that parental conflict in the endosperm may function as a driver of speciation by targeting regions and developmental stages critical for resource allocation and thus proper seed development.

Generation of Plant Material

Here, we used one inbred line (formed from ≥8 generations of self-fertilization) for each focal species (M. caespitosa, M. tilingii, and M. guttatus). The same inbred lines were used in previous studies of hybrid seed inviability in M. tilingii and M. guttatus (Garner et al., 2016) and M. caespitosa (Sandstedt et al., 2021). The M. caespitosa inbred line, TWN36, originates from a high-alpine population at 1594m in Twin Lakes, WA. The M. tilingii inbred line, LVR1, is derived from a population at 2751m in Yosemite Park, CA. The M. guttatus inbred line, DUN10, originates from a population in the Oregon Dunes National Recreation Area. 

In this study, we considered three intraspecific crosses (CxC, TxT, and GxG, where C = M. caespitosa, T = M. tilingii, and G = M. guttatus) and six interspecific crosses (CxT, TxC, TxG, GxT, CxG, GxC; maternal parent is always listed first). To generate diploid, experimental plants, we sowed 20-30 seeds for each inbred line on damp paper towels in petri dishes sealed with parafilm and cold-stratified them for 7 days to disrupt seed dormancy. After cold stratification, we transferred petri dishes to a growth chamber with 16-h days at 23°C and 8-h nights at 16°C. We transplanted seedlings into 3.5” pots with moist Fafard 4P growing mix (Sun Gro Horticulture, Agawam, MA) and placed the pots in the same growth chamber. Once plants began flowering, we randomly crossed within and between individuals (total plants: C = 22, T = 20, G = 16). For all crosses, we emasculated the maternal plant 1-3 days prior to each cross to prevent contamination from self-pollination.

To investigate species divergence in effective ploidy, we performed several interspecific, interploidy crosses: C4xxT, TxC4x; T4xxG, GxT4x; C4xxG, GxC4x (4x subscript indicates tetraploid). To generate synthetic tetraploid individuals, we treated 100-200 seeds of TWN36 and LVR1 with 0.1% or 0.2% colchicine and stored them in the dark for 24 hours (16 hours at 23°C and 8 hours at 16°C). The next day, we planted seeds onto Fafard 4P potting soil using a pipette and placed pots inside the growth chamber under typical light and temperature conditions (16-h days at 23°C and 8-h nights at 16°C). Once seeds germinated, we transplanted seedlings into 2.5” pots. After sufficient growth, we prepared samples for flow cytometry using a protocol adapted from Lu et al., 2017. Briefly, we extracted nuclei from one colchicine-treated sample and an internal control (2n Mimulus or Arabidopsis thaliana, Col-0) together in a single well. To extract nuclei, we chopped 100mg of leaf tissue (50mg colchicine-treated sample and 50 mg internal control) in 1mL of a pre-chilled lysis buffer (15mM Tris-HCl pH 7.5, 20mM NaCl, 80mM KCl, 0.5mM spermine, 5mM 2-ME, 0.2% TritonX-100). We stained nuclei with 4,6-Diamidino-2-phenylindole (DAPI), filtered nuclei for debris using a 40um Flowmi™ cell strainer, and aliquoted nuclei into a single well of a 96-well polypropylene plate. We assessed ploidy of each sample using a CytoFLEX (Beckman Coulter Life Sciences) flow cytometer. We calculated total DNA content using the following equation:

We generated three synthetic polyploids for TWN36 and six for LVR1. For each synthetic polyploid, 2C DNA content was nearly doubled compared to corresponding diploid lines (TWN36, 2C = 1.38 pg; TWN364x, 2C = 2.69  0.09 pg; LVR1, 2C = 1.26 pg; LVR14x, 2C = 2.64  0.05 pg). In some cases, we discovered that plants initially identified as tetraploid via flow cytometry were actually mixoploids. To ensure the crosses we performed were indeed interploidy, we determined the ploidy of the resulting progeny. From each interploidy cross, we planted 5-10 seeds per fruit, isolated nuclei from the resulting plants, and assessed 2C content using a flow cytometer for a few offspring as described above (3x TWN364xxLVR1, 2C = 1.92  0.04; 3x LVR1xTWN364x, 2C = 1.88  0.01 pg; 3x LVR14xxDUN10, 2C = 1.95  0.04 pg; 3x DUN10xLVR14x, 2C = 1.81  0.01 pg; 3x TWN364xxDUN10, 2C = 1.90  0.011 pg). We included data from interploidy crosses only when their progenies were confirmed to be triploids, or, in the case of 4x M. caespitosa, if we were using a confirmed stable polyploid line (i.e., self-fertilized at least one generation with polyploidy confirmed in the progeny). 

Measuring seed size and seed viability 

To measure seed size, we collected three replicate fruits per cross, with each fruit collected from a distinct plant. We imaged 50 seeds per fruit under a dissecting scope, for a total of 150 seeds per cross (except for one CxG fruit for which only 35 seeds were measured for a total of 135 seeds). Seed area was measured using ImageJ (Rasband, 1997). 

Using these same fruits, as well as fruits from interploidy crosses (2-5 fruits/cross, at least two fruits per cross from a distinct plant), we assessed seed viability using two different methods. First, we performed visual assessments of mature seeds, looking for irregular phenotypes (shriveled, wrinkled, or flat) known to be highly correlated with germination rates in these Mimulus species and their hybrids (Garner et al., 2016, Sandstedt et al., 2021). We scored the number of seeds that appeared round and plump (i.e., fully-developed) versus irregularly shaped (i.e., under-developed). Second, we performed Tetrazolium assays to assess seed viability on a subset of these same seeds (~100 seeds per fruit). For fruits generated from interploidy crosses and fruits that produced <100 seeds, we stained 32-63 seeds. We immersed seeds in a scarification solution (83.3% water, 16.6% commercial bleach, and 0.1% Triton X-100) and placed them on a shaker for 15 minutes. After scarification, we washed seeds five times with water and incubated seeds with 1% Tetrazolium at 30°C. Two days later, we scored the number of seeds that stained dark red (viable) versus pink or white (inviable). As noted above, we also planted a subset of the seeds from interploidy crosses to assess ploidy and germination rates generally reflected both seed viability measurements (data not shown).

Seed viability rescues

To assess whether aberrant endosperm development contributes to seed defects in interspecific crosses, we attempted to rescue seed viability with a sucrose-rich medium. We collected three fruits 8 to 12 days after pollination (DAP) from each intra- and interspecific cross (not including interploidy crosses), with each fruit collected from a distinct plant. Of the three fruits per cross, at least one fruit was collected 8 DAP (to maximize the chance of rescue). On average, we dissected 40 whole immature seeds per fruit (range = 25-57) and placed them on petri dishes with MS media containing 4% sucrose. We sealed petri dishes with parafilm and placed them at 23°C with constant light for 14 days before scoring germination.

Visualizing parent-of-origin effects during seed development 

To compare trajectories of seed development, we performed intra- and interspecific crosses, and we collected fruits 3, 4, 5, 6, 8, and 10 DAP. For consistency, we performed crosses and collected fruits at the same time of day.

To visualize early seed development, we collected fruits 3 and 4 DAP (N = 1 to 2 fruits per DAP per cross) and prepared them for clearing with Hoyer’s solution. We placed developing fruits in a 9 EtOH: 1 acetic acid fixative overnight. The following day, we washed fruits twice in 90% EtOH for 30 min per wash. We dissected immature seeds directly from the fruit onto a microscope slide with 100uL of 3 parts Hoyer’s solution (70% chloral hydrate, 4% glycerol, 5% gum arabic): 1 part 10% Gum Arabic and sealed the slide with a glass cover slip. We stored the microscope slides containing cleared, immature seeds at 4°C overnight. The next day, we imaged slides using the differential interference contrast (DIC) setting with the 20x objective on a Leica DMRB microscope. For each fruit, we scored the number of developing seeds with and without an intact chalazal haustorium (15-56 seeds per fruit; 32-111 seeds per cross per DAP); only seeds with visible embryos were scored. Additionally, we imaged an average of 11 seeds per fruit (3-15 seeds per fruit, 10-27 seeds per cross per DAP) to assess size differences in the endosperm and chalazal haustorium at 3 and 4 DAP. For the interploidy T4xxG cross, we imaged on average 18 seeds per fruit (14-26 seeds per fruit, 29-40 seeds per cross per DAP). We outlined and measured the endosperm in all seeds and the chalazal haustorium when present using ImageJ (Rasband, 1997). Because the chalazal haustorium was not present for all imaged seed, sample sizes for its measurements were lower. We selected and measured images that represented typical seed development at each time point.  

We defined the chalazal haustorium as two uninucleate cells that, together, form a continuous structure that penetrates toward the ovule hypostase cells (a group of tightly packed cells at the base of the ovule). To measure the chalazal haustorium, we began the outline near the epidermis of the seed (not including the hypostase cells) and extended it toward the micropylar region following Guilford & Fisk, 1952 (see their Figure 27). In addition, when measuring the endosperm, we started the outline at the same position near the epidermis of the ovule and extended it toward the opening of the micropylar haustorium.

To visualize later seed development (after 4 DAP when the seed coat is too thick to clear with Hoyer’s solution), we collected whole fruits at 5, 6, 8, and 10 DAP and stored them in a Formaldehyde Alcohol Acetic Acid fixative (10%:50%:5% + 35% water) for a minimum of 48 hours. After fixation, we dehydrated developing fruits with increasing concentrations of Tert Butyl Alcohol. Next, we washed fruits three times for two hours each with paraffin wax at 65°C before embedding them into a wax block. We sectioned wax blocks containing whole fruits into ribbons using a LIPSHAW Rotary Microtome (Model 45). Fruits collected at 5 and 6 DAP were sectioned into 12-um ribbons for better visualization of micropylar and chalazal domains, and fruits collected at 8 and 12 DAP were sectioned into 8-um ribbons. Next, we gently placed ribbons in a warm (~40°C) water bath and positioned them onto a microscope slide. We placed slides on a slide warmer overnight to adhere sections completely to the glass. In a staining series, we first used Xylene as a clearing agent and performed several washes with increasing concentrations of EtOH to effectively stain nuclei and cytoplasm (1% Safranin-O and 0.5% Fast Green, respectively). We further washed stained slides with EtOH and finished the series with Xylene. We sealed slides with a glass coverslip using Acrytol as the mounting medium.

We visualized slides using a Zeiss Axioskop 2 microscope with a 10x objective. For each fruit, we imaged at least 10 seeds with a developing embryo per fruit (except for severe embryo-lethal crosses: 10 DAP TxG, 8 seeds imaged; 10 DAP CxG, 1 seed imaged). We imaged at least five consecutive sections of each seed through the embryo. For all seeds imaged at 5 and 6 DAP, we scored the presence of the chalazal haustorium. Additionally, we categorized embryo development at 6, 8, and 10 DAP into four different stages: before globular to globular, late-globular to transition, early-heart to late-heart, and torpedo.