Pre‐ and postzygotic mechanisms preventing hybridization in co‐occurring species of the Impatiens purpureoviolacea complex

Abstract In the species‐rich genus Impatiens, few natural hybrids are known, even though closely related species often occur sympatrically. In this study, we aim to bridge the gap between micro‐ and macro‐evolution to disentangle pre‐ and postzygotic mechanisms that may prevent hybridization in the Impatiens purpureoviolacea complex from Central Africa. We analyzed habitat types, species distribution, pollination syndromes, pollinator dependency, genome sizes, and chromosome numbers of seven out of the ten species of the complex as well as of one natural hybrid and reconstructed the ancestral chromosome numbers of the complex. Several species of the complex occur in sympatry or geographically very close to each other. All of them are characterized by pre‐ and/or postzygotic mechanisms potentially preventing hybridization. We found four independent polyploidization events within the complex. The only known natural hybrid always appears as single individual and is self‐fertile. But the plants resulting from self‐pollinated seeds often die shortly after first flowering. These results indicate that the investigated mechanisms in combination may effectively but not absolutely prevent hybridization in Impatiens and probably occur in other genera with sympatric species as well.

Probably, the evolution in these genera included the development of effective mechanisms to prevent hybridization. In general, two classes of such mechanisms can be distinguished but often a combination of different barriers exists (Seehausen et al., 2014;Sobel & Chen, 2014): First, prezygotic mechanisms include for example geographical or habitat isolation as well as adaptation to different groups of pollinators, temporal variation in flowering time, and different reproductive systems (Arnold, 1997;Bradshaw & Schemske, 2003;Lumaret et al., 1987;Neri et al., 2017). Second, postzygotic mechanisms include for example inhibition of pollen tube growth, failure of normal seed development, and a reduced seed fertility and seedling fitness among other mechanisms (Lafon-Placette & Köhler, 2016;Lee et al., 2008;Merlin & Grant, 1986). Abortion or reduced fertility of seeds is often caused by failure of endosperm development, for example, if mother and father plants have unequal chromosome numbers (Birchler, 2014;Husband & Sabara, 2004;Ramsey & Schemske, 1998). Especially in species-rich lineages, highly diverse chromosome numbers often occur in combination with differences in genome sizes (e.g., Cai et al., 2019;Escudero et al., 2012;Han et al., 2020;Mota et al., 2016). Additionally, differences in other characters, such as habitat type and pollinator group (Glennon et al., 2012;Sobel et al., 2010), commonly exist in these species. All of these differences represent effective mechanisms preventing hybridization in many groups of plants (Birchler, 2014;Sobel et al., 2010), but few studies exist analyzing different mechanisms in larger clades with co-occurring species, even though such studies would provide deep insights into the evolution of these clades.
The species-rich genus Impatiens (Balsaminaceae; >1000 species) is an ideal group to study mechanisms potentially preventing hybridization (Janssens et al., 2009). It occurs mostly in the humid forests of the tropics and subtropics in Africa and Asia (Grey-Wilson, 1980a). In these habitats often several, sometimes even closely related Impatiens species occur sympatrically (e.g., Janeček et al., 2015;Kato et al., 1991;Ruchisansakun et al., 2016). However, until now only few natural hybrids have been found Grey-Wilson, 1980b,c;Tsukaya, 2004). Most of these hybrids occur in disturbed places in small to medium-sized populations (Grey-Wilson, 1980b,c). Furthermore, the proposed hybrid origin of several Impatiens species (Grey-Wilson, 1980b,c) has never been demonstrated and seems unlikely based on the recent molecular analyses on the genus (e.g., Janssens et al., 2009) and a few hybridization studies (e.g., Merlin & Grant, 1986;Ornduff, 1967;Tsukaya, 2004).
Due to the rare nature of hybrids but large numbers of cooccurring species, we can conclude that strong mechanisms preventing hybridization must exist in Impatiens. However, mostly prezygotic mechanisms have been studied in Impatiens: Besides isolation by geography and habitat type (Merlin & Grant, 1986) a common element preventing hybridization in Impatiens are switches between pollinator groups in closely related Impatiens species (Grey-Wilson, 1980a;Janeček et al., 2015;Lozada-Gobilard et al., 2019).
Additionally, Ruchisansakun et al. (2016) demonstrated that within the same habitat a group of species with asymmetric flowers-all pollinated by the same assemblage of bees-do not hybridize because each species deposits its pollen on different parts of the bee´s bodies.
In addition to the mentioned prezygotic mechanisms, also postzygotic mechanisms must exist in Impatiens. For example, Impatiens glandulifera and I. balfourii, two neophytic species occurring side by side in southern Europe, get visited by the same species of bumblebees (Ugoletti et al., 2013). Regular occurrence of heterospecific pollen on the stigmas inducing seed formation is documented (Ugoletti et al., 2013). However, no hybrids are known because hybrid seeds mostly fail to germinate in crossing experiments (Ugoletti et al., 2013). Consequently, strong genetic barriers probably exist between these distantly related species, preventing hybridization.
Differences in chromosome numbers are likely the reason for unsuccessful hybridization between previously mentioned I. glandulifera (2n = 18) and I. balfourii (2n = 14; Song et al., 2003). Similar to this example different chromosome numbers probably occur in many other sympatric Impatiens species because a large diversity of chromosome number is known within Impatiens (2n = 6 to 2n = 200 with a majority of species with 2n = 14 to 2n = 20; Jeelani et al., 2010;Song et al., 2003). However, chromosome number evolution has not systematically been studied in Impatiens.
A promising group to study mechanisms potentially preventing hybridization in closely related species is the Impatiens purpureoviolacea complex endemic to the mountain rainforests of the northwestern Albertine Rift Valley (in Rwanda, Burundi, and the Democratic Republic of the Congo). It originated in the Pliocene and started diversifying during the transition of Pliocene and Pleistocene, possibly triggered by an increased mountain uplifting and volcanic activity in the Albertine Rift . The clade consists of ten species that partly occur sympatrically or geographically close to each other. Most of them show a butterfly/long-tongued bee pollination syndrome with long, filiform, strongly enrolled flower spurs (Abrahamczyk et al., 2017;Fischer et al., 2021). Only two species have bucciniform spurs and are likely pollinated by birds . Even though several species of the Impatiens purpureoviolacea complex occur sympatrically or geographically close to each other and flower simultaneously only a single, rarely occurring hybrid is known . Therefore, we can assume that strong mechanisms preventing hybridization exist.
Here, we analyze mechanisms possibly preventing hybridization of seven out of ten species and one natural hybrid from the Impatiens purpureoviolacea complex. We study prezygotic (habitat types, geographical distribution, pollination syndromes, and pollinator dependency) and postzygotic (chromosome numbers and genome sizes) mechanisms that may prevent hybridization with special focus on the sympatric/geographically close species and put the traits into a phylogenetic context. Specifically, we form the following hypotheses: 1. Pre-and postzygotic mechanisms exist in the Impatiens purpureoviolacea complex that may prevent hybridization.
2. Co-occurring species are always separated by at least one pre-or postzygotic mechanism.

| Plant material
This study benefits from the extensive sampling by Eberhard Fischer since 1984 resulting into a recent revision of the Impatiens purpureoviolacea complex . Seven out of ten spe-

| Autonomous self-pollination tests and pollination syndromes
Autonomous self-pollination tests were conducted to test whether individual species of the Impatiens purpureoviolacea complex depart from the common pollinator dependency in most African Impatiens species (Lozada-Gobilard et al., 2019), which would represent a strong mechanism preventing hybridization. To test for the proportion of fruits that develop without pollinator activity, 20 flowers each of all accession of the Impatiens purpureoviolacea complex as well as of the hybrid I. × troupinii were marked in a pollinator-free greenhouse. Only fruits containing at least one well-developed seed were counted as successfully developed for the fruit set.
Additionally, manual self-pollinations were conducted for ten flowers of the wild-collected I. × troupinii to test whether seeds resulting from self-pollinations are viable. The 26 seeds resulting from these pollinations were sown out to test whether they are able to germinate and form adult plants.
Based on their pollination syndrome, most species of the Impatiens purpureoviolacea complex included into this study are pollinated by butterflies and bees. However, a single species, Impatiens gesneroidea, is probably pollinated by birds, which may act as a prezygotic mechanism preventing the hybridization with insect-pollinated species. The categorization of pollination syndromes was taken from Abrahamczyk et al. (2017).

| Species distribution and habitat type
The species of the Impatiens purpureoviolacea complex are all endemics to the northwestern Albertine Rift with some narrow endemics occurring in small elevation zones with specific habitats . Several species show overlapping distribution ranges. The distributions of the species of the Impatiens purpureoviolacea complex as well as their habitat types were mapped by Eberhard Fischer, the taxonomic specialist of the group based on a current revision , own observations in the field and the current margins of the mountain forests. If two species occur not in the same place but close to each other, with a distance of ≤2000 m, a distance bees and butterflies can fly (Araújo et al., 2004;Townsend & Levey, 2005) we treated them as geographically close, having the theoretical possibility for pollen transfer and thus to hybridize.

| Chromosome counts
We aimed to count chromosome numbers of all species of the Impatiens purpureoviolacea complex to be able to reconstruct its chromosome evolution and assess its importance as a postzygotic mechanisms preventing hybridization. However, due to chromosome structure and size chromosome counts were impossible for I. elwiraurszulae and I. versicolor. The numbers of chromosomes were counted in metaphase plates, which were obtained from actively growing root tips from pot-cultivated plants. For chromosome preparation, we used a protocol according Pijnacker and Ferwerda (1984) and Belyayev et al. (2018)

| Flow cytometry
To find out whether the species of the Impatiens purpureoviolacea complex differ in genome size, we used flow cytometry (FCM) to measure it. Nuclear DNA 2C-values (monoploid genome sizes) were estimated using propidium iodide FCM. Each sample preparation followed the two-step procedure (Otto, 1990

| Phylogenetic analysis
We generated a phylogenetic tree to reconstruct the chromosome evolution of the Impatiens purpureoviolacea complex. Sequence data of chloroplast atpB-rbcL and nuclear ImpDEF1 and ImpDEF2 were obtained from earlier phylogenetic and evolutionary studies on Impatiens (e.g., Janssens et al., 2009;Fischer et al., 2021, Table   S1 Appendix S1). Alignment of the sequences was carried out using the software program MAFFT (Katoh et al., 2002) (Johnson & Soltis, 1998;Pirie, 2015). For this, Maximum Likelihood (ML) trees of each data matrix were created using the RAxML search algorithm (Stamatakis et al., 2005) under the GTRGAMMA + I approximation of rate heterogeneity for ImpDEF1 and GTRGAMMA for ImpDEF2 and atpB-rbcL.
Best-fit nucleotide substitution models for the plastid and

| Statistical analysis
We conducted a t-test on the genome sizes of the species of the Impatiens purpureoviolacea complex with 16 vs. 32 chromosomes.
The analysis was conducted in R v 3.4.3 (R Development Core Team, 2017).

| Ancestral chromosome reconstruction
To reconstruct ancestral haploid chromosome numbers and infer the type of chromosome number transitions, we used ChromEvol v.

| Reproduction
All species of the Impatiens purpureoviolacea group are largely dependent on pollinators for seed production. Fruit set induced by autonomous selfing only ranges from 0 to 15% (Table 1).
Impatiens × troupinii is able to form viable seeds by autogamous and manual selfing. Fruit set of the autogamous treatment is 15%, while with manual self-pollination it is 60%. The 26 seeds (mean 4.3 ± 2.9 per fruit) resulting from manual self-pollinations were sown out and developed to 15 adult but not very robust plants with pale pink flowers, of which a handful survived until flowering.

| Prezygotic mechanisms preventing hybridization
Several of the seven species from the Impatiens purpureoviolacea complex included into this study show overlapping distribution ranges (

| Mechanisms preventing hybridization after pollination
Genome sizes of 17 accessions (= genetic individuals) from the Impatiens purpureoviolacea complex, including seven out of ten species and one natural hybrid, were measured ( Chromosome numbers were counted for nine accessions, including five species and one natural hybrid (Table 3). Additionally, chromosome numbers of six closely related outgroup species were taken from literature ( Table 1). The chromosome numbers of the closely related outgroup species are diverse, ranging from 2n = 10 to 2n = 20 with a majority of species with 2n = 16 (Figures 3 and 4).

| D ISCUSS I ON
To date, this is the only study in the species-rich genus Impatiens as well as one of the first studies in flowering plants in which a multidisciplinary approach is applied where reproductive, cytological, geographical, and phylogenetic information is combined to investigate pre-and postzygotic mechanisms that may prevent hybridization within an entire clade. Bridging the gap between micro-and macro-evolution, we are able to document by which mechanisms the diversity of the small Impatiens purpureoviolacea clade may have evolved. However, our approach may be applicable to explain the amazing diversity not only in the genus Impatiens but in many other species-rich genera with co-occurring species as well. Ideally, future studies using a similar approach should have a balanced design analyzing the traits of a number  I. urundiensis occur geographically close in Burundi . These self-compatible but largely pollinator-dependent species are surprisingly variable in their habitats as well as in reproductive and cytological traits, which reflects the distribution of traits in the entire genus (Abrahamczyk et al., 2017;Jeelani et al., 2010;Lozada-Gobilard et al., 2019;Song et al., 2003). However, this trait diversity may have evolved as an adaptation to local conditions as well as a mechanism to prevent hybridization.
Chromosome numbers and genome sizes are highly correlated to each other in the Impatiens purpureoviolacea complex. The reconstructions for the crown node (and using the ultrametric tree also for the stem node) of the clade indicated 2n = 16 chromosomes, the most common number of chromosomes in Impatiens (Song et al., 2003). This is also true for all other nodes within the I. purpureoviolacea complex. Therefore, four independent polyploidization events occurred within the clade (I. urundiensis, I. purpureoviolacea, I. elwiraurszulae, and I. gesneroidea). Since we see little morphological variability in chromosome structure and do not have any evidence for a combination of hybridization and polyploidization, we assume that all species are auto-polyploids. All of these polyploidization events took place in sympatry with a diploid species from the Impatiens purpureoviolacea complex. In addition, evolutionary changes of a second ecological trait (habitat and/or pollination syndrome) may act as a further mechanism preventing hybridization as well (Table 4). Just Impatiens ludewigii and I. purpureoviolacea occur geographically close to each other and only differ in chromosome numbers; however, the contact zone of both species is small. Such combinations of cytological and ecological mechanisms that may prevent hybridization have been reported repeatedly in a range of more or less species-rich genera with young radiations, for example, in Achillea (Asteraceae), Silene (Caryophyllaceae), Chamaenerion (Onagraceae), or Houstonia (Rubiaceae; e.g., Glennon et al., 2012;Husband & Sabara, 2004;Karrenberg et al., 2019;Ramsey, 2011).
These mechanisms often separate populations with different ploidy levels within the same species as well as between closely related species.
The only known natural hybrid of the Impatiens purpureoviolacea complex, I. × troupinii represents rare crossing events of the two auto-polyploids I. purpureoviolacea and I. gesneroidea (both 2n = 32).
Impatiens × troupinii has been observed in the wild only a few times since the early 1980s  ssp. kilimanjari and the insect-pollinated I. pseudoviola (Grimshaw & Grey-Wilson, 1997). However, the hybrid of the second subspecies of Impatiens kilimanjari (I. kilimanjari ssp. pocsii) and I. pseudoviola-I. × kaskazini-is vigorously growing and relatively common, but occurs only at anthropogenically disturbed places in the forest (Grimshaw & Grey-Wilson, 1997).
In conclusion, the rare occurrence of I. × troupinii and I. × lateritia and the occurrence of I. × kaskazini only at anthropogenically disturbed places indicate that different pollination syndromes and habitat are strong but no absolute mechanisms potentially preventing hybridization in Impatiens. However, under natural conditions the combination of these mechanisms probably works well prohibiting hybridization in Impatiens. Similar combinations of mechanisms preventing hybridization may occur in other species-rich genera with co-occurring, closely related species as well.

ACK N OWLED G M ENTS
We

CO N FLI C T O F I NTE R E S T
The authors declare no competing interests.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data are presented in the main body of the article and in the Appendix S1.