Comparative phylogeographic analysis of Cuscuta campestris and Cuscuta reflexa in Kenya: Implications for management of highly invasive vines

Eastern Africa is currently infested with dodders — highly invasive vines that threaten agriculture and biodiversity. We describe historical and ecological factors that have potentially shaped dodder invasion in this region. Our results reveal high levels of genetic diversity within dodder populations, which could enhance their ability to colonize new niches. Further, we identify two main population clusters in eastern and western Kenya, a pattern that is mirrored in an established invader ( Cuscuta campestris ) and a recent introduction ( C. reflexa ). Immediate control strategies are needed to limit deliberate or inadvertent seed/vine dispersal, eradicate existing populations, and monitor influx of propagules. impacts. We recommend (i) development of targeted efforts to curb new introductions to limit the potential for genetic variability and adaptation, (ii) prevention of seed and vine dispersal from current incursions, and (iii) complete eradication of existing populations where this is feasible.


Societal Impact Statement
Eastern Africa is currently infested with dodders-highly invasive vines that threaten agriculture and biodiversity. We describe historical and ecological factors that have potentially shaped dodder invasion in this region. Our results reveal high levels of genetic diversity within dodder populations, which could enhance their ability to colonize new niches. Further, we identify two main population clusters in eastern and western Kenya, a pattern that is mirrored in an established invader (Cuscuta campestris) and a recent introduction (C. reflexa). Immediate control strategies are needed to limit deliberate or inadvertent seed/vine dispersal, eradicate existing populations, and monitor influx of propagules.

Summary
• Invasion by dodder (Cuscuta spp.), holo-parasitic plants of the Convolvulaceae family, has recently surged in Eastern Africa. Particularly, colonization by two species, Cuscuta campestris and Cuscuta reflexa that have a broad host range and rapid invasiveness, has escalated. We examined the first assessment of genetic diversity of these invasive species in Kenya to inform subsequent management practices.
• A comparative phylogeographic analysis, based on chloroplast and nuclear ribosomal DNA, was performed to determine genetic diversities, population structure, and isolation of C. campestris and C. reflexa in Kenya.
• Results reveal high genetic variation in both species within populations across localities. C. campestris had more haplotypes and higher gene diversity (P = .0001) and heterozygosity (P = .004) than C. reflexa. Both species were geographically separated, evidenced by two distinct lineages that mostly corresponded to localities east and west of Kenya. Correlation between genetic and geographic distances revealed evidence of isolation by distance (IBD).
• We attribute these genetic variation patterns to establishment via similar invasion routes for the studied Cuscuta spp. and suggest that Cuscuta invasion in Kenya is in the early stages of establishment. Rapid eradication is urgently needed to prevent shifts to economically important hosts, which would have devastating 1 | INTRODUCTION Phylogeography examines distribution of natural populations and elucidates ecological and evolutionary processes that have shaped patterns of population structure (Avise et al., 1987). The ongoing surge in colonization by novel invasive alien species (IAS) or expansion by naturalized ones across habitats has increased the demand for insights into biological mechanisms that have influenced these invasions (Meyerson & Mooney, 2007;Schulz et al., 2019). The ability of an IAS to survive in a locality can depend on various factors, key among them being their introduction and adaptability to new environments, evolutionary history, and anthropogenic activities that either enhance or restrict dispersal (Dlugosch & Parker, 2008;Ward et al., 2008).
Understanding genetic variability can provide useful insights into patterns of local adaptation to changing environmental conditions and explain divergence among populations (Flanagan et al., 2018). In plants, pollen and seeds are key avenues for gene flow (Hamilton & Miller, 2002;Levin, 1981). A key determinant for the success of this process is dispersal of these materials, either biologically by vectors or abiotically by environmental agents, such as water and wind. Previous studies have shown that most plants adopt a leptokurtic gene flow pattern, where pollen and seeds are dispersed near a genotype's locality, and these are likely to be the most adapted (Bradshaw, 1972).
Conversely, plants that lack long-distance dispersal mechanisms are expected to have highly restricted gene flow; hence, they are expected to exhibit a high genetic structure (Loveless & Hamrick, 1984;Sork & Smouse, 2006). Successful expansion of most invasive weeds into new ranges depends on their life history traits, particularly high reproductive outputs and wide dispersal abilities that ensure colonization and establishment over heterogenous environments (Sork et al., 2016). This reproductive plasticity also ensures continuous establishment of new populations amid the changing environmental conditions. Since plants are sessile, their patterns of distribution and/or divergence across different geographical settings are subject to influence by isolation by distance (IBD), isolation by environment (IBE), and isolation by adaptation (IBA) (Wang & Bradburd, 2014). For instance, highly heterogenous environments between populations may affect flowering times or pollinators, whereas long distances may restrict movement of pollen or seeds.
Overall, these modes of isolation influence gene flow and affect phylogeography of plants.
All Cuscuta species are considered obligate holoparasites because they lack an established root system and therefore entirely depend on host plants for support and nourishment (Kaiser et al., 2015). They attach on a suitable host via haustorium, creating a conduit for siphoning of water, nutrients, and photosynthates (Dörr, 1969;Kuijt, 1969).
Additionally, taxonomic and molecular identification as well as character evolution and distribution of these species have been studied (García et al., 2014;Ren et al., 2020;Stefanovi c et al., 2007;Yuncker, 1932). To date, little is known regarding genetic variation and demographic patterns of these noxious weeds across habitats, with only a handful of reports describing population genetics (Elsiddig et al., 2018;Mutikainen & Koskela, 2002). Evaluating the genetic basis underlying successful invasion could offer valuable insights to inform development of control and management measures, considering that genetic variability is an important determinant of invasiveness (Kolbe et al., 2004;Tsutsui et al., 2000;Ueno et al., 2015). Specifically, IAS with a high genetic diversity are expected to have superior adaptability that enables them to cope with perturbing and stressful conditions in the new environments (Prentis et al., 2008).
The overarching goal of this study was to explore patterns of genetic diversity in two distinct Cuscuta species occurring in Kenya and elucidate historical as well as evolutionary characteristics that have shaped their distribution and expansion across the landscapes.
Field dodder (Cuscuta campestris Yuncker.), a North American species that is also native to the Caribbean islands, is naturalized in Africa, having been first identified in South Africa in 1918 (Yuncker, 1932) while the giant dodder (Cuscuta reflexa Roxb.), a tropical Asian species, was first reported in Mauritius in the 1970s (Holm et al., 1979).
Recently, we provided the first evidence of presence of this Indian species in continental Africa (Masanga et al., 2021). Both species share ecomorphological traits (propagating by vines and seed) that allow them to colonize new localities across different environments, and in Kenya, they are co-distributed to the east and west of the country. Moreover, they are generalist parasites that infect numerous plants across various angiosperm orders (Masanga et al., 2021). To date, the role of other characteristics, such as physiological similarities, degree of virulence, habitat preference, reproductive plasticity, and human-aided dispersal remains unknown. We hypothesized that populations of the two Cuscuta species could be genetically diverse, with this diversity potentially determining their success in colonizing current and subsequent expansion into new localities. Our findings suggest the existence of similar invasion routes for both species and emphasize the need to manage invasions prior to establishment.

| Sampling
A total of 10 and nine populations of C. campestris and C. reflexa, respectively, were collected from localities within their known distributions across Kenya. Identification and characterization of these species were carried out as earlier described in our previous work (Masanga et al., 2021). We focused on these species because (i) they are highly invasive and are currently expanding to various ecosystems across East Africa, and (ii) they have a wide host range, with potential to expand this parasitism to economically important tree crops, thereby posing danger to the agricultural sector. From each locality, we collected at least 5 individuals, at least 30 m apart, to minimize chances of resampling. Vines (5-10 cm), together with any flowers present, were collected and placed in silica gel, then stored at 4 C until DNA extraction. Sampled localities, across Cuscuta spp. ranges in Kenya are shown in Table 1 2.2 | DNA extraction, PCR, and sequencing DNA was extracted from the sampled materials, using a standard DNA isolation protocol described by Doyle and Doyle (1990), and then used for PCR targeting the chloroplast (ribulose bisphosphate carboxylase large subunit [rbcL]) and nuclear (internal transcribed spacer [ITS]) regions. For rbcL, we amplified a partial fragment using primers rbcL-512F and rbcL-1392R (McNeal et al., 2007), whereas in ITS, we targeted the 5.8S of the nuclear nrDNA region, alongside those flanking 18S and LSU (26S) genes (Baldwin, 1992). Amplifications were performed in 25-μl reactions using MyTaq™ DNA polymerase kit (Bioline, Meridian Biosciences), according to the manufacturer's instructions. Reaction mixes were subjected to an initial denaturation step of 95 C for 1 min, followed by 35 cycles of 95 C for 15 s, 30-s annealing (at appropriate temperatures for each primer) and a 1-min extension at 72 C. A final extension at 72 C for 10 min was also included. PCR products were purified using the Qiaquick™ PCR purification kit (Qiagen, USA), then sequenced, using the forward primer, on the ABI system at Macrogen (Macrogen Inc.). Sequences were cleaned in SeqMan Pro17 (DNASTAR Inc., Madison, WI, USA), to remove low quality reads and adapters and then aligned using ClustalX version 2.0 (Larkin et al., 2007). The resulting sequences were submitted to GenBank for the following accession numbers; C. campestris rbcL-MW581109-MW581141; MT849993-MT850001, and MT947606-MT9476223; C. reflexa rbcL-MW581080-MW581108; ITS-MT850025-MT850044 and MW080817-MW080836.

| Data analysis
The number of haplotypes (h), haplotype diversity (hd), and nucleotide diversity (π) were calculated using DnaSP 5.0 (Librado & Rozas, 2009). For phylogenetic analysis of haplotypes, we performed Bayesian inferencing (BI) using full sequences of the haplotypes. First, we used the program jModelTest version 2.1.6 (Posada, 2008) to determine the best fit model for both datasets and selected the general time-reversible (GTR) model with gamma distribution and assuming invariable sites (GTR + G + I) (Yang, 1994)  To further assess the relationship among haplotypes, we constructed median-joining (MJ) networks using popART v1.7 (Leigh & Bryant, 2015), with each indel treated as a single mutation event.
Next, we investigated demographic history of both Cuscuta species using neutrality tests and mismatch analyses, by calculating Tajima's D (Tajima, 1989) and Fu's F (Fu, 1997)  cies indicated that C. campestris had higher average gene diversity and heterozygosity across all populations than C. reflexa. Notably, this difference was statistically significant (P < .05) based on rbcL but not with regards to ITS (Table S1).

| Population and phylogeographic structure
We analyzed genetic differentiation among Cuscuta populations by comparing G ST and N ST , measures of differentiation that consider allelic frequencies and similarities among haplotypes, respectively.
Generally, a significantly higher N ST than G ST value is indicative of phylogenetic subdivision, whereas the opposite is true for a significantly higher G ST than N ST (Pons & Petit, 1996). Estimates, based on cpDNA,  (Figure 1). Similarly, median joining networks corroborated these results, as evidenced by two clusters that also corresponded to the two geographical regions. Interestingly, taxa from Nandi populations (NANC and NANR) clustered with those from eastern localities despite being geographically in the west. SAMOVAbased genetic analyses, targeting rbcL and ITS haplotypes, partitioned our data into two genetically distinct groups for both C. campestris and C. reflexa.
To assess variation across populations, we estimated and compared variance as well as proportions of variation among groups using

| Demographic analyses and geographical distribution of haplotypes
Since SAMOVA partitioned our taxa into two distinct groups, we analyzed potential population expansion based on these groups, using mismatch analysis and neutrality tests. Generally, negative Tajima (Table 3), neither of which also had a negative value of Fu's F (Table 3). Based on these data, we found limited evidence of population expansion in both species across sampled localities. Spatial distribution of the haplotypes in Kenya is shown in Figure 2. In summary, geographically proximal localities shared haplotypes from the same clade, with an overlap observed among shared ones. Notably, no haplotypes were shared between populations to the east and west of the country (Figure 2). Interestingly, Nandi, which is on the west of the Rift Valley depression and is geographically close to the western group, was genetically more like eastern populations.

| Isolation by distance
Kernel density estimates indicated a high concentration of points in Nandi and Kisumu for C. campestris and C. reflexa, respectively ( Figure 3). Consequently, we used these as reference points for calculating Euclidean distances between populations (Tables S2 and   S3). A correlation between geographical and genetic distances (Tables S4-S7) appeared to suggest isolation by distance, although only C. campestris (based on ITS) was statistically significant.

| DISCUSSION
We report genetic patterns that may have shaped successful invasion by C. campestris and C. reflexa, genetically distinct species, across Kenya. These two, alongside other Cuscuta species, have invaded various ecosystems across Eastern Africa (Masanga et al., 2021). We discuss these results with regards to how modes of reproduction and ability to disperse to new environments may have shaped the current invasion. These findings are expected to inform future management strategies for control of these noxious weeds.
The phylogeographic pattern of an invasive species is influenced by various factors, such as its life history characteristics, dispersal capabilities, environmental requirements, and association with ecology (Ditchfield, 2000). To understand the underlying invasion patterns, one requires to determine how much genetic divergence exists among individuals, haplotypes, and populations, as well as geographical distribution of the haplotypes (Ditchfield, 2000). Previous studies have recommended the use of multiple genes across genomes in a species, for detection of genetic structure among individuals, as these circumvent limitations associated with interpretation of historical processes brought about by a single gene (Bermingham & Moritz, 1998;Schaal et al., 1998;Templeton, 2002). In the present study, we used two genetic markers to analyze genetic variation and population structure in Cuscuta spp. Specifically, we chose rbcL due to its low-level genetic diversity, which offers high differentiation among populations and a slow rate of evolution (Gaut et al., 1992), as well as nrDNA (ITS) because it offers more variability (relative to coding regions) and has a low level of intraspecific divergence (Zhang & Hewitt, 2003 Note: Fixation indices represent average pairwise comparisons across all populations, n = 10 and n = 9 for C. campestris and C. reflexa, respectively. df, degrees of freedom; PV, percent variance; F ST , proportion of variation among populations; F CT , variance among groups. * P < .01.

T A B L E 3 Neutrality tests and mismatch distribution parameters in Cuscuta species
Pop Gene Tajima's D (P) Fu's F (P) SSD (P) Hrag (P) (except H4 in C. reflexa) were shared within, but not between, localities in the east and west of the country. We also observed that populations farthest from each other were genetically distinct, albeit at no statistical significance (except C. campestris based on ITS), suggesting a small isolation-by-distance effect. The only exception to this was observed in taxa from Nandi, which despite being geographically closer to western localities, clustered and shared haplotypes with eastern populations. Taken together, these results suggest existence of more than one center of genetic diversity, on either side of the country. We attribute this allopatric pattern of distribution to multiple introductions of propagules, and a complex interaction between the invasive potential of Cuscuta and the environment. The discovery of multiple centers of genetic diversity is particularly concerning given evidence that parasitic plants are expected to have faster rates of molecular evolution across all three plant genomes, relative to their nonparasitic counterparts (Bromham et al., 2013).
Although dodder seeds lack specialized adaptations for dispersion, long-distance spread through bird endozoochory (Costea et al., 2016(Costea et al., , 2019 and water (Ho & Costea, 2018) have been documented. In addition, we speculate that human-aided activities could be playing a key role in propagule exchange within populations. These activities could be direct, through movement of seeds and vines to new localities, as has previously been described (Knepper et al., 1990) (Barrett et al., 2008) and dispersal events (Arendt, 2015). These parameters shape the patterns of dispersal of propagules into a new environment, subsequent adaptation to the prevailing conditions and expansion. In many species, invaders that predominantly propagate vegetatively are expected to have a lower genetic diversity compared with their seed-propagated counterparts (Barrett et al., 2008). This asexual reproduction is advantageous when population sizes are small (Barrett et al., 2008;Lambrinos, 2001), although genetic bottlenecks may hinder successful expansion, especially in competitively more demanding environments. In addition, vegetative propagations limit a species' ability for genetic admixture, through reduced gene flow and low recombination (Eckert, 2002). As expected, all Cuscuta populations followed an isolation by distance pattern, with the only exception observed in Nandi. Despite being geographically closer to western localities, this population shared haplotypes and clustered with those from east of the country, possibly because it could be an introduction from these localities. Overall, the lack of haplotype sharing and the IBD pattern suggests constrained long-distance dispersal across western and eastern groups.
Interestingly, we noted some striking similarities in genetic variation between the two Cuscuta species under this study. First, both species exhibited an allopatric pattern of distribution, as evidenced by two clades that corresponded to localities in the west and east of Kenya. Second, both species exhibited significantly higher genetic differentiation (N ST > G ST ) across all populations, and third, the highest variation in both species was attributed to groups within, rather than among populations. Taken together, these findings suggest that both species may be employing near-similar historical characteristics during colonization and expansion into the observed localities. Conversely, t test results indicated that C. campestris populations had higher average gene diversities and heterozygosity than C. reflexa, across the sampled localities, possibly due to differences in demographic histories between the species. Particularly, C. campestris has been around longer than C. reflexa, thus has had more time for mixing and sharing of haplotypes among localities after introduction. We also speculate that C. campestris, a naturalized invader, could be predominantly propagating through seed, hence enhanced mixing among populations, whereas C. reflexa may be largely spreading via vines although this remains to be confirmed. C. reflexa has larger seeds than C. campestris, although previous researchers found no evidence of any association between seed size and geographical distribution range among members of the genus Cuscuta (Olszewski et al., 2020). Future investigations are expected to elucidate the relationship between reproductive plasticity and the observed genetic variability in Cuscuta spp.

| IMPLICATIONS FOR MANAGEMENT
Our findings have far-reaching implications for strategic management of alien invasive species, such as Cuscuta spp. Management of alien species invasions falls into four stages, namely, prevention, eradication, containment, and asset-based protection, with substantially improved cost to benefit ratios for management during early stages of prevention and eradication, relative to later invasion stages (Harris et al., 2018). Evidence from our previous research revealed little to economically important crops such as tea and coffee (Masanga et al., 2021). In other places, such as neighboring Uganda, Cuscuta spp., have been found parasitizing coffee (Coffea robusta) (Kagezi et al., 2021). We suggest that current Cuscuta spp. invasions in Kenya are at the early stages of establishment; therefore, rapid eradication of existing populations is urgently needed before spread and adaptation to economically important hosts. Emphasis should be placed on limiting new introductions across populations, before genetic mixing, as this may enhance genetic variability and increase adaptability.
Because seed plays an important role in gene flow and establishment of new populations, whereas vines ensure continuity of the existing ones, we propose that efforts should be put in place to prevent deliberate or inadvertent dispersal of these propagules, while eradicating existing populations.
F I G U R E 3 Isolation by distance showing the relationship between genetic and geographical distances among populations. The upper panel represents Cuscuta campestris datasets as follows; (a) density estimation of sampled points. (b) and (c) are correlations between genetic and geographical distances based on rbcL (R 2 = .2065, P = .1871) and ITS (R 2 = .5564, P = .01130), respectively. The lower panel represents C. reflexa datasets as follows; (d) density estimation of C. reflexa points. (e) and (f) are correlations based on rbcL (R 2 = .3191, P = .1130) and ITS (R 2 = .4262, P = .0567), respectively. Points marked in red (in graphs b, c, e, and f) indicate populations west of Kenya, whereas those in blue denote populations to the east. Density was estimated using a two-dimensional kernel density estimation in OriginPro v9.8. X and Y axes on the density plots represent latitude and longitude, respectively