Introduction
South Africa has been a major donor of invasive grasses to other parts of the world due to the perception that African C
4 grasses were more palatable to livestock and resistant to grazing than native grasses (Visser et al.
2016). Many African grasses have subsequently become invasive and highly damaging due to the combination of favourable traits, multiple introduction events (D’Antonio and Vitousek
1992; Visser et al.
2016), and high propagule pressure (Firn
2009). Historically, very few grasses have been targeted using biological control, mainly due to concerns of potential non-target attacks on economically valuable crops, tolerance to herbivory, and the belief that there were few or no specialist herbivores on grasses (Sutton et al.
2021a). These perceptions are, however, changing with the discovery of host-specific taxa (Goolsby et al.
2020), particularly eurytomid wasps in the
Tetramesa Walker (Hymenoptera: Eurytomidae) genus (Witt and McConnachie
2004; Sutton et al.
2021a,
b).
At the time of the review of grass biological control by Sutton et al. (
2021a), 23 invasive grass species were currently, or had already been, investigated as targets for biological control using herbivorous arthropods and fungal pathogens. The global weed biological control catalogue listed 14 releases of seven biological control agent species onto five invasive grass species between 1997 and 2019 (Winston et al.
2021; Supplementary Table S1). To date,
T. romana Walker is the only
Tetramesa species that has been used as a biological control agent for an alien invasive grass, namely
Arundo donax L. (giant reed or carrizo cane) in the USA (Goolsby and Moran
2009; Moran and Goolsby
2009).
Tetramesa romana is considered successful in the USA, particularly in the Rio Grande, Texas, where
A. donax biomass was estimated to be reduced by 32% in 2016, equating to 4.4 million $US a year in water savings (Goolsby et al.
2016; Moran et al.
2017). Following the success of this programme, there has been an increasing interest in the use of
Tetramesa as potential biological control agents for other alien invasive grasses (Lotfalizadeh et al.
2020; Sutton et al.
2021b).
The
Tetramesa genus currently comprises over 200 described species (Al-Barrak
2006; Lotfalizadeh et al.
2020; Natural History Museum
2021). They are highly host-specific, and typically specialise on only one species or genus, or on closely related genera, in the Poaceae (Phillips
1936; Claridge
1961). The larvae are endophagous feeders, and are either stem borers or gallers (Claridge
1961). Almost all of the sampling effort in collecting and describing
Tetramesa species has taken place in the Northern Hemisphere (Al-Barrak
2006) (Supplementary Fig. S1). Only four African species have been described, namely
T. aristidae Risbec from Senegal,
T. decaryi Risbec and
T. tananarivense Risbec from Madagascar, and
T. macalusoi De Stefani from Somalia (van Noort
2020). No
Tetramesa species have been described yet from South Africa (van Noort
2020), despite the existence of novel taxa in the region (Witt and McConnachie
2004; Sutton et al.
2021b).
In addition to
T. romana, a number of other
Tetramesa species are currently being considered as potential grass biological control agents. Witt and McConnachie (
2004) reported the presence of an unidentified
Tetramesa sp. on
Sporobolus africanus (Poir.) Robyns and Tournay,
S. natalensis (Steud.) T. Durand and Schinz, and
S. pyramidalis P. Beauv. in southern Africa. A second unidentified
Tetramesa species was found on
S. natalensis and
S. pyramidalis by Sutton et al. (
2021b) during native range surveys, where the wasp was found on only these two target weeds and not on any native congeners or other close relatives. Sutton et al. (
2021b) found that both
Tetramesa species significantly decreased tiller survival and had a deleterious impact on tiller reproduction. There are several other grasses that are indigenous to South Africa that are potential targets for biological control that may have other
Tetramesa species associated with them, particularly
Eragrostis curvula (Schrad.) Nees and
Andropogon gayanus Kunth (Olckers et al.
2021; Sutton et al.
2021b).
The Centre for Biological Control (CBC) at Rhodes University, South Africa, has been conducting field surveys for natural enemies on native grasses since 2017 across more than 200 sites and 70 grass species. The morphological uniformity of the adult and larval stages of the
Tetramesa makes it extremely challenging to identify different species (Dawah
1987; Ghajarieh et al.
2006). Even genus-level identifications delineating between the
Tetramesa and
Eurytoma have proven difficult (Henneicke et al.
1992), and reliable synapomorphies at even the family-level are still lacking (Lotfalizadeh et al.
2007; Gates
2008). The lack of a taxonomic backbone for
Tetramesa in South Africa and suitable morphological tools to distinguish between difference species has impeded our ability to assess the host specificity of field collected specimens and to establish pure laboratory cultures. The aims of the present work were therefore to use genetic techniques to distinguish between the
Tetramesa species, and from this make inferences about their predicted host ranges. Delineating species in this morphologically cryptic genus will advance efforts to identify potential biological control agents for the African grasses that have become invasive elsewhere.
Discussion
In general, very little taxonomic work has focused on Afro-tropical insects, particularly micro-Hymenoptera such as the
Tetramesa (van Noort et al.
2015; Berry and van Noort
2016; Hopkins et al.
2019). This investigation has provided new insights into the
Tetramesa assemblages on native African grasses by investigating their diversity, phylogenetic relationships, and host specificity. The work presented here has identified multiple
Tetramesa taxa that hold potential as biological control agents of invasive grasses, but some groups may be more suitably host-specific than others, and care must be taken to use an integrative taxonomic approach rather than relying solely on one line of evidence for identification.
Biological control outlook
The NPNS clades were found largely to be oligophagous feeders, and had a wider host range compared to the PNS groups. Biological control efforts should therefore prioritise the latter, and those that form part of the NPNS H. hirta and Andropogon gayanus clade, as well as PNS T. romana. This is based on the host-use patterns seen particularly in the COI phylogeny, where, with a few minor exceptions, each PNS clade was associated with a single host grass. The prospects for the biological control of the various target weeds are discussed below in the context of the phylogenies and inferences of host ranges. A promising species for M. maximus was found, but preliminary host rearing trials have indicated that this wasp can reproduce on Setaria sphacelata (Schumach.) Stapf and C.E. Hubb (unpublished data, Guy Sutton).
Eragrostis
The PNS E. curvula wasps (morphospecies ‘Tetramesa sp. 4’) belonged to the Tetramesa genus, but are likely not good candidates for biological control due to their broader host range on conspecific Eragrostis species. These could include non-target Australian and North American natives, that are either more closely related to the target weed than to the congeners from which they were collected, or very closely related to the congeners.
The nine PNS clades associated with Eragrostis hosts identified in the present COI phylogeny (spanning E. curvula, E. biflora, E. rigidior, E. trichophora, and E. plana) were delineated into eight GMYC taxa, and may represent unique species. Further taxonomic confirmation is, however, required to validate this observation. The low sequence divergence between the wasps on E. curvula clades PNS.ECUR.1 – PNS.ECUR.3 and E. trichophora PNS. ETRICH.1 (ranging from p-distances of 2.4 – 3.9%) suggests that host-specificity testing of these two wasps is important, as each taxon may be able to utilise both host plants, which could preclude their use as biological control agents. No-choice host-specificity testing should be completed, as it seems unlikely, although not impossible, that the same species of wasp would have such different feeding modes on two different host plant species, namely galling E. trichophora and mining the stems of E. curvula. Tetramesa sp. 26 on E. rigidior (PNS.ERIG.1) may not be an appropriate biological control agent, as it was also recorded on E. gummiflua in the field and therefore has a relatively broad host range. The second Tetramesa sp. 26 clade (PNS.ERIG.2), however, may be useful as it was specific to only E. rigidior. Further sampling focusing on these species is required.
Further work should focus on a population genetics analysis of both the PNS
E. curvula Tetramesa wasps and their hosts from a wider geographic range to test whether host plant form may be driving genetic divergence.
E. curvula has been developed as a pasture grass and has had numerous varieties developed. Such genetic breeding has led to distinct forms of the grass that vary morphologically and chemically (e.g.? phosphate and fibre concentrations) (Leigh
1961). In the case of different
E. curvula haplotypes in the native range, it will be vital to confirm matches to haplotypes in the invaded range to ascertain points of origin and select the most appropriate
Tetramesa population(s) as control agents. For example, Harms et al. (
2021) identified a novel haplotype of the alien invasive water weed
Hydrilla verticillata (L.f.) Royle, and found that the source population was from north-eastern China. This allowed the researchers to conduct targeted herbivore guild collections from this region.
The NPNS
E. curvula wasp that was assigned the morphospecies name ‘
Tetramesa sp. 5’ does not appear to belong to the
Tetramesa genus, based on the present nuclear 28S results. It is possible that the
Tetramesa as it currently stands should be divided into two genera: one comprising the predominantly PNS groups, and the other the NPNS clades. This would need to be achieved using an integrative approach, where genetic methods are used in conjunction with expert taxonomists. Both
Tetramesa sp. 4 and
Tetramesa sp. 5 are currently being used in host-specificity tests on native Australian
Eragrostis species, as this will be the ultimate deciding factor in their utility as biological control agents.
Tetramesa sp. 4 has already been found in Australia on field-collected
E. curuvla (Sutton et al.
2023), making it vital to assess its distribution and possible non-target effects.
Sporobolus pyramidalis
The PNS
Sporobolus wasps revealed two potential taxa, namely on
S. pyramidalis (PNS.SPYR.1) and
S. africanus (PNS.SAFR.1). The GMYC analysis suggested a split of the PNS.SPYR.1 clade into two taxa, but this was an artefact of geographic substructuring (these were from
Tetramesa collections in the KwaZulu-Natal and Eastern Cape provinces). Both field-based and no-choice tests have revealed that
Tetramesa sp. 1 (PNS.SPYR.1) is not able to utilise
S. africanus (Sutton et al.
2021b), but the host specificity of the
S. africanus wasps should be investigated further. The
Tetramesa in clade PNS.SAFR.1 is a novel group that requires host specificity testing mirroring the work conducted on
Tetramesa sp. 1 by Sutton et al. (
2021b).
Sutton et al. (
2021b) have already found that
Tetramesa sp. 1 is suitably host specific and damaging to
S. pyramidalis and
S. natalensis in the field. Additionally, this wasp has been recently imported into quarantine in Australia for further host specificity testing.
Both the PNS and NPNS
Sporobolus wasps (named morphospecies ‘
Tetramesa sp. 1’ and ‘
Tetramesa sp. 2’) have been shown to be host-specific in the field, where sp. 1 was found to be the more damaging of the two (Sutton et al.
2021b). The present phylogenetic results suggest that the NPNS
Tetramesa sp. 2 wasp does not belong to the
Tetramesa genus.
Tetramesa sp. 2 was identified morphologically as belonging to the
Tetramesa genus, but our genetic analyses suggest that they are sufficiently different to be considered a new or different genus. This highlights the need to update the taxonomy of the group.
The present genetic results suggest that the NPNS
Tetramesa sp. 2 wasp may be able to feed on
E. curvula (two wasp samples out of the nine that were collected on
S. pyramidalis), but field host range surveys by Sutton et al. (
2021b) have found that the wasp was specific to only
S. pyramidalis and
S. natalensis. Laboratory-based host specificity tests have also been conducted on over 20 non-target species, including eight native South African
Sporobolus species, with no non-target feeding recorded to date (Guy Sutton, unpublished data).
This scenario is an example of why it is imperative to have correctly identified each prospective agent in a biological control programme, and to have a solid understanding of their life histories and interspecific interactions in order to achieve the greatest level of damage. Since Tetramesa sp. 2 may be able to use E. curvula as a marginal host in the field, and that the wasp is less damaging than Tetramesa sp. 1, it might be prudent to only use Tetramesa sp. 1 for biological control.
Hyparrhenia hirta and Andropogon gayanus
The NPNS
Tetramesa collected on
Hyparrhenia hirta and
Andropogon gayanus that formed a sister group to the PNS clade were host-specific and formed a monophyletic clade in both gene trees. This is not surprising, since
H. hirta and
A. gayanus are phylogenetically closely related, both being in the tribe Andropogoneae (Skendzic et al.
2007), and structurally similar (i.e., both are tall-statured grasses) (Canavan et al.
2019b).
The COI GMYC species delimitation results suggested that this group may comprise three cryptic species, and could be considered as potential biological control agents because they are unique and host specific.
Hyparrhenia hirta has naturalised in the USA and many parts of Europe, but it is particularly problematic in Australia, where it has become highly invasive (Chejara et al.
2010). It has also been reported as a problematic weed in wheat fields in Pakistan (Hussain et al.
2004). Similarly,
A. gayanus has invaded the northern regions of Australia’s tropical savannas (Rossiter-Rachor et al.
2009). Invaded regions such as these could benefit from the
Tetramesa biological control agents identified here, particularly in mitigating intense bushfires and devalued land. Although there were only two specimens in this group, it would be worth conducting additional surveys on
A. gayanus in Zimbabwe, as there is likely to be a large degree of species diversity on the grass. Additional field surveys on other close relatives within the Andropogoneae are also required to assess their potential as biocontrol agents.
Southern African Tetramesa phylogenetics
The nuclear 28S rRNA marker was effective in broadly separating what are likely true
Tetramesa from other described or undescribed genera, where a suggested sequence divergence of ∼ 3 to 3.5% can be used for genus-level delimitation. Our interspecific p-distances were comparable to Chen et al. (
2004), who reported interspecific distances of 1.69–13.5% for their eurytomid data set. Of the four 28S sequences identified as
Tetramesa gleaned from GenBank, two fell within the proposed
Tetramesa clade (sourced from Chen et al. (
2004) and Munro et al. (
2011)) while the other two (Gillespie et al. (
2005), and a sequence from unpublished material (
https://www.ncbi.nlm.nih.gov/nuccore/DQ080114), fell in a NPNS polytomy that was more similar to some
Eurytoma sequences, and with a divergence of as much as 7.1% from the
T. romana group. This high nuclear divergence suggests that these NPNS GenBank specimens were either misidentified, or that the
Tetramesa genus needs to be split using morphological and molecular tools. Similarly, all the COI sequences deposited on BOLD that were identified as
Tetramesa (collected in Canada and Germany) showed sequence divergences of nearly 20% compared to
T. romana and
T. bambusae, which is a clear indication that these specimens likely belong to a different genus. The lack of genetic data across the entire genus is an impediment, and more sequence data is need to revise the genus.
In agreement with Chen et al. (
2004), the nuclear 28S rRNA marker is preferred as an initial identification guide, because it yielded well-supported basal nodes and broader-scale taxonomic relationships that will be useful for genus-level taxonomic revisions. The COI marker should be used as second filter to focus on specific target groups identified in the nuclear gene tree, as it provided greater resolution within the PNS group and revealed some potential cryptic species and/or genetically distinct populations. Numerous polytomies, however, were a confounding factor in inferring how these groups are evolutionarily related. Additionally, the COI phylogeny did not produce the same monophyletic
Tetramesa group as was seen using the 28S marker, and yielded surprisingly high sequence divergence values between the Northern Hemisphere
T. romana and
T. bambusae groups (10.8%). It was unexpected that the NPNS
H. hirta and
A. gayanus and the
T. romana clade in particular did not cluster with the PNS group in the COI phylogeny as it did in the 28S phylogeny, and that the COI genetic divergences were unusually high for these groups.
It is known that the COI marker could yield unexpected results when delimiting species that have undergone a rapid, recent radiation, as the lack of recombination in the mitochondrial genome can lead to an overestimation of sequence divergences due to the accumulation and retention of mutations (Hupalo et al.
2023). Comparatively, due to recombination, nuclear markers will thus more readily indicate renewed gene flow between previously isolated populations (e.g., due to climatic cycles and habitat changes), which can result in conflicting phylogenies (Eyer et al.
2017; Després
2019). Additionally, it is likely that these wasps can reproduce facultatively via thelytokous parthenogenesis (i.e., a form of asexual reproduction where diploid daughters are produced from unfertilised eggs) (Moran and Goolsby
2009). Compared to the default reproductive mode of arrhenotoky in the Hymenoptera (i.e., unfertilised eggs develop into haploid males and fertilized eggs develop into diploid females), thelytoky can lead to increased levels of homozygosity over time (Mateo Leach et al.
2009). It is even possible that infection by different
Wolbachia strains can cause reproductive isolation between intraspecific populations (e.g., in
Nasonia wasps (Bordenstein et al.
2001)). Future phylogenetic studies could investigate a wider range of nuclear markers—both ribosomal and protein-coding—and also the relationship between
Wolbachia and their
Tetramesa hosts in order to determine the degree to which these endosymbionts might be affecting genetic diversity, and under what conditions the wasps change their mode of reproduction.
This study has presented the first phylogenetic analysis of southern African
Tetramesa, and has revealed multiple taxa that are new to science. Combined with field host range data, we have identified at least five
Tetramesa taxa that show promise as biological control agents (these were namely clades PNS.SPYR.1, PNS.SAFR.1, NPNS.HHIR.2, NPNS.HHIR.3, and NPNS.AGAYA.1 in Fig.
2 and Supplementary Fig. S6). These wasps appear to be host-specific to
S. pyramidalis,
S. africanus,
H. hirta, and
A. gayanus. Based on field host range data,
Tetramesa sp. 4 on
E. curvula showed evidence of oligophagy across congenerics, and is unlikely to be suitably host-specific for release in Australia where there is a very large diversity of
Eragrostis species. Recommendations for future research include: (1) incorporating multiple other nuclear regions and ddRADseq techniques into phylogenetic analyses to refine divergence thresholds for delimiting true
Tetramesa from other eurytomid genera, as well as to resolve the high incidence of polytomies observed, (2) using phylogenetic methods in conjunction with morphological and ecological data to revise the taxonomy of the group (with the possibility of creating subgenera), and (3) conducting further host-specificity tests, impact assessments, and potential hybridisation trials with the different
Tetramesa taxa to determine their efficacy as biological control agents.
Acknowledgements
The African lovegrass biocontrol project is supported by funding from the Australian Government Department of Agriculture, Fisheries and Forestry as part of its Rural Research and Development for Profit program and the AgriFutures Australia Programme. This project has been assisted by the New South Wales (NSW) Government through its Environmental Trust. The NSW Environmental Trust and the NSW Weed Biocontrol Taskforce are two of the cash contributors for African Love grass research under the national project `Underpinning agricultural productivity and biosecurity by weed biological control’, supported by the Australian Government programme Rural Research and Development for Profit (RRnD4P) (Round 4). Additional funding was provided by the Department of Forestry, Fisheries and the Environment: Natural Resource Management programme (DFFE: NRM) and the South African National Research Foundation (NRF) Research Chairs Initiative of the Department of Science and Technology. Any opinions, findings, conclusions and recommendations are those of the authors and the funders do not accept liability in this regard. We also wish to thank Lenin D. Chari for collecting specimens, the Albany museum for the use of their microscope and photographic software, Shelley Edwards for the use of the Zoology and Entomology Molecular Laboratory (ZEML) facilities, and Abigail Kirkaldy for assistance in photographing specimens.