This research was carried out following the evaluation and approval of the sampling protocols by the South African Institute for Aquatic Biodiversity (SAIAB) ethics committee (Ref: 2014/03). Permits to carry out this research were obtained from the Parks and Wildlife Authority of Zimbabwe.

The EZH freshwater ecoregion receives a mean annual precipitation of 900 to 3000 mm/a (Mazvimavi 2010) which sustain perennial flow in the rivers and streams of this region. There are four major river systems that drain this ecoregion: the Lower Zambezi, Pungwe, Save and Buzi systems (Fig. 1). Two tributaries of the Lower Zambezi (the Nyangombe and Kairezi rivers) drain the western and northern parts of this ecoregion. Three localities were sampled in the Nyangombe subcatchment and two localities were sampled in the Kairezi subcatchment (Fig. 1; Suppl. material 1). The Pungwe River flows eastwards from the EZH, and the river has a length of about 400 km from its source to its discharge point into the Indian Ocean near Beira in Mozambique. A total of 17 localities covering four major subcatchments of the Pungwe River system (i.e., the Rwera, Pungwe mainstream, Honde and Nyamukwarara) were sampled (Fig. 1; Suppl. material 1). Riparian zones of the Rwera and mainstream Pungwe subcatchments are covered with remnants of the Afromontane rainforests, whereas miombo woodlands are the dominant vegetation types in the riparian zones of the Honde and Nyamukwarara subcatchments. The Buzi River system has three eastward draining subcatchments (the Buzi, Rusitu and Revue). Two localities were sampled in the Rusitu subcatchment (Haruni and Upper Rusitu) which drains the Chimanimani Mountains. The Odzi River, a major subcatchment of the Save River system, drains southwards from the EZH (Fig. 1). Two localities were sampled in the Odzi subcatchment (the Nyamazi and Nyanyadzi rivers). A single locality in the Upper Save was also sampled to collect specimens and tissue samples from the type system of Zaireichthys monomotapa (Suppl. material 1).

Fishes were collected in December 2013 and 2014 using a Samus-725M electrofisher. Captured fishes were anaesthetised with clove oil, digitally photographed and a small piece of muscle tissue was dissected from the right side of each specimen and preserved in 95% ethanol in the field for genetic analysis. Tissue samples were stored at -80^oC at the South African Institute for Aquatic Biodiversity (SAIAB), Grahamstown. Voucher specimens were fixed in 10% formalin in the field. They were then transferred through 10% and 50% to 70% ethanol for long-term storage. All voucher specimens were deposited into the fish collection facility at SAIAB as reference material. Species were identified using regional identification keys and their known geographic distributions according to Skelton (2001) and Marshall (2011).

DNA was extracted from preserved tissue using the salting out method (Sunnucks et al. 1996). DNA concentration was quantified using a Nanodrop ND-1000 (Nanodrop Technologies, Inc). A fragment of the mitochondrial cytochrome oxidase subunit I (COI) gene was amplified by polymerase chain reaction (PCR) using universal fish DNA barcoding primer sets: VF2-T1 and VR1-T1 (Ivanova et al. 2007), FishF1 and FishR1 or FishF2 and FishR2 (Ward et al. 2005). PCRs were performed with a Veriti 96 well thermal cycler (Applied Biosystems) and each reaction mixture (25 µL) contained 100–200 ng template DNA, 14.4 µL of water, 2.5 µL deoxynucleotide triphosphate (dNTP) (10 mM), 2.5 mM MgCl₂, 2.5 µL PCR buffer (10X), 0.5 µL of each primer (20 pmol) and 0.1 µL Taq DNA polymerase (Southern Cross Biotechnology, Cape Town). The PCR amplification profile was 94 °C for 3 min, followed by 38 cycles of 94 °C for 30 s, 55 °C for 30 s and 72 °C for 50 s, and then final extension at 72 °C for 7 min. PCR products were purified with Exosap (Applied Biosystems), cycle-sequenced using BigDye Cycle Sequencing Kit (Applied Biosystems, Foster City, CA, USA) and sequenced at SAIAB using an ABI 3730xl DNA Analyzer (Applied Biosystems). Sequences generated from this study were submitted to GenBank (accession numbers: Amphilius (MH431952 - MH432017); Chiloglanis (MH432018 - MH432062); Hippopotamyrus (MH432063 - MH432086); Zaireichthys (MH432087 - MH432119); Suppl. material 1). BOLD sequences that were included in this study are also listed in Suppl. material 1.

Sequences were cleaned, aligned and trimmed to equal lengths using the program SeqMan7.2.1 (DNASTAR, Madison, WI, USA). The appropriate models of sequence evolution for each genus were selected using the Akaike’s Information Criteria (AIC) (Burnham and Anderson 2002) as implemented in jModeltest 2 (Darriba et al. 2012). Phylogenetic relationships among unique haplotypes within each genus were inferred using MrBayes 3.1.2 (Ronquist et al. 2012). The analysis for each genus had two replicate searches of 10 million generations with four Markov chains. Trees were sampled every 1000 generations to obtain 10000 sampled trees. TRACER 1.5 (Rambaut and Drummond 2007) was used to assess if the chains had converged and determine the burn-in. We discarded 10% of the sampled trees as burn-in and the remainder were used to calculate the consensus tree and Bayesian posterior probabilities. Model corrected genetic distances between unique lineages identified for each genus were calculated using PAUP (Swofford 2003). To explore the possible taxonomic distinctiveness of the genetic lineages that will be uncovered from the Eastern Highlands, sequences of topotypes (i.e., samples collected from or within the vicinity of the type localities of currently described species), whenever available, were also included in the analyses.

The General Mixed Yule Coalescent (GMYC) method (Pons et al. 2006) was used to delineate candidate species or operational taxonomic units (OTUs). The GMYC is a robust method that models branching thresholds for intraspecific (coalescent) and interspecific (speciation/diversification) patterns (Fujisawa and Barraclough 2013). This approach has been widely applied in a number of studies to identify cryptic diversity within morphologically defined species (e.g., Crivellaro et al. 2017; Grabowski et al. 2017; Kordbacheh et al. 2017). The ultrametric trees for GMYC analyses were estimated in BEAST (Drummond and Rambaut 2007) using two priors (Yule model and Coalescent model with constant population size) and two rates of molecular evolution (constant and relaxed clock). For each of the four genera considered in the present study, we compared three trees which were built based on the following combinations of priors and rates of molecular evolution: (i) yule model and a constant clock, (ii) yule model and a relaxed clock, and (iii) coalescent model with constant population size and a constant clock. The GMYC analyses were conducted with the package ‘splits’ (Species Limits by Threshold Statistics (http://r-forge.r-project.org/projects/splits) using R v.3.4.1 (R Development Core Team 2011).

The COI dataset for Chiloglanis consisted of 73 sequences (included 30 BOLD sequences; Suppl. material 1) and an edited alignment of 534 bp with 175 polymorphic sites which resulted in 48 unique haplotypes. Bayesian analysis divided haplotypes of C. neumanni from the EZH ecoregion into two well-supported clades (Fig. 2a). Estimates of the number of candidate species using GMYC ranged from eight to 16, excluding outgroups (Fig. 2a). The coalescent model with constant population size and a constant clock (CONC) gave the most conservative estimate of the number of candidate species, and these were largely consistent with the clades inferred using the Bayesian analysis, with the exception of one sample in the C. sp. ‘rough skin’ clade and two samples in the C. sp. ‘dwarf’ clade which were assigned to separate clusters from these two major clades (Fig. 2a). Both the Yule model and a constant clock (YULE), and the Yule model and a relaxed clock (RELA) clearly overestimated the number of candidate species due to over splitting of samples in the C. sp. ‘rough skin’ and C. sp. ‘dwarf’ clades (Fig. 2a). A conservative approach was therefore considered in the present study, and six candidate species or molecular operation taxonomic units (MOTUs) were delimited within C. neumanni from the EZH ecoregion and the adjacent Shire and Ruo rivers in Malawi (Fig. 2a). The first major clade in Fig. 2a contained four candidate species, C. sp. ‘rough skin’, C. sp. ‘Zambezi’, C. sp. ‘Pungwe’, and C. sp. ‘Shire’, with model corrected genetic divergences between these candidate species ranging from 1.35–7.60% (Table 1). The second major clade contained two candidate species, C. sp. ‘dwarf’ and C. sp. ‘Nyangombe’ which were separated by 4.01–5.27% sequence divergence.

Figure 2a. 

Bayesian phylogenetic tree based on mtDNA cytochrome oxidase sub unit I (COI) sequences showing the candidate species or molecular operational taxonomic units (MOTUs) identified within Chiloglanis neumanni from the Eastern Zimbabwe Highlands freshwater ecoregion. Well supported nodes are shown by a solid circle. The indicated candidate species or MOTUs were identified using the GMYC method based on trees that were built using three different combinations of priors and rates of molecular evolution: (i) coalescent model with constant population size and a constant clock (CONC), (ii) yule model and a constant clock (YULE) and (iii) yule model and a relaxed clock (RELA).

Chiloglanis sp. ‘rough skin’ comprised haplotypes from the Pungwe and Buzi river systems (Fig. 2b). Chiloglanis sp. ‘Zambezi’ contained haplotypes from the Nyangombe River in the EZH ecoregion, as well from a tributary that flows into Lake Cahora Bassa and from the Cubango River which is part of the Okavango River system (Fig. 2b). Chiloglanis sp. ‘Pungwe’ contained haplotypes that were confined to the Pungwe River system (Fig. 2c). Chiloglanis sp. ‘Shire’ was confined to the Shire and Ruo rivers (Fig. 2b). Chiloglanis sp. ‘dwarf’ was distributed in the Pungwe and Ruo rivers, while C. sp. ‘Nyangombe’ was only recorded from the Nyangombe subcatchment during the present study (Fig. 2c). The genetic divergences between almost all the candidate species identified from the present study are consistent with interspecific genetic divergences between morphologically distinguishable Chiloglanis species, for example C. anoterus and C. bifurcus from the Inkomati River system which are separated by 2.85–3.59% COI sequence divergence (Table 1). It is also important to note that all the candidate species identified from the EZH freshwater ecoregion are deeply divergent from C. pretoriae (16.59–20.89% sequence divergence; Table 1; Fig. 2a), a name that was previously assigned to the EZH ecoregion suckermouth catlets. Unfortunately, sequences of both C. neumanni sensu stricto and C. emarginatus, two other species names that were previously used for the EZH freshwater ecoregion suckermouth catlets, were not available for comparison in the present study.

Figure 2b. 

The distribution of Chiloglanis sp. ‘rough skin’ (orange circle), Chiloglanis sp. ‘Zambezi’ (blue circle) and Chloglanis sp. ‘Shire’ (purple circle) in the Eastern Zimbabwe Highlands freshwater ecoregion and adjacent areas.

Figure 2c. 

The distribution of Chiloglanis sp. ‘dwarf’ (green circle), Chiloglanis sp. ‘Nyangombe’ (blue circle) and Chloglanis sp. ‘Pungwe’ (red circle) in the Eastern Zimbabwe Highlands freshwater ecoregion and adjacent areas.

The COI dataset for Amphilius consisted of 79 sequences (included 16 BOLD sequences; Suppl. material 1) and an edited alignment of 572 bp with 166 polymorphic sites which resulted in 34 unique haplotypes. Bayesian analysis divided samples of A. natalensis and A. uranoscopus into two separate and well-supported clades (Fig. 3a). Estimates of the number of candidate species within A. natalensis using GMYC ranged from one to six (Fig. 3a). Yule model and a relaxed clock (RELA) clearly underestimated the number of candidate species in A. natalensis sensu lato due to the geographic isolation and disjunct distribution of the populations within this ‘species’. Five candidate species were therefore delineated based on trees generated using the Yule model and a constant clock (YULE), which gave the next most conservative estimate on the number of species (Fig. 3a). Two of these candidate species, A. sp. ‘natalensis Pungwe’ and A. sp. ‘natalensis Buzi’ are confined to the Pungwe and Buzi river systems, respectively (Fig. 3b). These two candidate species are allopatrically distributed and are separated by substantial genetic divergence (Fig. 3a; 6.86–8.58% genetic divergence; Table 2). Divergences within candidate species ranged from 0.00–0.72% in A. sp. ‘Pungwe’ and from 0.00–1.30% in A. sp. ‘Buzi’. Amphilius sp. ‘Pungwe’ was collected from multiple localities in the Pungwe River system and a single locality in the Kairezi River, a tributary of the Lower Zambezi, while A. sp. ‘Buzi’ was collected from the Haruni, Rusitu, and the mainstream Buzi River (Fig. 3b). The present study revealed that A. sp. ‘Pungwe’ and A. sp. ‘Buzi’ are genetically distinct (10.44–12.71% sequence divergence, Table 2; Fig. 3a) from A. natalensis sensu stricto based on a specimen collected from the Umgeni River (the type system for A. natalensis) in South Africa.

Figure 3a. 

Bayesian phylogenetic tree based on mtDNA cytochrome oxidase sub unit I (COI) sequences showing the candidate species or molecular operational taxonomic units (MOTUs) identified within Amphilius uranoscopus and A. natalensis from the Eastern Highlands of Zimbabwe. Well supported nodes are shown by a solid circle. The indicated candidate species or OTUs were identified using the GMYC method based on trees that were built using three different combinations of priors and rates of molecular evolution: (i) yule model and a constant clock (YULE), (ii) coalescent model with constant population size and a constant clock (CONC) and (iii) yule model and a relaxed clock (RELA).

Figure 3b. 

The distribution of Amphilius sp. ‘natalensis Buzi’ (dark blue circle), Amphilius sp. ‘natalensis Pungwe’ (red circle) and Amphilius sp. ‘natalensis Ruo’ (light blue circle), Amphilius sp. ‘uranoscopus Save’ (purple triangle), Amphilius sp. ‘uranoscopus Buzi’ (black triangle), Amphilius sp. ‘uranoscopus Pungwe’ (green triangle), Amphilius sp. ‘uranoscopus Zambezi’ (orange triangle) and Amphilius sp. ‘uranoscopus Ruo’ (yellow triangle) in the Eastern Zimbabwe Highlands freshwater ecoregion and adjacent areas.

Bayesian analysis revealed a well-supported shallow clade corresponding to the species currently recognised as A. uranoscopus in southern Africa (Fig. 3a). Estimates of the number of candidate species within A. uranoscopus using GMYC gave consistent results for the EZH ecoregion (four candidate species) based on both the Yule model and a constant clock (YULE) and the coalescent model with constant population size and a constant clock (CONC) models (Fig. 3a). All the candidate species within A. uranoscopus from the EZH ecoregion are allopatrically distributed. Haplotypes within Amphilius sp. ‘uranoscopus Save’ were collected from the Nyamazi River, a tributary of the Odzi River, and from the mainstream Nyanyadzi River (Fig. 3b). Both A. sp. ‘uranoscopus Buzi’ and A. sp. ‘uranoscopus Pungwe’ were confined to the Buzi and Pungwe river systems, respectively (Fig. 3b). Haplotypes within A. sp. ‘uranoscopus LZ’ occurred in the Nyangombe and Kairezi rivers, which are both tributaries of the Lower Zambezi (Fig. 3b). The candidate species within A. uranoscopus sensu lato had shallow divergences among them (0.54–2.56% sequence divergence; Table 3) compared to the deep divergences found among lineages within A. natalensis sensu lato (Table 2). Notable exceptions were samples of A. uranoscopus collected from the Ruo River (BOLD sequence MAFW032) which is the type river for A. hargeri and from the Cubango River (BOLD sequence ANGFW075) which is the type river for A. cubangoensis which were respectively deeply and moderately divergent from all the other lineages within the A. uranoscopus sensu lato clade (Table 3). The analysis also showed that the Ruo and Cubango samples of A. uranoscopus were genetically distinct, being separated by 7.23% (Table 3).

The edited alignment of 47 Zaireichthys sequences (included 15 BOLD sequences, Suppl. material 1) was 534 base pairs in length with 151 polymorphic sites which resulted in 21 unique haplotypes. GMYC analyses delimited Z. monomotapa sensu stricto which is distributed in the Save and Pungwe river systems, and identified one candidate species from the EZH ecoregion, Z. sp. ‘slender’, which occurred in the mainstream Nyangombe River and its tributary, the Chidiya River (Fig. 4a, 4b). BOLD sequences (SAFW899, SAFW900, SAFW905, SAFW907, SAFW908) for individuals collected from three streams which flow into Lake Cahora Bassa, as well as sequences MAFW12911 and MAFW13611 for individuals collected from two tributaries which flow into Lake Malawi also belonged to this clade (Fig. 4b; Suppl. material 1), indicating that this taxon may be widespread in the Lower Zambezi and Lake Malawi catchments (Fig. 4b). Zaireichthys monomotapa sensu stricto and Z. sp. ‘slender’ were separated by 9.26–11.00% sequence divergence, while divergences within these taxa ranged from 0.0–0.96%. Zaireichthys sp. ‘leopard spot’, a candidate species from the Ruo River, was basal to Z. monomotapa sensu stricto (Fig. 4a), with a divergence of 3.22–3.80% between these two sister taxa. A second candidate species, Z. sp. ‘Chilwa’, within the Z. monomotapa complex occurred in the Montepuez River system in northern Mozambique and the Lake Chilwa system in Malawi is also shown in Fig. 4a.

Figure 4a. 

Bayesian phylogenetic tree based on mtDNA cytochrome oxidase sub unit I (COI) sequences showing the candidate species or molecular operational taxonomic units (MOTUs) identified within Zaireichthys monomotapa from the Eastern Zimbabwe Highlands freshwater ecoregion. Well supported nodes are shown by a solid circle. The indicated candidate species or MOTUs were identified using the GMYC method based on trees that were built using three different combinations of priors and rates of molecular evolution: (i) yule model and a constant clock, (ii) yule model and a relaxed clock, and (iii) coalescent model with constant population size and a constant clock.

Figure 4b. 

The distribution of Zaireichthys monomotapa sensu stricto (green circle), Zaireichthys sp. ‘slender’ (red circle) and Zaireichthys sp. ‘leopard spot’ (orange circle) in the Eastern Zimbabwe Highlands freshwater ecoregion and adjacent areas.

The edited alignment of 27 Hippopotamyrus sequences (included 11 BOLD sequences, Suppl. material 1), including a single Cyphomyrus cubangoensis sequence, was 580 base pairs and had 88 polymorphic sites which defined 17 haplotypes. Phylogenetic inference assigned the Pungwe and Buzi populations of H. ansorgii to two distinct clades that were separated by 2.85–3.27% divergence (Fig. 5a; Table 4). Divergences within lineages were 0.0–0.35% for H. sp. ‘Pungwe’ and 0.00–1.25% for H. sp. ‘Buzi’. GMYC analysis identified three clusters (or candidate species): (i) samples from the Kwanza, Upper Zambezi and the Buzi River system, (ii) samples from the Kunene River system and the Ruo River, and (iii) samples from the Pungwe River system represented a candidate species, H. sp. ‘Pungwe’ (Fig. 5a, 5b). The genetic divergences between lineages within these clusters are presented in Table 4.

Figure 5a. 

Bayesian phylogenetic tree based on mtDNA cytochrome oxidase sub unit I (COI) sequences showing the candidate species or molecular operational taxonomic units (MOTUs) identified within Hippopotamyrus ansorgii from the Eastern Zimbabwe Highlands freshwater ecoregion. Well supported nodes are shown by a solid circle. The indicated candidate species or MOTUs were identified using the GMYC method based on trees that were built using three different combinations of priors and rates of molecular evolution: (i) yule model and a constant clock, (ii) yule model and a relaxed clock, and (iii) coalescent model with constant population size and a constant clock.

Figure 5b. 

The distribution of Hippopotamyrus sp. ‘Buzi’, (red circle), Hippopotamyrus sp. ‘Pungwe’ (blue circle) and Hippopotamyrus sp. ‘Ruo’ (green circle) in the Eastern Zimbabwe Highlands freshwater ecoregion and adjacent areas.

This study represents the first fine scale geographical survey and genetic exploration to determine the extent of hidden diversity in stream fishes of the Eastern Zimbabwe Highlands (EZH) freshwater ecoregion. Species delimitation assessment using the GMYC method revealed existence of 16 molecular operational taxonomic units (MOTUs) or putative species in five currently recognised nominal species (i.e., six MOTUs in Chiloglanis neumanni, two MOTUs in Amphilius natalensis, four MOTUs in A. uranoscopus, two MOTUs in Z. monomotapa and two MOTUs in H. ansorgii) collected from the EZH freshwater ecoregion. Given that all these five ‘species’ (with the exception of Z. monomotapa sensu stricto) were described from systems outside the EZH freshwater ecoregion, 15 of the 16 identified MOTUs within these ‘species’ are likely to represent new species that were previously unrecognised by science. Although the GMYC is a robust method for species delimitation (Fujisawa and Barraclough 2013), the entities identified in the present study were not designated as ‘distinct species’, but rather considered to represent ‘candidate or potential species’ pending critical evaluation using integrative approaches that combine molecular, morphological and ecological data to determine their taxonomic integrity. Ongoing assessments by the authors have revealed consistent morphological differences between Chiloglanis sp. ‘rough skin’ and C. sp. ‘dwarf’, and formal descriptions of these newly identified species are underway.

While the present study’s main focus was on the EZH freshwater ecoregion, it is important to indicate that the diversity of stream fishes in the broader Eastern Afromontane Region may be much higher than currently documented as highlighted by the presence of other candidate species outside the EZH freshwater ecoregion. These include Amphilius sp. ‘Ruo’, Zaireichthys sp. ‘Chilwa’ and Hippopotamyrus sp. ‘Ruo’, as well as lineages within the A. uranoscopus complex. Additional fine-scale geographic surveys are therefore required to fill missing sampling gaps to more accurately map the distribution ranges of these lineages and potentially identify additional hidden diversity. Results of the present study also showed that samples collected from the type localities of the three synonyms of A. uranoscopus from southern Africa, A. hargeri (BOLD sequence MAFW032), A. brevidorsalis (sequence MB9404BUZ) and A. cubangoensis (BOLD sequence ANGFW075) were genetically differentiated from other populations of A. uranoscopus sensu lato (see Fig. 4a). This warrants further taxonomic investigations to determine whether these synonyms represent valid species that may need to be resurrected.

The existence of such high taxonomic diversity within a small portion of the Eastern Afromontane Region is consistent with findings from a number of previous studies that have uncovered substantial hidden diversity and narrow range endemic species (or lineages) within several stream fishes that were previously thought to have wide geographic ranges in southern Africa (e.g., Chakona et al. 2013; Goodier et al. 2011; Maake et al. 2014; Morris et al. 2015; Swartz et al. 2007, 2009; Wishart et al. 2006). Results of the present study thus add to the growing board of evidence that shows that a large proportion of freshwater fishes in southern Africa remain scientifically undocumented, because many river systems remain poorly explored as much of the previous research effort and application of molecular approaches has predominantly focussed on fishes of the Cape Fold Ecoregion (Ellender et al. 2017). Despite being one of the geographically and taxonomically well explored regions in southern Africa, new species and deeply divergent genetic lineages continue to be discovered within almost all fish taxonomic groups of the Cape Fold Ecoregion (Chakona and Swartz, 2013; Chakona and Skelton, 2017; Chakona et al. 2013, 2014; Wishart et al. 2006), and estimates indicate that there are about 43 undescribed species within the 21 currently recognised fish species of this region (Linder et al. 2010). The discovery of hidden diversity in the EZH freshwater ecoregion adds to the growing evidence for the existence of high species-level diversity within a number of fish species from high altitude streams in southern, east and west Africa (e.g., Friel and Vigliotta 2011; Morris et al. 2016; Schmidt and Pezold 2011; Schmidt et al. 2014, 2015, 2016; Thomson 2013; Thomson and Page 2010). Given that many regions in southern Africa, in particular the subtropical and tropical river systems, have not been adequately explored, and the use of modern approaches for rapid species discovery remains very limited, additional diversity is likely to remain hidden within other wide-ranging ‘species’ in the region.

Findings from the present study have considerable implications for aquatic biodiversity conservation in the EZH freshwater ecoregion, and the broader Eastern Afromontane region. Existence of such high cryptic diversity within five ‘species’ from a few mountain tributaries which represent a very small portion of the Eastern Afromontane region indicates that the overall conservation value of this globally important biodiversity hotspot has been severely underestimated. This is because the current biodiversity estimates and level of endemism in this region does not include cryptic diversity within stream fishes. Because many of the stream fishes from the EZH freshwater ecoregion are thought to have wide geographic ranges, they are considered to be of least conservation concern (Tweddle et al. 2009). This has resulted in stream fishes being neglected from ongoing conservation efforts in the EZH freshwater ecoregion, and the broader Eastern Afromontane region. This can be seen in the calls for research funding, where the primary focus is on other vertebrate groups such as birds, amphibians, reptiles and small mammals (e.g., see http://www.cepf.net/grants/project_database/Pages/default.aspx#). The present study uncovers a classic example where underestimation of taxonomic diversity and poor understanding of the spatial distribution of species has misdirected conservation efforts, to the extent that the EZH freshwater ecoregion is currently not listed among the priority freshwater Key Biodiversity Areas within the Eastern Afromontane Biodiversity Hotspot (see CEPF 2012). This is however unfortunate because the highly sensitive Afromontane streams and rivers in this region have been severely transformed and are experiencing ongoing human impacts, including illegal mining activities, deforestation, increased sedimentation, uncontrolled burning and introduction of non-native invasive piscivorous species (Kadye and Magadza 2008; Kadye et al. 2013). For example, the rocky streams which drain the Chimanimani Mountains used to have intact indigenous riparian vegetation, had clear water and perennial flow, but surveys in 2013 revealed that human encroachment, increased agricultural activities and the associated loss of riparian vegetation has transformed these streams into sluggish flowing, and highly turbid and heavily silted habitats. This increases the risk of losing sensitive fish species and other aquatic taxa from the EZH freshwater ecoregion whose value as an endemic hotspot has been previously underestimated.