Introduction

Koalas (Phascolarctos cinereus) are an iconic Australian marsupial species listed as vulnerable on the IUCN ‘red list’ of threatened species (Woinarski and Burbidge 2016). While a large part of their ongoing population decline is due to habitat loss, disease threats are additionally limiting population viability (Polkinghorne et al. 2013). Following European settlement, koala populations across Australia declined significantly due to hunting in the 1890s to 1920s, with southern populations nearing extinction. During this time, small refuge populations were established on offshore Victorian islands and these koalas have been subsequently used to restock most of their former southern range. These southern populations (Victoria and South Australia (SA)) are genetically distinct from the northern animals (Queensland (QLD) and New South Wales) (Neaves et al. 2016; Ruiz-Rodriguez et al. 2016). Southern koalas are in over-abundance in some parts of their range and they have also recently been shown to have lower rates of the two major disease threats affecting koala population viability, chlamydiosis and koala retrovirus (KoRV) (Simmons et al. 2012; Polkinghorne et al. 2013; Legione et al. 2016; Fabijan et al. 2019a, 2019b; Sarker et al. 2019).

Koalas in the southern areas have restricted genetic diversity when compared with their northern counterparts (Houlden et al. 1996, 1999; Taylor et al. 1997; Fowler et al. 1998; Lau et al. 2014; Ruiz-Rodriguez et al. 2016; Kjeldsen et al. 2018) likely as an effect of the founder, bottleneck and inbreeding effects in these populations. Low diversity places populations at risk of increased expression of recessive deleterious traits. A number of pathologies with a suspected genetic basis have been reported in the southern populations at a higher rate than in the northern animals. Testicular abnormalities have been recorded in animals derived from the Kangaroo Island population in SA including testicular aplasia (Seymour et al. 2001; Cristescu et al. 2009), asymmetry and atrophy (Speight et al. 2018) with rates as high as 12.9% of affected males reported in the Eyre Peninsula population in SA. In contrast testicular aplasia was found to be absent in the outbred Pilliga ranges in New South Wales, with the rates of testicular abnormalities closely linked to the inbreeding co-efficients of the populations (Seymour et al. 2001).

Another suspected genetic disorder is the renal disease oxalate nephrosis, which has a very high prevalence in the Mount Lofty Ranges population in SA, with up to 32% of deaths attributed to this cause (Speight et al. 2018). Koalas with this disease have kidney failure associated with deposition of calcium oxalate crystals in the tubules of the kidneys. It is also apparent that even clinically normal animals in the Mount Lofty and Kangaroo Island populations have elevated levels of oxalate in their urine when compared with an unaffected northern population from southeast QLD (Speight et al. 2013, 2014). Genetic causes of renal oxalate disease are reported in other species (Danpure and Rumsby 2004; Heiene et al. 2009; Vidgren et al. 2012) but these are yet to be identified in southern koalas.

Microcephaly, colloquially known as “pin-head” has also been reported in southern animals, particularly by wildlife care groups working with certain isolated populations in Victoria (Australian Senate 2011), though with no formal scientific studies of the problem it is hard to gauge the prevalence of this condition. Other skeletal disorders reported across the range of koalas that may have a genetic component include craniofacial tumours (primarily osteochondromas) (Sutton 1986; Canfield et al. 1990; Hanger and Loader 2014), microphthalmia and scoliosis (Higgins et al. 2018).

Approaches to analysing the genetic diversity of bottlenecked wildlife populations have generally focussed on examining a limited range of markers. Probably the most popular technique has been the use of variability in short tandem repeats (STR) also known as microsatellite markers via PCR and visualisation or sequencing of variations in the number and size of these repeats across individuals and populations. PCR amplification and sequencing of variable gene regions such as whole or partial mitochondrial genomes or genes or major histocompatibility (MHC) loci has also been extensively deployed. The advent of the availability of reference genomes has also led to the use of SNP analysis, typically performed using a microarray from a related species for which extensive genetic information on SNP variability is available (Hindrikson et al. 2017; O'Brien et al. 2017; Gienapp et al. 2017; Cammen et al. 2016). With newer sequencing technologies becoming more affordable partial genome sequencing technologies such as RADseq (restriction site associated DNA sequencing) (Andrews et al. 2016; Cammen et al. 2016) have also been deployed. The lack of good quality reference sequences and the costs of whole genome sequencing have meant that more comprehensive methods such as whole genome resequencing and GWAS as deployed in humans and some high value domestic species has remained limited (Jones and Good 2016). The numbers of animals required for meaningful analysis of population variance in association with phenotypic traits has remained prohibitive for wildlife species. However studies in North American puma have demonstrated that for species with significant genetic bottlenecks and large tracts of heterozygosity in their genomes, whole genome sequencing of individuals can provide startlingly in depth information on inbred populations (Saremi et al. 2019). The use of RNAseq (exome capture) in human and livestock species studies of genetic disorders is a widely used compromise in terms of ability to provide in-depth SNP sequencing for larger numbers of individuals for a reasonable cost (in comparison to whole or partial genome sequencing) (Jones and Good 2016). Its use in wildlife species has however remained limited, presumably because of the increased difficulty in preservation and sample quality from field sample collection sites as well as the lack of reference genomes to facilitate mapping of sequencing reads (Cammen et al. 2016).

This paper reports a novel use of RNAseq data to identify the likely underlying genetic variation responsible for a number of suspect genetic disorders in a bottlenecked wildlife population (southern koalas). The paper reports the identification of high quality variants in transcriptomic data and differential gene expression of genes in lymph node tissue from a total of 29 koalas from two populations, one from south east QLD (10 animals), representative of the “northern” genotype and one from the Mount Lofty Ranges in SA (19 animals) representative of the “southern” genotype. This paper provides the first evidence for an underlying genetic basis for several disorders affecting southern koala populations and sets the stage for more detailed studies to inform breeding, translocation and restocking decisions in this iconic vulnerable species.

Methods

Ethics

Ethical approval for this study was granted by the University of Queensland Animal Ethics Committee, permit number ANFRA/SVS/461/12 and ANRFA/SVS/445/15, the Queensland Government Department of Environment and Heritage Protection permit number WISP11989112, the University of Adelaide Animal Ethics Committee permit number S-2013-198 and the South Australian Government Department of Environment, Water and Natural Resources Scientific Research Permit Y26054.

Samples

Samples were collected from wild-rescued koalas that had been euthanised for clinical reasons and submitted for post-mortem examinations from South East QLD (Greater Brisbane) (n = 10) and SA (Mount Lofty Ranges) (n = 19) Fig. 1. The mainland Mount Lofty Ranges koala population in SA primarily originated from the Kangaroo Island population (derived in turn from Victorian island populations) but also included koalas from QLD and New South Wales (Lindsay 1950; Bergin 1978). Age was determined by dentition and the amount of wear on the upper premolar (Martin and Handasyde 1999). Submandibular lymph nodes were collected within 2–6 h of death into RNALater® and stored at − 80 °C.

Fig. 1
figure 1

Map of koala sampling locations, Mt Lofty Ranges marked with a blue star, Greater Brisbane with a green star. Approximate current range of wild koalas marked with a blue line (current distribution derived from Atlas of living Australia http://www.ala.org.au accessed 22/12/20) estimated demarcation between southern and northern genotypes marked with a green line

Full size image

Details of the koalas, sample quality and read quantity are provided in Online resource 1. A summary of koala details is provided in Table 1. Details of the comparative pathology, chlamydial and Koala retrovirus (KoRV) status of the animals included in this study are reported elsewhere (Tarlinton et al. 2017; Sarker et al. 2019; Fabijan et al. 2020).

Table 1 Details of the koalas used in this study
Full size table

RNA extraction and sequencing

Total RNA was extracted using an RNeasy Mini kit with on column DNAase1 digestion (Qiagen). RNA quantity and quality were assessed via anXpose spectrophotometer (Bioke) and Agilent 2100 Bioanalyzer. mRNA was prepared for sequencing using the Illumina TruSeq stranded mRNA library prep kit and 100 base pair, paired end sequencing was performed on an Illumina HiSeq. Reads were trimmed with skewer (Jiang et al. 2014) using 3′ quality cut off of 3 (− q = 3) and mean quality of 20 (− Q = 20) prior to mapping to the available koala genome contig assembly GCF_002099425.1_phaCin_unsw_v4.1 from NCBI genomes for SNP detection or gene expression analysis.

Genome mapping and SNP calling

Trimmed reads were mapped to the koala contigs using Hisat2 (Kim et al. 2019). Sorted bam files were processed using opossum (Oikkonen and Lise 2017) to allow robust identification of SNPs from transcriptome data. SNPs were called from combined bam files using samtools (Li et al. 2009)mpileup and output genotype likelihoods generated in bcf format (-g -b -u options). Bcftools (BCFtools)was used to call multiallelic SNPs (-vm options). Identified SNPs were filtered using vcftools (Danecek et al. 2011). SNPs where a base was missing in more than half the samples, with minor allele frequency less than 0.2, reads of minimum quality < 30 were removed (--max-missing 0.5 --maf 0.2 --minQ 30 --recode). For each SNP a population FST was calculated using Weir and Cockerham’s algorithm (Wier and Cockerham 1984) (--weir-fst-pop option). Locations and predicted effects of SNPs on the function of known genes were determined using SNPEff (Cingolani et al. 2012) with maximum up/down stream distance set as 2000 bp.

Gene expression analysis

To quantitate the expression of known genes, sequence alignment and read quantification was performed using the pseudo-alignment-based tool Kallisto v0.43.0 (Bray et al. 2016). Transcripts with a mean TPM < 1 for both groups (QLD and SA) were removed. Differential expression of transcripts between QLD and SA animals was determined using Sleuth v0.28.1 (Pimentel et al. 2017). Transcripts with a false detection rate corrected p value < 0.05 and a mean log2 fold change (log2FC) greater than 1 and less than minus 1 were considered to be differentially expressed. False detection rate correction was performed using the Benjamini–Hochberg method (Benjamini and Hochberg 1995). To account for possible confounding factors analysis was repeated with a model comparing location alone against one containing location + age (as approximated by tooth wear class) + sex. This is by a likelihood ratio test in sleuth of a “full model” (~ location + age + sex) vs a”reduced model” (~ location).

Protein structural changes

Protein structural changes due to SNPs in the gene SLC26A6 were predicted using the Phyre2 software suite (Kelley et al. 2015).

Gene set enrichment analysis was conducted using a hypergeometric test in ToppGene. (Chen et al. 2009) Benjamini–Hochberg corrected p values < 0.05 are reported.

Results

Identification of SNP variants between populations

A total of 215,834 unique variants were identified. To remove SNPs with high coverage variability, those where a base was identified in   ≤ 66% of both QLD and SA populations were removed. This resulted in a final set of 135,191 SNPs. The mean fixation index (FST) of these SNPs was 0.412; when filtered for those SNPs with strongest evidence of divergence between the two populations (FST ≥ 0.8), 24,981 variants remained (23,145 SNPs and 1836 INDELs, see Online resource 2). A high FST is evidence of a history of separation of genes or populations but could be biased by variants on the sex chromosomes. Therefore, it was important to test if any of the high FST variants were biased by sex of the animals. As the genome of the koala is currently not assembled into chromosomes, but rather as contigs, separation of variants that are located on the sex chromosomes was not immediately possible. As a proxy to identify any variants associated with sex of samples, a Wilcoxon rank sum test was used to test the frequency of each different genotype in male vs female animals. Only a single variant (contig NW_018344261.1 position 1080527) showed a significant association with sex (p value = 0.048). This however was not significant after correcting for multiple hypothesis testing. Therefore, we propose that the SNPs reported are not biased by the ratio of sexes of the animals in the two populations.

Annotation of the variants using SNPEff highlighted the majority of variants are located in introns (34,488 variants). As the mapped reads are from RNAseq, possible sources of high coverage in introns are a mix of mis-annotation, immature mRNA sequencing or possible genomic DNA contamination. To keep analysis here as robust as possible, these predicted intronic variants were not used in later analysis. The majority of other annotations were in non-coding regions (3′ UTR = 10,256, downstream = 7658, upstream = 2504, 5′UTR = 1700). Of the SNPs in coding regions, the majority were synonymous variants (5589) and a smaller proportion were non-synonymous variants (3320).

Functional annotation of population variants

SNPEff characterises the predicted impact of SNP on gene function. This annotation ranges from unknown “modifier” to low, moderate and high impact. Removing those genes with limited impact (“modifier”) resulted in 3669 unique variant positions affecting 2168 unique genes (Online resource 2). Gene set enrichment showed significant enrichment in poor brain function in humans (DisGeNET (Piñero et al. 2020) codes C3714756, C0917816, C0423903, C4020876, C1843367, C0025362 respectively) and Human Phenotypes relating to brain abnormalities (Human Phenotype Ontologies (Köhler et al. 2019) codes 0002060, 0100547, 0011282, 0011283 respectively). A total of 349 genes were associated with these terms which contained a total of 2341 variants (Online resource 3 and 4).

Differential gene expression between populations

A comparison of gene expression between the QLD and SA koalas identified 104 statistically significant differentially expressed transcripts in the lymph node tissue (Online resource 5). Forty of these were downregulated in SA animals (compared with QLD animals) and 64 of these were upregulated in SA animals (compared with QLD animals). As samples were recovered from wild rescue euthanised animals and couldn’t be perfectly matched between locations there may be confounding factors which contribute to gene expression changes observed. Factors that could be more confidently controlled for were sex and age. However, age is an approximation based on tooth wear class (see online resource 1) and so is used with caution. Using the same tools but with a model for differential expression analysis controlling for age and sex, 51 statistically significant differentially expressed transcripts were detected (q value < 0.05, 26 upregulated in SA, 25 downregulated). These are highlighted in online resource 5 in column “qval (~ location + age + sex)”.

There were a variety of transcripts downregulated in the SA animals that have a plausible link to testicular abnormalities, including several with testes specific expression or a role in sex steroid metabolism (at least in the mouse and human models) including Meiotic Kinetochore factor (MEIKIN), serine protease inhibitor Kazal-type 10-like (important in sperm capacitance), lamin tail domain containing 1 (LMNTD1), multiple EGF like domains 11 (MEGF11), growth arrest specific 8 (GAS8), aldo–keto reductase family 1 member C1 homolog and critically, the androgen receptor (AR). There were also a number of transcripts that are usually testis- expressed that were upregulated in the SA animals including ST6 N-acetylgalactosaminide alpha-2,6-sialyltransferase 2 (ST6GALNAC2), family with sequence similarity to 204 member A (FAM204A), anaphase promoting complex subunit 16 (ANAPC16), testis expressed 11 (TEX11), zinc finger protein 345-like (testes expressed transcription factor), nicotinamide nucleotide adenylyltransferase 3 (NMNAT3) and PHD finger protein 24-like. However, none of the SA koalas in the study cohort exhibited testicular abnormalities.

In terms of anion-exchange transport proteins known to be linked to oxalate metabolism disturbances (and therefore to oxalate nephrosis), solute carrier family 26 member 6 (SLC26A6) was upregulated in the lymph node of SA animals, of which 21% (4/19) were affected by clinical oxalate nephrosis. The functions of this protein (in humans) include apical membrane anion-exchange, a role in the pH buffering system and mediation of intestinal chloride absorption and oxalate secretion.

Similar to the genes enriched in the SNP analysis there were a large number of differentially expressed genes between the southern and northern koala groups that are linked to known brain expression and facial deformity syndromes in people. Including calcium dependent secretion activator (CADPS), protocadherin gamma-C4, family with sequence similarity 204 member A (FAM204A), vesicle amine transport 1 like (VAT1L), alkaline ceramidase 3 (ACER3), nestin (NES), ArfGAP with coiled-coil, ankyrin repeat and PH domains 3 (ACAP3), dishevelled binding antagonist of beta catenin 1 (DACT1), neuronal PAS domain protein 1 (NPAS1), prosaposin-like, acyl-CoA oxidase 2 (ACOX2). None of the SA koalas in the study cohort exhibited brain or facial deformities, but two koalas showed scoliosis and kyphosis.

There were a small number of genes differentially expressed that could plausibly be linked to the overwhelming KoRV expression in the QLD koalas and lower prevalence in SA koalas. These genes include a C–C motif chemokine 4-like transcript and integrin alpha-M-like, prostaglandin E synthase (PTGES) which were upregulated in the SA animals. C–C motif chemokine 4 binds directly to CCR5 (one of the receptors for HIV in humans) and inhibits CCR5-mediated entry of HIV-1 and HIV replication. The solute carrier family 29 member 1 (Augustine blood group) (SLC29A1), which was downregulated in SA animals, is thought to be a DC-sign-like pathogen-associated molecular pattern (PAMP) that is potentially involved in viral recognition. Similarly, down regulation of several MHC1 linked transcripts (HLA class I histocompatibility antigen, alpha chain G-like, class I histocompatibility antigen, Gogo-OKO alpha chain-like), which could be involved in KoRV recognition, were seen in the SA animals.

Several of the differentially expressed genes appear to be transcript variants of the same locus, in some cases different transcripts of the same gene are upregulated or downregulated in the SA animals, these include Subtilisin-like protease one (2 variants, one upregulated, one downregulated), SUN domain-containing protein 2-like (4 variants, 3 upregulated, 1 down regulated), bridging integrator 2 (BIN2) (1 up regulated, 1 down regulated), pyroglutamyl-peptidase I (PGPEP1) (1 upregulated, 1 down regulated) and syncytin-1- like mRNA (2 up regulated and 1 down regulated).

SNPs associated with differentially expressed genes

A total of 241 variants were detected within 52 transcripts from 47 unique differentially expressed genes (see online resource 6). A permutation test to determine the joint probability any transcript is differentially expressed and also contains a SNP showed that this is unlikely to have occurred by chance alone (p < 1 × 10−16, calculated from probability of 10,000 sample permutations). Of these the majority (116) are located in annotated introns or downstream of genes (38 downstream and 35 3′ UTR). A smaller proportion of variants were upstream (7 variants) and 5′ UTR located (4 variants). A total of 40 variants were located in coding regions (21 synonymous, 14 missense and other splice variants). A Kruskal–Wallis rank sum test was used to test for significant differences of within gene expression between observed genotypes at each variant positions (that is, if different SNP genotypes are associated with different levels of cis gene expression). A total of 80 variant positions showed significantly different gene expression between genotypes (Benjamini–Hochberg corrected p value < 0.05). These are within 8 genes: CUTA, LETMD1, CD209 antigen-like protein E (LOC110202118), class I histocompatibility antigen (LOC110203128), SLC26A6, SLC29A1, ST6GALNAC2 and NPAS1. As with all variants studied, the majority (48) were in introns with 24 in the coding sequences (17 synonymous and 7 missense) and the majority appear to be fixed within the SA animals compared with the QLD animals. Three genes had SNPs in the 3′ UTR; LETMD1 (3 SNPs); LOC110202118 and ST6GALNAC2 (each with one SNP each) which may be related to gene expression changes through mRNA degradation or miRNA binding.

The functional relationship of some of these genes to the disorders in the southern populations is not clear (CUTA, LETMD1, SLC29A1, ST6GALNAC2) and the population variation seen in these SNPs may be the result of genetic divergence of the two geographically separated populations, through genetic drift or due to the genetic bottlenecks that the SA population has undergone. Two of these genes (LOC110202118: CD209 antigen-like protein E, LOC110203128: class I histocompatibility antigen) were linked to immune function and may be related to either expression in the face of KoRV infection or differing resistance to KoRV infection (as has been reported for some rare MHC alleles in HIV). NPAS1 is expressed in neuronal tissue and may be linked to the plethora of other neuronal genes differentially expressed between populations, potentially explaining reports of facial and neural deformities in some southern populations.

The SLC26A6 variants that were differentially expressed are alternatively spliced transcripts of this gene at the 5′ end (Fig. 2). There were two nucleic acid changes compared with the reference sequence present. The first was a T to C at position 11 in transcript XM_020998995.1, resulting in a leucine to proline change in the amino acid sequence at this location (and in the corresponding location in transcript XM_020999000.1). This could result in a shortening of the initial alpha helix of this protein’s secondary structure (Online resource 7). The second was a G to C change at position 1924 in XM_020998995.1 and the corresponding location in all other differentially expressed alternatively spliced transcripts of this gene resulting in a conservative valine to leucine change that nonetheless could also result in a shortening of an alpha helix in the protein secondary structure (Online resource 7). In both cases the QLD animals were either homozygous or heterozygous for the version present in the reference genome while the SA animals were uniformly homozygous for the alternate SNP (Fig. 3) displaying a significantly higher expression. The four animals from SA with confirmed pathology consistent with oxalate nephrosis (samples SAR1352A10, SAR2433A1, SAR2433A5 and SAR2433A6) show higher than average expression of transcript XM_020999000.1 compared to other SA animals (Fig. 4).

Fig. 2
figure 2

Sequence alignment of the 5′ ends of the alternate transcripts of SLC26A6 differentially expressed between SA and QLD animals, location of T-C SNP at position 11 in transcript XM_020998995.1 is marked with a blue triangle

Full size image
Fig. 3
figure 3

Percentage of animals with each combination of SNPs in SLC26A6 transcripts (./. stands for missing in that transcript. Line 1 = G to C at position 1924 (G1924C) of XM_020998995.1, line 2 = T to C at position 11 (T11C) of XM_020998995.1. Orange = QLD animals, Blue = SA animals. QLD animals are homo- or heterozygous for reference nucleotide. The SA animals were uniformly homozygous for alternate nucleotides

Full size image
Fig. 4
figure 4

Transcript expression levels of the missense variants of transcript XM_020998995.1 at position 11 in SLC26A6. SA animals (triangles) with the G genotype (C in mRNA) have a higher expression level (TPM normalised expression) than QLD animals (circles) with either the A (T in mRNA) or G (C in mRNA) genotype. The four animals from SA with confirmed pathology consistent with oxalate nephrosis (samples SAR1352A10, SAR2433A1, SAR2433A5 and SAR2433A6) are highlighted as red triangles

Full size image

Discussion

This paper takes a novel approach in the use of RNAseq data from two geographically distinct koala populations (one of which is genetically restricted) to examine differentially expressed genes and differentially segregated genes that are potentially linked to disease syndromes seen in the genetically restricted population. While such approaches are used in human and domesticated or semi-domesticated species to explore specific genetic traits (Cammen et al. 2016; Steiner et al. 2013; Jones and Good 2016), their application to wildlife species has not been previously developed.

The most striking link between the data in this study and disease in southern koalas is that of the anion transporter SLC26A6 which transports anions, including oxalate, in the kidney and intestine. This gene is both differentially expressed and demonstrates differential segregation of SNPs between the populations. This anion transporter has been the focus of several recent studies in mice and rat models, where knockout animals with this gene functionally removed demonstrated a propensity to the development of hyperoxaluria and/or oxalate urolithiasis (Freel et al. 2006; Jiang et al. 2006). Conversely, overexpression of renal SLC26A6 has also been found to lead to increased oxalate excretion and renal oxalate deposition in rats (Jiang et al. 2018), showing the complexity of the mechanisms of oxalate secretion. Case control studies in humans with oxalate urolithiasis are also exploring the role of SLC26A6 in oxalate uroliths in people, so far demonstrating a link between high expression of the gene (whether functional or not is not known) and oxalate kidney stones in people (Monico et al. 2008; Lu et al. 2016).

Based on our findings it is likely that SLC26A6 is either directly or indirectly linked to the high prevalence of oxalate nephrosis and hyperoxaluria seen in the SA animals. The SNPs in the SLC26A6 gene may be fixed in this population, and in the animals tested in this study this appears so. However only four of the 19 koalas (21%) in the study showed clinical oxalate nephrosis, suggesting that more complex factors may be involved in disease manifestation. It is tempting to speculate that the SNPs identified here affect the protein tertiary structure and reduce its effectiveness in oxalate secretion, driving a compensatory (but ineffective) feedback loop, increasing mRNA production of the SLC gene (as it was upregulated). Whilst the four animals with clinical signs of oxalate nephrosis show high expression, it is, however, not clear what effects the SNPs identified in this study have on the protein’s structure and function, or whether the primary defect is in this protein or another in the oxalate metabolic pathways, thereby driving an increased expression of SLC26A6 to increase oxalate secretion. Other genes of interest for the cause of oxalate nephrosis in koalas (Speight et al. 2018), including those that have been well described in primary hyperoxaluria in humans, are not expressed in lymph node tissue in other animals (Danpure and Rumsby 2004), and hence would likely not have been detected in the current study. If these genes were also involved in koalas with oxalate nephrosis, the SLC26A6 upregulation could possibly reflect a systemic response to increased oxalate load in SA koalas. Therefore, a targeted approach to investigating genes of interest and their expression is needed, with comparison between large cohorts of affected and non-affected koalas.

In the case of the marked propensity in the SA population to testicular abnormalities such as aplasia, the downregulated expression of the androgen receptor, a key gene in genital development and differentiation in mammals (Krausz 2011) and associated with male infertility in humans, gives an indication of which primary developmental pathways could be involved with this syndrome in koalas. It is possible that there is perturbation and compensatory increase in some genes in testicular development pathways due to aberrant decrease in critical regulators of sex differentiation such as the androgen receptor. Of critical note, there is no particular reason why these genes should be differentially expressed in lymph node tissue (which has little role in the functions of these genes), indicating that there may be a systemic (and not tissue specific) change in expression of these genes. None of the 12 male SA koalas in the study cohort were affected by testicular abnormality, hence further studies are needed in affected individuals to investigate expression of these genes in testis tissue.

The presence of scoliosis and kyphosis which has been reported in koalas in the SA population, and was present in two SA koalas in the study cohort, may be linked to the large number of genes associated with brain development as many genetic neuromuscular developmental disorders in humans present with scoliosis as part of the phenotype (Mary et al. 2018; Tully et al. 2018). However, the large number of genes differing between the populations in this study indicate that this phenotype is probably multifactorial in koalas. In addition the underlying defect (or defects) may not be in these genes per se but the regulatory networks controlling them, though in this respect it is important to note that transcription factors known to drive expression of these gene networks were not differentially expressed in this data set.

The “syncytin-1” like transcripts (Online resource 9), which are differentially expressed in different directions (2 upregulated and one down regulated in SA animals), are variants of a “type D” retroviral env gene (Henzy and Johnson 2013) and appear to be an endogenous retrovirus closely related to the previously described brush tailed possum (Trichosurus vulpeca) retrovirus TvERV (Online resource 9) (Baillie and Wilkins 2001). Up-regulation of human endogenous retroviral transcripts (HERVs) in response to exogenous retroviral infection has been well described for HIV infection of human cells (van der Kuyl 2012) and with the advent of long read deep sequencing capable of reliably distinguishing differing loci of HERV groups (there are up to several hundred of each type in a genome) it has become evident that HIV infection preferentially upregulates transcription from some loci while repressing others (Young et al. 2018), the mechanisms of which are not yet clear. It would not be unreasonable to expect that the high levels of KoRV expression in the QLD group (compared with the SA group) (Tarlinton et al. 2017) are affecting expression of other endogenous retroviruses in these animals similar to the effects HIV infection has on HERVs.

The results give an insight into the genetic basis of testicular abnormalities, oxalate nephrosis, and microcephaly syndromes seen in the genetically restricted SA populations and provide a starting point for targeted testing for the underlying genetic mutations causing these problems. They also provide a warning to managers of koala populations as to the types of developmental disorders that may be seen with further restriction of this population, such as might occur with groups used as founder populations in new areas.

There are limitations when working with a wild population which would ideally be addressed in a more targeted study. A definitive genetic study for SNPs associated with disease in these populations should involve a genome wide association study of DNA, a more carefully controlled “case” population and a much larger sample set than in this study. Case and control sample sizes in similar human studies of complex diseases are typically in the thousands (Spencer et al. 2009). However, there are a number of practical difficulties in achieving this in a vulnerable wildlife species where sampling is by necessity opportunistic. Given these limitations, the data provided here provides clear indication of which metabolic pathways may be involved in or altered by heritable genetic disorders in SA koalas and will provide a valuable starting point for population managers interested in the range of genetic diversity seen in different koala populations in Australia. In addition, the recent bushfire crisis with its subsequent effects of a drastic reduction in koala numbers has provided clear urgency to understanding which animals and populations are best suited to breeding and translocation programmes to maintain genetic diversity and population health in this vulnerable species.