As predicted, substantial differentiation exists in gene expression between lip and tumor tissues. We observed differential expression patterns consistent with DFTD’s origin as a Schwann cell tumor [22], including genes involved in nervous system functions such as synaptic signaling and ion transport, as well as characteristics common to cancers [5]. Lips, in contrast, had differentially expressed genes characteristic of muscle and epidermal tissue function. Within hosts, we found surprisingly little transcriptional variation in between the sexes, regardless of infection status, despite clear sex-biased differences in resistance and tolerance of DFTD [31, 36]. Perhaps most interestingly, we found significant variation in gene expression among DFTD tumors collected from different localities, potentially associated with different tumor strains, environmental effects (either within-host or extrahost) or host-pathogen coevolutionary relationships. However, we found no association between detected SNP variants in transcribed genomic regions and gene expression, consistent with results from previous studies that showed variants in regulatory regions [37], which are not captured in RNA-seq approaches, likely affect observed phenotypic variation.
Gene expression characteristics of DFTD
Despite considerable genotypic and phenotypic diversity among cancers [47], there are relatively common mutational drivers that are necessary for the genesis and proliferation of cancerous cells. Functions of these drivers include: the sustenance of proliferative signaling, evasion of growth suppressors, resistance to apoptosis, induction of angiogenesis, activation of invasion and metastasis, reprogramming of energy metabolism, and evasion of immune response [48, 49]. Similarly, we observed differential expression of genes and enrichment of pathways in DFTD associated with the extracellular matrix, the cell cycle, DNA replication and repair, and immune function. The extracellular matrix is fundamental to maintaining tissue homeostasis. Dysregulation of the extracellular matrix can both cause – and occur as – a response to the development and proliferation of cancerous cells, facilitating uncontrolled proliferation, angiogenesis, tissue migration and invasion, and metastasis [50–52]. Cell cycle ‘checkpoints’ detect DNA damage or errors in DNA replication and chromosome organization and are regulated by critical tumor suppressor genes that prevent proliferation of defective and potentially cancerous cells [53]. A key example of such a gene is TP53, which is activated in response to DNA damage, initiates the inhibition of cell proliferation and regulates apoptosis [54]. Loss of TP53 function is common to many cancers [54, 55]. We found various pathways associated with TP53, such as G1 and G2 cell cycle checkpoints, to be downregulated in DFTD relative to normal lip tissue. In turn, upregulation of pathways associated with mitosis and the cell cycle that likely lead to increased cell proliferation were also found in DFTD.
Despite downregulation of cell cycle checkpoint and apoptosis pathways, we observed upregulation of homology-directed DNA repair (HDR) in DFTD. Dysfunctional HDR often results in genomic instability and can lead to carcinogenesis [56, 57]; however, cancers can become resistant to standard cancer treatments aimed at inducing DNA damage through reactivation of HDR, emphasizing that cancer cell proliferation is not dependent on HDR suppression [58]. Active yet dysfunctional HDR can lead to errors in double-stranded break repair that produce gross chromosomal rearrangements [59], such as those that are evident in DFTD [21, 60, 61].
Although all cancers exhibit some form of immune avoidance, DFTD and other transmissible tumors are unique in that they can evade host MHC recognition of non-self-cells [24]. In DFTD, MHC avoidance is achieved through the ERBB-STAT3 axis [41]. Specifically, DFTD became transmissible when a normal Schwann cell tumor gained a variant exhibiting overexpression of ERBB3, which overactivates the transcription factor STAT3 and blocks the production of MHC I [23, 24, 41]. Without MHC I expression, DFTD cells are unable to be recognized as foreign by the host. Similar to previous work, we found the differential expression of other genes associated with STAT3, including upregulation of MMP2, HDAC5, and downregulation of PTGIS [41]. However, in contrast to this previous study that found upregulation of the TRIM28 protein, which is typically activated in response to STAT3 signaling, we detected slight downregulation in expression of this gene. TRIM28 is often overexpressed in cancer [62]; however, its expression has been shown to be predictive of tumor class in human glioblastomas [63], and positively correlated with tumor size and development stage, whilst being negatively correlated with patient survival in human hepatocellular carcinoma [64]. Lack of detectable TRIM28 overexpression may thus reflect size or developmental stage variation among the sampled DFTD tumors.
We also observed downregulation of Notch signaling, which is critical to Schwann cell development and required for differentiation of Schwann cell precursors (SCP). Notch activation upregulates ERBB2, which acts as a receptor for NRG1, which in turn is critical for transitioning SCPs to immature Schwann cells [65, 66]. In the absence of Notch signaling, a scarcity of ERBB2 receptors would lead to reduced sensitivity of SCPs to NRG1, despite its overexpression. Dysfunction of these pathways suggests a decoupling of DFTD cell proliferation from typical Schwann cell developmental controls.
Gene expression in devil lip tissue is not associated with infection status or sex
DFTD infection has no effect on gene expression in devil lip tissue. This is surprising as DFTD has significant effects on host physiology and can elicit host immune response [67]. Tumor growth leads to increasing metabolic demands concurrent with difficulty feeding, ulcerations, metastases, and secondary infections that produce a progressive loss of body condition and almost universal mortality within 12 months following visible tumor development [18, 31]. In humans, gene expression in normal tissues adjacent to tumors reflects a state that is intermediate between healthy and cancerous, with commonly expressed pathways including pro-inflammatory responses induced by the tumor [68]. We chose lip tissue due to its proximity to the mouth, where DFTD allografts typically implant, believing that changes in gene expression would be greatest in tissues local to tumors. However, DFTD tumors do not always occur on the lips or in the mouth, and there was likely variation in the proximity of lip biopsies to the site of tumor growth. Ethical and experimental concerns necessitated a consistent sampling strategy for healthy tissues, preventing individual adjustments of biopsy locations to accommodate variation in tumor location. In addition, lip tissues may have been inappropriate for detecting systemic immunological or metabolic changes. For example, systemic immune responses associated with tumor growth tend to involve the accumulation of immune cells in the peripheral blood or lymphoid tissues, rather than in the epidermis [69]. Biological functions associated with immune response were not underrepresented in lip tissues, suggesting that the lack of differential expression was not specifically due to a lack of immune expression in these tissues.
Despite a lack of a direct immune response in devil lips to DFTD infection, we observed upregulation of immune-associated genes in BR relative to TKN devils, irrespective of infection status. This difference is not necessarily associated with DFTD and may be due to differential exposure to other infectious agents or innate differences in immune function between genetically distinct populations. However, DFTD is an overwhelmingly strong selective force in devils [34, 36, 70] and likely drives immune adaptation in affected populations. Further, documented immune responses to DFTD [67] suggest that differential host immune expression between localities may alter the microenvironment faced by invading tumor cells. We thus recommend further gene expression studies targeting immunologically active host cells; specifically, to investigate systemic host responses to DFTD using blood samples as well as a refined approach for detecting localized responses by targeting healthy tissues < 1 cm from DFTD tumors.
Male and female devils exhibit different responses to DFTD, with females losing body condition at a significantly slower rate when infected and genetic evidence of adaptations that result in greater survival rates among females [31, 36]. Although DFTD infection produced no transcriptomic response in lip tissues for either sex, we found several genes that were differentially expressed between males and females generally, regardless of infection status. Consistent with previous work comparing sexes in uninfected devils [71], the X-linked gene FRMD7 was downregulated in males (log2FC = -4.01). FRMD7 is putatively involved in fatty acid metabolism and has been associated with skin disorders, serving as a potential factor in differential susceptibility between sexes to DFTD [71]. Additionally, we found six other genes that were differentially expressed between sexes. Four of these were uncharacterized, while HMGB3 and MECP2 – both also X-linked – were downregulated in males. HMGB3 is a DNA-binding protein that helps maintain stem cell populations and is overexpressed in some human cancers via the Wnt signaling pathway [72, 73]. Further, HMGB3 affects nucleic acid recognition and innate immune system activation, and its upregulation has been associated with allograft rejection [74–77]. Therefore, higher baseline expression of HMGB3 in female devils may enhance the innate immune response to DFTD relative to males. Recently, MECP2 was identified as an oncogene through induction of the MAPK and PI3K growth factor signaling pathways and is overexpressed in many human cancers [78]. However, it is unclear how MECP2 expression in normal host tissues could affect DFTD progression.
DFTD tumor gene expression varies geographically
We observed considerable variation in tumor gene expression among the sampled localities, which varied in the length of time since DFTD arrival. Interestingly, gene expression patterns that we observed in DFTD relative to host lips were more intense (i.e., more differentially expressed genes and greater log2 fold-changes) in WPP tumors – where DFTD has been present longest among our study sites – than in other populations. Specifically, WPP tumors exhibited upregulation of mitosis and downregulation of translation, DNA damage checkpoints, and immune function relative to tumors from BR, which had the shortest time since DFTD arrival. Varying intensity of DFTD-characteristic gene expression may be due to differences in the ratios of normal-to-tumor cells within tumor biopsies, potentially driven by subtle differences in tumor morphology among localities (e.g., differences in the extent to which tumor tissue is delineated from the surrounding host tissue, or the distinctness of tumor margins). On the other hand, expression changes in cancer-associated genes are not only linked to tumorigenesis but can be directly correlated with tumor aggressiveness and overall prognoses in human cancer patients [79, 80]. Meanwhile, incrementally greater gene expression changes through time can produce progressively more severe phenotypes or reflect more advanced stages of tumor development [e.g., 81, 82]. The strength of expression in DFTD-associated pathways therefore may be associated with DFTD phenotypic variation, such as growth rates, although no such differences (nor differences in tumor morphology) between the studied localities have been documented.
Gene expression in tumors from TKN reflected an intermediate state between WPP and BR but were differentiated by sex (Fig. 2c; Additional file 7: Figure S7). That is, tumors from TKN males exhibited patterns of gene expression that were characteristic of WPP tumors from both sexes and tumors from TKN females resembled male and female tumors from BR (see Fig. 2c). Perhaps coincidentally, previous work demonstrating more rapid body condition loss in DFTD-infected males than in females was conducted in TKN [31], while no similar analysis of host body condition has been performed for either of our other study sites. We do not have sufficient serial volume measurements for the tumors in our study to directly associate gene expression with tumor growth rates. In the absence of data comparing tumor growth rates or host body condition between TKN, WPP, and BR, it is difficult to establish a link between relative levels of DFTD-characteristic gene expression the severity of disease.
Although it remains unclear whether transcriptional variation in DFTD occurs among localities as a result of neutral or adaptive differences among strains, there is an interesting temporal trend that correlates with expression changes. DFTD first arrived in WPP in 2006 and has thus had the longest amount of time to adapt to devils in that locality. DFTD then emerged in TKN in 2011, and tumors there showed intermediate expression values between WPP and BR, where DFTD has been for the shortest time (since 2016). Thus, expression variation among study sites may reflect ongoing DFTD adaptation to devils within populations, which themselves have demonstrated rapid evolutionary responses to DFTD [34, 36]. Additionally, multiple DFTD lineages with differing degrees of pathogenicity have been observed. Specifically, in WPP, DFTD arrival in 2006 was characterized by initially low mortality rates compared to localities that had experienced DFTD for longer [83]. Karyotype analysis confirmed the existence of a distinct tetraploid strain at WPP that was replaced by a more virulent diploid strain in 2012, causing an immediate increase in disease prevalence and population decline [84]. In addition, recent phylodynamic analysis indicates multiple tumor lineages exhibiting differences in transmission rates, demonstrating epidemiological variation among distinct tumor strains [85]. Given its recent emergence, our BR tumor samples may represent a novel DFTD lineage present near the advancing disease front. Such a lineage was not observed at a broad spatiotemporal scale [85] but may be evident through more intensive sampling of the most recently emerged tumors near and on the west coast of Tasmania.
We detected weak genetic structure in tumors that broadly reflected transcriptomic variation among localities and may reflect different tumor strains. However, we acknowledge that the persistence of a small number of host variants in the tumor dataset may produce a residual signal of host population structure, despite rigorous bioinformatic filtering to exclude known devil variants genotyped from lip RNAseq data as well high-coverage whole genome sequences (see Additional file 3: Text S3). Further, although we identified a number of variants in tumors that were predicted to affect the function of genes differentially expressed among localities, no significant associations were detected. Thus, transcriptional variation in tumors may be purely regulatory in origin, or we may have lacked sufficient statistical power (19 samples) to identify genotype-expression associations. Alternatively, population genetic structure and adaptive and plastic responses to the local environment in hosts can produce variation in immune function that drive differential responses to wildlife disease [14, 86]. Our study sites have similar vegetation communities and experience similar climatic conditions but decrease in elevation from east to west. Region-specific adaptation of devils to local environmental conditions exists, as well as significant selection imposed by DFTD following its arrival in naïve populations [34, 36, 70]. Different transcriptomic responses of DFTD to devils from different localities may thus also be explained by immunological variation among devil populations that is driven by environmental differences that lead to differential exposures to non-DFTD infectious agents.