The transcriptome landscape of the carcinogenic treatment response in the blind mole rat: insights into cancer resistance mechanisms

Background Spalax, the blind mole rat, developed an extraordinary cancer resistance during 40 million years of evolution in a subterranean, hypoxic, thus DNA damaging, habitat. In 50 years of Spalax research, no spontaneous cancer development has been observed. The mechanisms underlying this resistance are still not clarified. We investigated the genetic difference between Spalax and mice that might enable the Spalax relative resistance to cancer development. We compared Spalax and mice responses to a treatment with the carcinogen 3-Methylcholantrene, as a model to assess Spalax’ cancer-resistance. Results We compared RNA-Seq data of untreated Spalax to Spalax with a tumor and identified a high number of differentially expressed genes. We filtered these genes by their expression in tolerant Spalax that resisted the 3MCA, and in mice, and found 25 genes with a consistent expression pattern in the samples susceptible to cancer among species. Contrasting the expressed genes in Spalax with benign granulomas to those in Spalax with malignant fibrosarcomas elucidated significant differences in several pathways, mainly related to the extracellular matrix and the immune system. We found a central cluster of ECM genes that differ greatly between conditions. Further analysis of these genes revealed potential microRNA targets. We also found higher levels of gene expression of some DNA repair pathways in Spalax than in other murines, like the majority of Fanconi Anemia pathway. Conclusion The comparison of the treated with the untreated tissue revealed a regulatory complex that might give an answer how Spalax is able to restrict the tumor growth. By remodeling the extracellular matrix, the possible growth is limited, and the proliferation of cancer cells was potentially prevented. We hypothesize that this regulatory cluster plays a major role in the cancer resistance of Spalax. Furthermore, we identified 25 additional candidate genes that showed a distinct expression pattern in untreated or tolerant Spalax compared to animals that developed a developed either a benign or malignant tumor. While further study is necessary, we believe that these genes may serve as candidate markers in cancer detection. Electronic supplementary material The online version of this article (10.1186/s12864-018-5417-z) contains supplementary material, which is available to authorized users.


Background
The long-lived (> 20 years) subterranean rodent Spalax (genus Spalax ehrenbergi complex) is a solitary inhabitant of sealed underground burrows in the Eastern Mediterranean region [1]. While these tunnels protect the animals from predation and climatic extremes, they are also prone to sharp drops in O 2 levels, which can reach 7% in the rainy season [2]. During these periods, Spalax performs intensive digging work under low O 2 conditions and rapid re-oxygenation. Spalax displays an extraordinary hypoxia tolerance that has evolved over 40 million years of the species existence in habitats with frequent drops in oxygen. In the laboratory, Spalax can survive 3% O 2 levels, which makes it one of the most hypoxia-tolerant animals known [3,4]. The extraordinary hypoxia tolerance of Spalax is achieved through several morphological and physiological mechanisms of respiration-related genes and their regulation [3][4][5] including high blood vessel density, which results in a shorter diffusion distance for oxygen and high levels of activity compared to other murine species [3][4][5].
The mitochondria's increased production of reactive oxygen species has been observed [6] under hypoxia/reoxygenation cycles, which in turn leads to oxidative stress and damage to the DNA. Because hypoxia also causes dNTP depletion and the repression of DNA repair pathways [5] Spalax has fine-tuned various hypoxia-tolerance strategies during its life history, which seems to have driven a large degree of resistance against cancer development in order to allow survival and fitness in its subterranean habitat. Indeed, spontaneous cancer has never been observed during decades of research on Spalax. A key mechanism that contributes to cancer prevention is the high repair capacity constantly high expression of DNA repair and editing machinery, especially genes associated with the Fanconi anemia DNA repair pathway in Spalax [5,7,8]. Another notable mechanism is the heparanase alternatively spliced variant, which inhibits extracellular matrix degradation, tumor growth, and metastasis [9].
A previous study has reported that Spalax individuals are predominantly resistant to carcinogenic compound treatment and that fibroblasts play a prominent role in this resistance [10] In that study, Spalax fibroblasts were able to inhibit growth and kill adjacent cancer cells of humans and other mammals via direct cell-to-cell interaction in co-culture models, or via fibroblast-generated conditioned medium when transferred to cancer cells growing alone. This scenario was accompanied by decreased cancer cell viability, disturbed cell cycle progression, nuclei deformation, and mitochondrial fragmentation.
In the current investigation, we have attempted to elucidate some of Spalax's molecular mechanisms that enable cancer resistance at the gene-regulation level. We compared the response of Spalax and mice to a single subcutaneous injection of the chemical carcinogen 3-methylcholanthrene (3MCA) to the left flank of the neck. All 12 injected mice developed the expected fibrosarcoma tumor 10 to 14 weeks following the injection. Out of the 22 injected Spalax individuals, however, two animals developed a benign granuloma after 14 and 16 months, and only two developed the expected fibrosarcoma 18 months [10] and 30 months (reported here) after the injection.
By comparing the transcriptome of Spalax's and mice's injected versus uninjected tissues and untreated samples, we unraveled differences in the regulation of some of the major genes and pathways that seem to contribute to Spalax's significantly high cancer resistance. We found a higher elevation of anti-cancer pathways in Spalax and that many of the differentially regulated genes belong to three major groups: the innate immune system, the extracellular matrix (ECM), and DNA repair machinery. We have concluded that the stronger activity of the Spalax's innate immune system and the elevated expression of tumor suppressor genes related to the ECM compared to those in mice, may play an important role in Spalax tumor surveillance. These regulations, together with the Spalax's basic higher levels of expression of genes that belong to DNA damage repair pathways and DNA metabolism, may enable Spalax's higher tolerance to carcinogens (in particular 3MCA) and to its generally extraordinary cancer resistance.

Results
Spalax and mice were treated with the carcinogen 3MCA; muscle tissue samples were then collected from the injection site, the developed tumor, the opposite (uninjected) side, or the untreated control (Details shown in Table 1). Since we had samples from various conditions, we tried to organize the investigation in the following fashion. To uncover a possible genetic predisposition to cancer resistance and cancer influence on the surrounding tissue, we compared tissue from untreated Spalax to tissue taken from the uninjected side of the neck of Spalax with tumor growth. We extended this comparison to tissue from mice (which we consider tumor susceptible) to identify genes related to universal cancer development (Fig. 1).
In an attempt to uncover cancer-resistance mechanisms and to further understand what makes Spalax generally different in its non-susceptibility to cancer induction, we compared samples from tumor-resistant Spalax with samples retrieved from four Spalax individuals that failed to resist chemical carcinogenesis to different degrees.

Identification of genes expressed in carcinogen-tolerant Spalax
We took samples from various conditions in Spalax and mice (Table 1). Four samples of untreated Spalax animals as the control group were used, as well as four samples from animals that were injected but did not develop tumors after 18 and 30 months, which we term "Spalax tolerant." Out of 22 treated Spalax, four developed a tumor: two a malignant fibrosarcoma and two a benign granuloma. The granuloma samples were pathologically defined as spindle cell proliferation reflecting fibrosis at the site of an incompletely resolved inflammation, whereas fibrosarcomas were described as highly mitotic and heavily inflamed spindle and epithelioid cell tumors [10]. From these animals, samples were taken from the injected side, or After filtering, trimming, and mapping these RNA-Seq samples, we investigated the effects of the 3MCA treatment and the relation between these samples. We then clustered the expression values (Fig. 2). The analysis showed a separation of the samples by phenotype first, and then by species. The samples of Spalax untreated control and "Spalax tolerant" were direct neighbors. Close to this subgroup were the UIS samples of Spalax with fibrosarcoma and then granuloma. Mice control and mice UIS samples were also paired together but were still closer to the non-carcinogenic tissues of Spalax than to the carcinogenic tissue samples. All samples taken from tumor tissue were clustered in the same branch, first pairing samples from Spalax and then mice.
Because the development of tumors has an extensive impact on gene expression, it is generally difficult to distinguish between expression changes that allow tumor growth from those that are subsequently induced by the tumor. To assess how cancer development in Spalax affects genome expression in the neighboring tissues, we compared the Spalax control with the Spalax UIS from the animals with tumors by pooling them together. We discovered 68 annotated genes to be differentially expressed (Additional file 1: Table S2). To determine which of these 68 genes could be considered as plausible candidates for cancer susceptibility, we added mice samples to the analysis. Mice have a very low cancer resistance, so we were interested in the expression behavior of these genes in mice samples. We clustered the 68 genes according to their expression patterns and included all data sets at our disposal, excluding tumor tissue, namely Spalax UIS (of granuloma and fibrosarcoma bearing animals), mice control, and mice UIS (fibrosarcoma bearing). In this part of the study we omitted the tumor tissue to avoid introducing the large variances found in tumors.
In this clustering, we found 25 genes with distinct expression patterns (Fig. 3). Twelve of these genes displayed a general up-regulation in the susceptible samples compared to the Spalax control. These genes contain known cancer-related genes such as Jak3, Foxo1, and Cdkn1a (p21). The remaining 13 genes showed a general down-regulation in the susceptible tissue samples, i.e., of mice and tumor-bearing Spalax. We found known cancer-related genes such as Ntrk1, Serpina6, Nme3, and Sstr4. We also found significant differences in the level of expression of mitochondrial uncoupling protein (UCP1) in tumor-resistant samples (Spalax control and Spalax UIS cancer negative) in comparison with the cancer-sensitive group (mouse control, mouse UIS, and Spalax UIS cancer positive). The low level of UCP1 expression in the tolerant samples may have reflected a high level of natural defenses against cancer via mitochondrial homeostasis. Nucleoside diphosphate alkylase 2 (NME2), which is an active cancer suppressor, was highly expressed in Spalax tissues (Fig. 3).
We then conducted microRNA analysis in Spalax and mice tumor-bearing individuals. microRNAs are small (2 2 nucleotides) non-coding RNA molecules. They act as post-transcriptional modifiers that can silence a gene. We found nine microRNAs that were connected to the candidate genes, either as microRNA regulators or as potential targets. Differential expression analysis was performed for the comparison of UIS vs. IS in granulomas and fibrosarcomas in Spalax as well as for fibrosarcomas in mice (Additional file 1: Table S3). We found 27 microRNAs that targeted Cdkn1a and Foxo1. Since microRNAs act as suppressors of gene expression, we expected the micro-RNAs and the target genes to have opposing expression patterns in the three comparisons between the UIS and IS in Spalax granuloma, Spalax fibrosarcoma, and mice fibrosarcoma. Three microRNAs that target Foxo1 showed this pattern (miR-182, miR-223, and miR-96), and one for Cdkn1a (miR-296), as shown in Fig. 4. Although which genes regulate these microRNAs remain unknown to date, we hypothesize this pattern as a partial explanation for the regulation we observed.

Investigation of Spalax individuals that developed cancer
Since cancer development is a relatively rare event in Spalax, we were especially interested in investigating the  four animals that did develop tumors, specifically the transcriptome differences between those that managed to limit the growth to a benign granuloma and those that formed a malignant fibrosarcoma, similarly to mice.
To explore the differences between the two groups, we compared the uninjected side of the treated Spalax with the injected, tumor side of the animals using gene expression analysis. In the comparison of the samples of granuloma tissues, we found 2316 differentially expressed genes (DEGs). The fibrosarcoma samples showed 3293 DEGs. The intersection of these two gene sets consisted of 1736 genes. We split the functional analysis into three parts: 1) genes that were differentially expressed in both tumor types (Additional file 1: Table S7); 2) genes that were differentially expressed only in animals with granulomas (Additional file 1: Table S8); and 3) genes that were differentially expressed only in animals with fibrosarcomas (Additional file 1: Table S9).
Functional analysis of the DEGs showed that in all three comparisons, the dominant categories were the extracellular matrix, the citrate cycle, the cell cycle, and the immune system. These categories are all typically influenced by cancer development. We also directly compared the tissue from the uninjected sides of Spalax with fibrosarcoma and Spalax with granuloma (section 3.3). We found only 84 DEGs, indicating that despite the different tumors, the tissues were very similar. Once again, ECM-related categories dominated the enrichment analysis.
Since several of the tests showed an enrichment of genes belonging to the extracellular matrix and the immune system, we focused our investigation accordingly.
We focused on genes that were related to the extracellular matrix and collagen formation, as they were found to be significantly differentially expressed in both fibrosarcoma and granuloma tumors. This analysis showed that genes were generally up-regulated in the ECM categories (Additional file 1: Table S1). This finding is not surprising, since the modification of the extracellular matrix is a hallmark of cancer that allows growth and invasion into neighboring tissue and stimulates blood vessel growth. We continued to investigate the ECM-related genes on their connectivity, stability in expression (i.e., low variance between the samples of the same condition), and the known effects on tumor growth. We found a highly connected network of seven genes, namely Aspn, Chad, Fbln, Kera, Lum, Omd, and Sfrp1. (See Table 2 for a description of these genes.) The genes are connected via shared protein domains and co-expression, which were found using GeneMANIA annotation software [11] and various publications [12][13][14]. Figure 5 shows the connected genes, where Lum and Aspn are highly connected hub genes. Since it is difficult to determine the regulatory direction of co-expression, and a shared protein domain is by its nature undirected, it is also difficult to distinguish between regulators and targets. Our analysis showed that this gene complex was generally up-regulated in Spalax granulomas and mice fibrosarcomas (compared to uninjected tissue) while being down-regulated in Spalax fibrosarcomas.
In addition to the cellular matrix genes, we also found connections to the cell growth genes Igf2, Jun, and  [60]. Aspn inhibits TGFβ1, a biomarker of poor prognosis in cancer [61]. High asporin expression is significantly associated with less aggressive tumors, which could explain the benign growth.
Chad (Chondroadherin) expression has been linked to significant decreases in hepatocellular carcinoma, both in mRNA and protein levels. Chad abundance correlates with differentiation and metastasis, while the reduction of Chad levels significantly enhances proliferation and migration [39]. Chad has a α2β1 integrin binding sequence that previous studies have found counter cancer development [62].
Fbln5 (Fibulin-5) is frequently silenced in lung cancer and suppresses cell invasion by inhibiting the Wnt/β-catenin pathway [63]. Fbln5 has also been shown to reduce reactive oxygen species (ROS) production by modulating cell-ECM interactions [64]. Excessive ROS production results in cellular toxicity, which can promote tumor growth.
Lum (Lumican) has different effects on cancer development, depending on the type of tumor and whether Lum is expressed in the cell or in the adjacent stromal tissue [65]. Lum also acts as an oncogene, but if present in the surrounding stroma, it can restrict cancer progression and migration.
Sfrp1 (Secreted frizzled-related protein 1) is an antagonist of Wnt signaling, which gives Sfrp1 its tumor-suppressive effect. The loss of Sfrp1 has been hypothesized to activate MAPK or the non-canonical Wnt pathway [66]. One mechanism of how this loss can occur is the methylation of Lum via HP1α.

Kera
(Keratocan) encodes a protein that is involved in corneal transparency. Mutations can cause cornea plana [67].
Omd (Osteomodulin) is an extracellular matrix keratan sulfate proteoglycan that is also connected to bone development [68].

Igf2
(Insulin-like growth factor 2) is a mitogenic peptide hormone expressed by liver and many other tissues. It is closely associated with cancer [69,70].

Jun
Jun is part of the transcription factor activator protein 1 and has been closely associated with cancer development [15,72,73].
Cebpa. While they are not directly involved in matrix generation, they are connected to the central ECM gene Lum. Jun is part of the hetero-dimer AP-1, and centrally involved in prostate cell migration and invasion. Furthermore, AP-1 was reported having a significant role in melanomas [15]. Cebpa is down-regulated in gastric cancer, and potentially useful for its treatment [16]. Nevertheless, Igf2 and Cebpa were clearly regulated differently in Spalax fibrosarcoma samples than in the granuloma samples.
In mice, the picture was more ambiguous. While the genes related to cellular matrix complex are generally up-regulated, the intensity of the up-regulation is much smaller, and often not statistically significant.
While not presented in this network, other important ECM-related genes were also found to be differentially regulated in Spalax and mice. The gene Creb3l1 was found to be more highly expressed in Spalax granulomas compared to the fibrosarcomas of both mice and Spalax. In addition, Emilin2 is up-regulated and inhibits the Wnt-signaling pathway [17]. Other genes were also found to be up-regulated, including Bmp7, which antagonizes transforming growth factor β1 (TGFβ1)-mediated fibrosis through suppressing epithelial-mesenchymal transition [18], and Mmp8, which prevents metastasis formation through the modulation of tumor cell adhesion and invasion [19].
We then performed microRNA analysis of the ECM genes. We used the microRNA data to explain the impact of the genes involved. We found nine microRNAs that were connected to the ECM genes, either as micoRNA regulators or as potential targets. Differential expression  Table S4). Four of these nine microRNAs were found to be regulated by Cebpa, yet none matched the expression pattern of Cebpa as a gene. We believe that an additional regulatory mechanism must be responsible for this expression pattern. We must point out, that the knowledge about microRNAs and genes were taken from mice annotation. This has to be taken into consideration when evaluating the impact of these findings.
In this part of the study, we compared the two different tumor tissues (granuloma and fibrosarcoma) to uninjected tissue. We identified a tight network of genes from the extracellular matrix, which then allowed us to hypothesize that the modulation of ECM in Spalax granuloma largely contributes to host resistance.

Direct comparison of uninjected tissue of Spalax with tumor
In general, the transformation from normal cell to cancer cell is accompanied by an extensive altering of gene expression. In the current experimental setup, it was difficult to determine which changes in the observed major shift in gene expression were the drivers of cancer development and which changes were secondary to tumorigenesis. This situation is especially important in order to understand the difference between Spalax and mice. We started by comparing the uninjected side (UIS) with the injected side (IS) in Spalax to determine the extent of genetic rewiring. We considered a |log2FC| ≥ 2 as a threshold for highly differentially expressed genes (see Additional file 1: Table  S6). Among all DEGs in the described experiments, we filtered the genes that belong to these categories: cell cycle, citrate cycle, immune system, extracellular matrix, hypoxia response, and degenerative diseases ( Table 3). The ECM has the most differentially expressed genes of these categories, followed by cell cycle.
To investigate the differences between the species, we included the mice tumor samples in our analysis. We looked at genes that showed an expression difference between the species of |log2fc| ≥ 1. The highest differences were found in hypoxia-related genes (88/ 164 genes, 54% of the group members). The second group was related to the cell cycle control (84/239, 35% of the group members), the third was the immune system group (43/125 genes, 34%), and the fourth was the ECM group (57/269, 21%).
The elevation of Tlr4, Ifngr1, Ifnar2, Stat1, and Irf1 is presumably a part of an integrated pathway. The levels of Stat1 in Spalax in response to 3MCA injection were found to be five times higher compared to mice. These elevated Stat1 levels seemed to contribute to the tumor surveillance Interferons-I, which were found to systemically activate natural killer cell activity.
We also found stronger down-regulation of mTOR mRNA expression in Spalax compared to mice. In this regard, the elevation of the hypoxia-responsive Ddit4 in Spalax, compared to its down-regulation in mice, should also be noted, as up-regulation of Ddit4 expression mediates mTOR inhibition and growth inhibition in cancer cells [20].
By comparing the control, untreated Spalax, and mice groups, we found higher basic levels of expression of 75% of the 54 genes that belong to the Fanconi Anemia pathway. Fanconi Anemia genes are involved in the homologous recombinational repair of DNA double-strand damage, inter-strand crosslinks repair, and mismatch repair. The ratios between the control groups of the mRNA  expression (Spalax/mice) were found to be high for most of the genes (Table 4 and Fig. 6).

Discussion
The treatment with 3MCA resulted in the development of fibrosarcoma in 12 out of 12 mice within 3 months, while only 4 out of 22 Spalax developed tumors during this time; two of them developed a benign granuloma by 14 and 16 months, and two had developed the expected fibrosarcoma by 18 and 30 months following the carcinogenic treatment. For malignant tumors only, Fisher's exact testing showed significant differences (p-values of 3.32 × 10 −6 and 1.66 × 10 − 7 for all tumor outcomes).
We showed in this study that Spalax is highly resistant to the carcinogenic effects of 3MCA injection. While the investigation of "confirmed" cancer resistance is important for the understanding of cancer mechanisms, the investigation of failed cancer resistance is no less valuable. Of special interest in this study is the fact that four Spalax failed to escape tumor growth in different ways, where two of them managed to "contain" the process, which ended in the formation of benign tumors. Although two cases of each type were the smallest sample size for statistics purposes, this examination provided an opportunity to gain deep insights into the carcinogenesis of these subterranean mammals.
The first investigation of the samples via clustering showed that the most significant division between the samples was between tumor tissue and non-treated tissue, with the genus as the second divider (Fig. 2). Since tumors are generally considered to be organs developing in the context of organisms [21], tumor growth has a larger impact on the overall expression than the species factor. The Spalax tolerant samples clustered together with the Spalax control, which shows that the Spalax that did not develop a tumor were indeed affected very little by the treatment. The UISs of the fibrosarcoma and granuloma individuals were positioned on neighboring branches of the graph, showing that the tumor did seem to have a major impact on genome expression in the cells of neighboring tissue. Interestingly, the UISs of the fibrosarcoma tissue was closer to the Spalax control/tolerant than the UISs of the granuloma tissue, which was also visible during other parts of the investigation. We hypothesize that Spalax with benign tumors mounted substantial defense mechanisms, which were reflected in the extensive changes in gene expression in the neighboring tissue.
We found that the mice control and mice UIS of fibrosarcoma had their own branches and were still closer to the untreated samples of Spalax than to the clade with all the tumor tissues, which implies that extensive rewiring in the gene expression caused by the tumor growth occurred. We found 25 genes that behaved consistently different in animals that developed any type of tumor regardless of the species (Fig. 3). While some of these genes have been described in the literature as being related to cancer, our work can extend that knowledge by showing how these genes react in different species and conditions. Some of these genes are known cancer-promoting or suppressing genes such as Cdkn1a and Foxo1. While the precise mechanisms are unknown, and extensive investigations will be required to validate these findings, we propose these genes/gene networks as candidate genes that contribute to cancer resistance in Spalax.
Since the population of Spalax is not homogeneous, like any population of non-inbred wild animals, it is natural to assume that among the heterogeneous group will be those that cannot avoid the carcinogenic stimulus. On the contrary, the cancer-prone population of the laboratory mice demonstrated rapid response to carcinogen treatment by the development of fibrosarcoma. Thus, tissues for analysis from untreated and UIS treated mice were considered to be samples with very low threshold levels. To determine whether individual tumor sensitivity in Spalax is genetically pre-determined, we organized the samples for comparative analysis as follows [1]: cancer sensitive (mouse untreated and UIS of tumor-bearing mice as well as UIS of Spalax, in which we observed the neoplastic processes) and [2] cancer resistant (Spalax control plus Spalax tumor-negative UIS). See Fig. 3. Rmi2 6x Top3 5x Wrn 3x Telo2 5x List of Fanconi anemia genes that are up-regulated between Spalax control and mice control During tumor progression, already mutated cells constantly co-interact with stromal host cells, which can either promote or suppress the growth of malignant cells [22]. In this context, the outcome depends to a large extent on the abilities of the stromal host cells to maintain homeostasis, integrity, and oxidative metabolism. According to the paradigm of tumor cell metabolic parasitism, host cells "fuel" the anabolic growth of cancer cells through energy transfer [23].
The role of the mitochondrial metabolism in the development of cancer is of great interest. Mitochondrial uncoupling protein 1, or UCP 1 (also known as thermogenin), was originally associated with the brown adipose tissue of mammals. UCP1 promotes proton leakage from the inner mitochondrial membrane independent of ADP phosphorylation; this leakage uncouples respiration from ATP synthesis, thus dispersing the energy of oxidation in the form of heat [24,25]. Previous studies have demonstrated the unique changes in the UCP1 sequences of the naked mole rat and have discussed these sequences' possible roles in these mammals' thermoregulation and longevity [26]. UCPs are currently being investigated for their potential role in cancer initiation and development [27]. In studying the role of UCP1-overexpressing cancer-associated fibroblasts (hTERT-BJ1) in human breast cancer cell (MDA-MB-231) growth, a previous Fig. 6 The Fanconi Anemia pathway study has demonstrated increased β-oxidation, ketone body production, and the release of ATP-rich vesicles, all of which "fuel" tumor growth [23]. Conversely, the induction of mitochondrial dysfunction via UCP transfection of MDA-MB-231 cells results in significant reductions in tumor growth. Thus, if a rapidly growing tumor does not receive "fuel" from surrounding tissues, then the tumor may undergo involution by "eating itself." The comparative analysis of cancer-sensitive and cancer-resistant samples in our study revealed a dramatic difference in the levels of UCP1 gene expression (Fig. 3). We can assume that a high level of expression of UCP in tissues may represent a predictive marker for carcinogenesis susceptibility.
Another prognostic marker of cancer development, nucleoside diphosphate alkylase 2 (NME2), was differentially expressed in our study's mouse samples (low expression level) and in Spalax samples (high expression level), including samples derived from the UIS of cancer-sensitive Spalax (Fig. 3). NME2 have been extensively studied in the past for their cancer-suppressing activities [28]. Previous studies have shown that an overexpression of NME2 reduces the migration and invasion of gastric cancer cells to the cellular matrix in vivo and in vitro. Subsequently, NME2 expression is associated with the well-differentiated and less invasive histology of gastric cancer. The high level of expression of NME2 gene in the Spalax-sensitive samples found in our study may explain to a certain extent the significant delay of tumor development and the tumors' benign nature in half the cases. In addition, Somatostatin Receptor 4 (SSTR4), which we found to be more highly expressed in Spalax tissues (Fig. 3), could inhibit the proliferation of normal cells or tumor cells and suppress the formation of tumor vessels via releasing hormone-inhibiting tumor growth [29].
Even though four Spalax animals failed to resist the treatment with 3MCA in our study, this failure occurred in two distinct ways. We explored the differences between the benign granuloma and the malignant fibrosarcoma. One of our main findings was a cluster of key regulators in the extracellular matrix pathway. The ECM gene cluster including the Lum, Sfrp1, Fbln5, Chad, Kera, Aspn, and Omd genes was found to be distinctly more activated in granulomas than in fibrosarcomas. Especially noteworthy is that the activation of highly connected tumor suppressors and anti-metastatic genes such as Lum and Aspn may have explained the stunted growth of the tumor. The cell growth genes Igf2 and Cebpa are also up-regulated in granuloma. These genes are generally up-regulated in cancer cells, although Cebpa has also been reported to act as a tumor suppressor [30]. It was interesting to note that these cancer genes were activated in the benign tumor in our study, while the malignant growth seemed to have deactivated these genes..
The extracellular matrix plays a significant role in cancer, with both tumor promoting and inhibiting effects [31]. The switch in the citrate cycle is likely related to the Warburg effect, where the tumor switches from aerobic to anaerobic glycolysis. This switch enables the tumor to use the available glucose for other processes, such as cell division [32]. Removing cell cycle checkpoints is a basic requirement for the uninhibited proliferation of tumor cells, while cell cycle arrest is a common mechanism involved in tumor suppression [33]. Regulating the genes of the immune system is a crucial factor to prevent the organism from attacking the tumor [34].
Creb3l1 is an endoplasmic reticulum gene that is a transcription factor and a metastasis suppressor. It has been found to repress the expression of the genes that regulate metastasis, invasion, and angiogenesis in both breast cancer and bladder cancer [35]. We found in our study that the expression of Creb3l1 was higher in the granulomas of Spalax compared to the fibrosarcomas and may have been an important factor in preventing the development of malignant fibrosarcoma.
We attempted to elucidate the connections between these ECM genes in our study. As shown in Figs 5, 13 interactions were found to belong to the class of shared protein domain or co-expression. Because the nature of these interactions makes it impossible to infer a direction of the regulation, it is difficult to determine which gene induced a reaction and which genes were influenced. We did find a few differentially expressed micro-RNAs, mainly regulated by Cebpa. The expression patterns of the microRNAs and Cebpa did not match, however, thus indicating the presence of complex regulation activities and additional interactions that we are not aware of. Another explanation is, that there are species specific microRNAs, that are not present in mice.
The hypoxic environment Spalax lives in causes DNA damage through the occurrence of reactive oxygen species production. Spalax has adapted to this over the years by developing a very efficient DNA repair mechanism [36] that subsequently prevents persistent DNA damage and ultimately prevents cancer development by repairing harmful mutations. 3MCA is a DNA-damaging agent, and Spalax reacts to this damage by activating DNA-repair genes, especially the Fanconi anemia genes. In our study, most of the Fanconi anemia genes already had a higher expression in the untreated control groups. Similar results have been shown in a previous comparison between the transcriptome of the brain and liver [5,7] (Additional file 1: Table S5).
In addition to its enhancement of the DNA-repair effect, the hypoxic habitat of Spalax also appears to make the expression of hypoxia-related genes more attuned to countering cancer development. We found that more tumor suppressors up-regulated and more tumor promoters down-regulated in Spalax compared to mice in the hypoxia-responsive gene group. In this regard, one of the most important hypoxia-responsive genes is mTOR, which is a master regulator of cell growth control [37]. Previous studies have implicated downstream mTOR-regulated processes in the hallmarks of cancer: proliferative signaling, metabolic reprogramming, angiogenesis, and metastasis. We suggest that the down-regulated expression of mTOR in the hypoxic conditions found in the tumor microenvironment may lead to declines in the metabolic rate and may be an adaptive mechanism that Spalax uses to resist cancer development.
Another very important barrier to cancer development is the immune system's tumor surveillance. In our analysis, compared to mice, Spalax displayed a stronger and more robust activation of innate immune response of the tumor suppressors responsible for tumor surveillance. The elevated expression of Tlr4, Ifngr1, Ifnar2, Stat1, and Irf1 that was revealed upon carcinogen treatment appeared to reflect a part of an integrated pathway reaction. Previous studies have shown that these innate immune genes are important players in tumor surveillance [38,39]. The expression of Stat1 in our study might have been the major exception because of its higher basal expression in the untreated control group. An earlier study reported that Stat1 can inhibit the growth of benign and neoplastic cells by regulating the transcription and expression of several pro-apoptotic and anti-proliferative genes [40]. Irf1 is a transcriptional regulator and tumor suppressor that suppresses tumor cell growth and stimulates an immune response against tumor cells [41]. We believe that these changes in gene expression contribute to the Spalax-specific relative tolerance to the carcinogenic 3MCA compared to mice and might be a part of the layer of genes that enable Spalax's high cancer resistance.
As a potential additional defense, we found a stronger up-regulation of the immune system in Spalax compared to mice, especially of the innate immune-response tumor suppressors.
One of the major risk factors for cancer is aging. Spalax is known for its extreme longevity; the species can reach 20 years old in captivity, which is unusual for a mammal with such a small body mass. Since Spalax is also a solitary non-model organism that does not breed in captivity, the exact ages of specimens cannot be precisely determined, which could have been a confounding factor in our analysis. Due to this resistance, only four of the animals in our study developed a tumor, of two different phenotypes: granuloma and fibrosarcoma. These different factors created a high degree of variance in the data, since the samples were taken from only two different animals per tumor type and from animals of different sexes and ages. We addressed this variance in our study by rigorous normalization and filtering of the data.

Conclusion
We identified 25 genes in this study that we consider important for cancer development. Some of these genes are known to be related to cancer, while the roles of others are still unknown; we propose the latter as potential candidates for further cancer-resistance studies.
This study of the rare cases of failed cancer resistance in Spalax showed a tight network of genes that are responsible for the extracellular matrix remodeling. We consider these genes to be important for the difference between malignant and benign tumor development. This study has also pointed out several differences between Spalax and the model organism mice. The DNA repair genes of the Fanconi Anemia pathway were found to be especially important factors in cancer resistance.

Animals and tissue extraction
All animal protocols were approved by the University of Haifa Institutional Ethics Committee (Permit# 193/ 10).The experimental design and the animals tested in the experiments are listed in Table 1. Spalax was captured in the field and kept in individual cages in the Animal Facility of the Institute of Evolution, University of Haifa. Mice were purchased from Harlan Laboratories (currently Envigo, Jerusalem, Israel). Twenty-Two Spalax and 12 mice (C57BL/6) were treated with 3MCA. All animals were kept at 21°-23°C in a 12:12 light-dark cycle with free access to food and water. Untreated control Spalax animals were anesthetized with isoflurane when biopsies were taken (4% mixed in oxygen for the initial anesthesia; followed by 3% isoflurane flow for 20 min), then animals were kept in the Animal Facility under veterinary observation for recovery. Animals with tumor growth and control mice received 10% isoflurane, continued for more than 2 min after breathing stopped, when the samples were taken.
The 3MCA treatment was done according to Manov et al.'s [10] paper. A single subcutaneous injection of 3MCA was administered to the left-flank of the neck of the Spalax and mice. When the tumor length reached about 2-3 cm, it was completely removed. The most proximal~0.5 cm was harvested for RNA extraction, after pathological confirmation defining it as fibrosarcoma. For the control animals or the resistant Spalax individuals a similar sized neck muscle biopsy was removed for RNA extraction. The ages of the mice were 3 to 4 months. The age of Spalax cannot be accurately determined, as the species cannot be bred in captivity and are instead captured in the field; hence only their weight upon capture and the years they survive in captivity can serve as a reference of their age. Accordingly, we estimate the age of the treated animals to be between 2 years old to over 10 years old. Animals were checked for tumor growth once a week until such growth was palpable. The animals were then checked two to three times a week.
For histological examination, samples were fixed in 4% paraformaldehyde dissolved in PBS; they were then dehydrated in increasing concentrations of ethanol and embedded in paraffin. Form these blocks, five-micrometer sections were cut and stained with hematoxylin and eosin for microscopic histopathological examination to assess their tumorigenesis status.

RNA extraction and sequencing
RNA was extracted using the Qiagen miRNeasy Minikit according to the manufacturer's protocols. In short, frozen tissues were homogenized in QIAzol lysis reagent (Qiagen). After the addition of chloroform, the solution was vortexed and centrifuged. Ethanol was added to the aqueous phase and passed through a Qiagen cleanup column. Total RNA was eluted in 30 ul volumes. Quality was assessed on an Agilent bioanalyzer, and concentrations were determined by Qubit. Sequencing was performed by the Functional Genomics Unit of the W. M. Keck Center at the University of Illinois, USA. 1μg of total RNA was used to construct the libraries using the TruSeq Small RNA Sample Prep kit (Illumina). The strand-specific RNA-Seq libraries were prepared with Illumina's "TruSeq Stranded RNA Sample Prep" kit. The libraries were pooled in equimolar concentrations, and the pool was quantitated by qPCR and sequenced on one lane (small RNAs) or two lanes (RNA-Seq) for 101 cycles on a HiSeq2500 using a TruSeq SBS sequencing kit, version 4. Fastq files were generated and demultiplexed with bcl2fastq v1.8.4 conversion software (Illumina).

RNA-Seq filtering and mapping
Sequencing was performed by the Functional Genomics Unit of the W. M. Keck Center at the University of Illinois. The single-end RNA-Seq data were evaluated for their quality with Fastqc (www.bioinformatics.babraham.ac.uk). The reads were trimmed from adapters and regions of low quality using Trimmomatic 0.33 [42]. The filtered fastq files were stored in the Gene Expression Omnibus repository (GSE117501). To map the Spalax genome, we used the genome published by Fang et al. as a reference [8]. For mice, we used the reference genome published by NCBI. The reads were mapped to their respective genomes using TopHat2 [43] along standard parameters. Sorting and indexing were done using samtools [44], and the genes were counted using htseq-count [45]. The alignments were visualized using IGV [46,47].

Differential expression analysis
Gene expression analysis was conducted using EdgeR [48] and DESeq2 [49]. Genes were considered differentially expressed if they had an adjusted p value ≤0.05. While both algorithms produced similar results, we used the consensus of both results as being differentially expressed. Functional annotation was done in R [50] using the packages topGO [51], KEGGREST [52], and biomarRt [53] for the GO [54] and KEGG [55] annotation. Gene descriptions and gene synonyms were acquired through Ensembl [56]. Gene relations were visualized using Cytoscape [46]. Additional gene information was gathered with GeneMANIA [11] and the web-based tool cBioPortal [57].

microRNA sequencing, filtering and mapping
Libraries were constructed using the TruSeq Small RNA Sample Prep kit (Illumina, CA, USA). The libraries were quantitated by qPCR and sequenced on one lane for 101 cycles from one end of the fragments on a HiSeq 4000 using a HiSeq 4000 sequencing kit, version 1. Fastq files were generated and demultiplexed with bcl2fastq v2.17.1.14 conversion software (Illumina). Adaptors were trimmed from the reads, which were then mapped using mirDeep [58]. Differential expression analysis was conducted using DESeq2 [49], while microRNA targets and regulators were taken from OncomiRDB [59].

Additional file
Additional file 1: Table S1. RNA-Expression of Spalax. RNA expression divided between the different samples. Table S2. 68 differentially expressed genes in carcinogen resistant Spalax. Table S3. miRNA analysis in Spalax candidate genes. Differential expression analysis of miRNAs in Spalax samples. Table S4. miRNA analysis in Spalax ECM genes. Differential expression analysis of miRNAs in Spalax samples of ECM related genes. Table S5. Expression values of Fanconi Anemia genes. Differential expression analysis for Fanconi Anemia genes in Spalax. Table S6. Differential expression analysis of Spalax UIS vs tumor. A |log2FC| ≥ 2 is a threshold for highly differentially expressed genes. We filtered the genes that belong to these categories: cell cycle, citrate cycle, immune system, extracellular matrix (ECM), hypoxia response, and degenerative diseases. Table S7. Differentially expressed genes in fibrosarcomas and granulomas. Table S8. Differentially expressed genes in granulomas only. Table S9.