Molecular Expression Profile Reveals Potential Biomarkers and Therapeutic Targets in Canine Endometrial Lesions

Cystic endometrial hyperplasia (CEH), mucometra, and pyometra are common uterine diseases in intact dogs, with pyometra being a life threatening disease. This study aimed to determine the gene expression profile of these lesions and potential biomarkers for closed-cervix pyometra, the most severe condition. Total RNA was extracted from 69 fresh endometrium samples collected from 21 healthy female dogs during diestrus, 16 CEH, 15 mucometra and 17 pyometra (eight open and nine closed-cervixes). Global gene expression was detected using the Affymetrix Canine Gene 1.0 ST Array. Unsupervised analysis revealed two clusters, one mainly composed of diestrus and CEH samples and the other by 12/15 mucometra and all pyometra samples. When comparing pyometra with other groups, 189 differentially expressed genes were detected. SLPI, PTGS2/COX2, MMP1, S100A8, S100A9 and IL8 were among the top up-regulated genes detected in pyometra, further confirmed by external expression data. Notably, a particular molecular profile in pyometra from animals previously treated with exogenous progesterone compounds was observed in comparison with pyometra from untreated dogs as well as with other groups irrespective of exogenous hormone treatment status. In addition to S100A8 and S100A9 genes, overexpression of the inflammatory cytokines IL1B, TNF and IL6 as well as LTF were detected in the pyometra from treated animals. Interestingly, closed pyometra was more frequently detected in treated dogs (64% versus 33%), with IL1B, TNF, LBP and CXCL10 among the most relevant overexpressed genes. This molecular signature associated with potential biomarkers and therapeutic targets, such as CXCL10 and COX2, should guide future clinical studies. Based on the gene expression profile we suggested that pyometra from progesterone treated dogs is a distinct molecular entity.


Introduction
We obtained written owner consent before including any subject in the experiment. The animals were treated according to the norms of the Veterinary Hospital.
Endometrial samples were collected from female dogs, admitted to the Veterinary Hospital for elective or therapeutic ovariohysterectomy (OHE). The sample was composed of 16 cystic endometrial hyperplasia (CEH), 15 mucometra and 17 pyometra specimens (8 open and 9 closed-cervix pyometra samples). Two uterine samples with similar morphological features (approximately 1 cm 2 each) per case were collected from mid-portion of the right and left horn for molecular and histological analyses, respectively. Luminal exudate was excluded from the endometrial samples. Normal endometrium, CEH, mucometra and pyometra cases were diagnosed based on clinical information, physical examination, radiographic and/or ultrasonographic images, and further histologically confirmed by two experienced pathologists (FAV and RLA). The determination of estrus cycle phase was performed through anamnesis, clinical control, vaginal cytology and serum progesterone levels [9,10]. Diestrus phase was also subdivided into early, mid-and late diestrus (S1 Table). Fifty-five animals were in diestrus and 14 in anestrus. Normal endometrial samples were obtained from 21 healthy female dogs during early diestrus. Seventeen of 69 dogs (1 from diestrus, 4 from CEH, 2 from mucometra and 10 from pyometra groups) were previously treated with exogenous progesterone-like compounds to suppress estrous signs. Clinical features of all animals are shown in Table 1. Statistical analysis included Chi-square or Fisher exact test to determine the association between the categorical variables.

Blood Samples and Analysis
Before surgery, blood samples for hematological and progesterone (P4) serum levels were collected from the jugular vein using the vacuum system EDTA (BD Vacutainer Blood Collection Tube-BD, Franklin Lakes, New Jersey, USA) and EDTA-free (BD Vacutainer Serum Tube-BD, Franklin Lakes, New Jersey, EUA) respectively. Biochemical and hematological analysis were performed following routine protocol. Hematological parameters included hematocrit (HCT), white blood cell count (WBC) and platelet count (PLT). Serum P4 levels were measured by chemiluminescence immunoassay according to Tahir et al. [11]. Kruskal-Wallis or Mann-Whitney tests were applied to compare hematological data and P4 levels between the groups. Statistical analysis was carried out using SPSS version 17.0 (SPSS) and the GraphPad Prism 5 (GraphPad Software Inc.) software.

Tissue samples and Histopathological Examination
Uterine tissues were collected immediately after OHE. Samples were both formalin-fixed for histopathology and snap-frozen in liquid nitrogen, the latter being immediately stored at -80°C for mRNA extraction. Formalin-fixed material was paraffin embedded, with hematoxylin and eosin (HE) slides prepared for histological diagnosis (CEH, mucometra or pyometra). Representative photomicrographs of endometrium tissue sections from diestrus, CEH, mucometra and pyometra groups are shown in S1 Fig.

RNA extraction
Fresh frozen tissue samples were macrodissected using sterile scalpel blades, based on areas of endometrium identified following HE evaluation. Tissue samples were submitted to cleavage using lysing tubes in Precellys R tissue homogenizer equipment (BioAmerica Inc, Florida, USA). Total RNA was extracted using TRIzol reagent (Invitrogen Life Technologies Inc., Carlsbad, CA, USA) and the mRNA purified using the RNeasy MiniKit (Qiagen), according to the manufacturer's recommendations. RNA samples were quantified on a Nano-Drop ND-8000 spectrophotometer (Thermo Scientific, Wilmington, NC, USA) and analyzed using Agilent 2100 Bioanalyzer 6000 Nanochip (Agilent Technologies Inc., Waldbronn, BW, Germany). Only samples with a RIN (RNA integrity number) higher than 7.0 were considered for use in the gene expression experiments.

Clinical data
Eleven different breeds, with the majority being mixed-breed dog, were represented in all groups (Table 1). Higher serum levels of progesterone were observed in diestrus group compared with CEH, mucometra and pyometra (P<0.0001; P<0.0001; P = 0.006, respectively; S1 Fig and Table 1). Although pyometra samples showed a higher P4 levels in comparison with other uterine disorders, it was not statistically significant. Interestingly, significantly higher serum levels of P4 were detected in dogs with closed pyometra when compared with open pyometra (P = 0.0006, S2 Fig).

Molecular analysis
Unsupervised hierarchical clustering analysis revealed two main clusters, with mucometra and pyometra samples showing a strong trend to cluster together, and diestrus and CEH samples mainly grouped in the other cluster (Fig 1). Pyometra showed 189 differentially expressed genes in comparison with other groups (diestrus, CEH and mucometra), with 169 overexpressed and 20 underexpressed (S2 Table). The SLPI (secretory leukocyte peptidase inhibitor) gene was detected as having the highest fold change, being 30 times more expressed in pyometra compared with other endometrial tissues (S2 Table). Among the highest upregulated genes in pyometra, three metalloproteinase genes (MMP13, MMP1 and MMP12) and three S100 family members (S100A12, S100A8 and S100A9) were observed. Conversely, the EPHA7 (ephrin receptor A7) gene was detected as having the lowest fold change in pyometra. Canonical pathway and network analysis by IPA were performed for the differentially expressed genes in pyometra. Network analysis showed multiple interactions between the MMP genes and S100 family genes detected with the highest fold change (Fig 2A). In addition, important interactions between overexpressed CXCL8 gene, a chemokine that is one of the major mediators of the inflammatory response, and other genes were highlighted, including the PTGS2/COX2 gene ( Fig 2B). A comparison between gene expression data in pyometra versus normal endometrium using the Hagman et al. (2009) [8] findings and our data (pyometra vs. diestrus) was also performed. Concordance of 58 differentially expressed genes was identified, with 33 being underexpressed and 25 overexpressed in pyometra ( Table 2). The SLPI (secretory leukocyte peptidase inhibitor) gene was detected as being the top up-regulated gene in both studies. In addition, overexpression of the MMP and S100 family genes was confirmed. To identify specific altered genes in pyometra, the list of differentially expressed genes detected in pyometra versus other groups (diestrus, CEH and mucometra) was compared with the list of 58 common genes detected by Hagman et al [8] data's comparison. As a result, only 29 (50%) differentially expressed genes remained on the list, with 22 being overexpressed and 7 underexpressed ( Table 2, highlighted in bold).
The unsupervised clustering analysis (Fig 1) also revealed two distinct groups of pyometra, one of them being closer to diestrus and CEH samples, with the other clustering with CEH and mucometra samples, yet still isolated from them. Interestingly, the latter comprised a group of pyometra specimens obtained from bitches previously treated with exogenous progesterone compounds. In contrast, the pyometra samples that clustered near the diestrus and CEH samples were mostly obtained from untreated dogs (Fig 1). These findings indicated different molecular alterations in pyometra as a result of previous progesterone treatment. Estrous cycle phase and cervix conditions (open or closed) in each sample are also shown and were randomly distributed over the groups (Fig 1). In addition, the molecular profile was not correlated with histopathological features in each group (data not shown).
Further supervised clustering analysis was performed in CEH, mucometra and pyometra samples according to exogenous progesterone treatment status (Fig 3A). In accordance with previous unsupervised analysis results, two major clusters were observed, being one composed by all pyometra samples from treated dogs and the other cluster comprised by the pyometra samples from untreated animals as well as all by the other groups irrespective of treatment status ( Fig 3A). These data suggest that pyometra from hormone-treated dogs present a distinct molecular signature. Therefore, the molecular profile of pyometra according to previous hormone treatment was investigated. Supervised clustering analysis revealed 194 differentially expressed genes, in treated (N = 10) versus untreated (N = 7) dogs, with 57 being underexpressed and 134 overexpressed (Fig 3B). The top 20 differentially expressed genes detected in this analysis are shown in Table 3. Interestingly, three members of the S100 family (S100A8, S100A9 and S100A12) presented the highest fold changes among the up-regulated genes in pyometra from the hormone-treated group (Table 3). In addition, pro-inflammatory cytokines and chemokines (TNF, IL1B, IL6, and CCL3, among others) were also upregulated in the treated animals. Network analysis showed TNF as a central gene with multiple connections with other genes (Fig 2C) and an overexpression of the transcript factor E2F1 gene interacting with different genes (Fig 2D).
In order to identify a molecular signature for closed pyometra, a life-threatening condition, the expression profile of closed pyometra compared with open pyometra was investigated ( Fig  3C). Eighty-two differentially expressed genes were detected, but no significant difference was observed after Bonferroni correction. Characterization of exclusive molecular alterations in closed and open pyometra was also sought, with the aim of identifying putative biomarkers. Firstly, two lists of significant genes exclusively expressed in open pyometra versus diestrus and closed pyometra versus diestrus were generated and further compared to reveal exclusively altered genes in each group. Interestingly, closed pyometra revealed 70 exclusively altered genes, while open pyometra had 34 (S3 and S4 Tables, respectively). The top five-upregulated genes in closed pyometra were LBP, CCL3, IL1B, CXCL10 and ITGAM; while in open pyometra were FABP3, IL7, TNC, SDC1 and CLDN2. With the aim of revealing potential biomarkers and therapeutic targets for closed pyometra, a gene set enrichment analysis by IPA was performed, revealing 21 genes exclusively expressed in closed pyometra ( Table 4). The IL1B gene, which encodes a proinflammatory cytokine, was detected as the highest upregulated gene (Fold change = 9.29, Table 4). Overexpression of the CXCL10, NNMT, MMP8, F3 and TNF genes was also identified in closed pyometra (Table 4). Thereafter, PPI networks were constructed using NAViGaTOR based on genes with altered expression detected exclusively in closed and open pyometra, in order to highlight potential biomarkers and/or therapeutic targets to therapy for each condition (Fig 4). In closed pyometra, four up-regulated genes were revealed as potential biomarkers and therapeutic targets, including CXCL10, IL1B, KDR, and TNF. The LBP gene, detected with the highest fold change, was indicated as a potential diagnostic marker in closed pyometra. In open pyometra, four overexpressed genes (ITGAV, FGFR3, SRC and PTGS1) were indicated as potential biomarkers and drug targets.

Discussion
In this study, the global expression profile of pyometra was described in comparison to diestrus, CEH and mucometra, with a significant number of cases in each group. A distinct molecular profile for pyometra in female dogs previously treated with exogenous progesterone compounds compared with untreated dogs was described, indicating that they are, in fact, distinct molecular entities. This study also explored the molecular profile of closed pyometra, a condition that remains both a diagnostic and therapeutic challenge, in order to identify putative biomarkers and/or molecular therapeutic targets. Unsupervised analysis of transcriptomic profiles allowed the identification of two major groups, with one of them having essentially diestrus and CEH samples, and the other composed of 7/16 CEH samples, 12/15 mucometra and all pyometra specimens. These data corroborate the current knowledge that CEH may predispose to the development of mucometra and pyometra. CEH can lead to endometrial thickening with subsequent accumulation of fluids within uterine cavity, increasing the risk of pyometra or mucometra development [1]. A cross study validation was performed via the comparison of our data (pyometra versus diestrus) with the Hagman's data [8] revealing 58 genes in common in both studies, thus confirming our findings. However, 50% of these genes (29/58) remained as differentially expressed in pyometra after the comparison between pyometra versus other groups (diestrus, CEH and mucometra). This analysis, allowed the identification of altered genes exclusive to pyometra. The SLPI gene identified as the top overexpressed gene in pyometra encodes an antimicrobial peptide secreted by epithelial tissues. In addition, SPLI modulates infection and inflammation by neutralizing lipopolysaccharide (LPS) from Gram-negative bacteria impairing innate immune activation by toll-like receptors (TLRs) [12]. SLPI was previously reported as expressed in woman reproductive tract and induced by progesterone [13]. Up-regulation of the SPLI in pyometra may be endometrium-protective, both against microorganisms and from immune-mediated tissue damage.
Notably, both supervised analyses according to hormone treatment status evaluating all uterine disease samples (CEH, mucometra and pyometra) and only pyometra group confirmed a molecular signature for pyometra in treated dogs. These findings highlight the molecular alterations underlying the well-known association of pyometra with progesterone [2], [4]. Overexpression of S100A8, S100A9 and S100A12 genes were detected in pyometra, particularly in the hormone-treated animals. Several S100 proteins, including S100A8, S100A9 and S100A12, have been identified as endogenous danger-associated molecular patterns (DAMPs). DAMPs are intracellular molecules, which are released following cell death and can be recognized by the innate immune system, signaling tissue damage [14], [15]. S100A8 and S100A9 stimulate the production of the pro-inflammatory cytokines TNF, IL-6, IL-1B and IL-8 [14]. Accordingly, the IL1B and IL8 genes were detected as overexpressed in pyometra group, being IL1B, IL6 and TNF overexpressed particularly in pyometra of hormone-treated animals. Drugs targeting the S100A8/S100A9 complex leading to modulation of inflammatory response have been proposed in the treatment of cardiovascular disease [14]. Thus, high levels of S100A8 and S100A9 in pyometra may be a result of tissue damage leading to amplification and/or perpetuation of the local inflammation, being an attractive therapeutic target.
Overexpression of PTGS2/COX2 was detected in pyometra, confirming previous report [16]. COX-2, also detected as differentially expressed in pyometra of hormone-treated dogs, is the more important source of prostaglandins and thromboxane A2 in inflammation. Furthermore, COX-2 is downstream of TLR signaling after activation by endogenous S100 proteins and other stimulus like LPS [15]. Thus, in pyometra a continuous activation of TLR by both endogenous and exogenous stimulus can lead to an exacerbated inflammatory response. COX-2 interacts directly with CXCL8/IL-8 and CXCL14 (Fig 2B), chemokines that were also overexpressed in pyometra. High serum levels of IL-8 were observed in dogs with pyometra, particularly in those that developed Systemic Inflammatory Response Syndrome (SIRS), suggesting that IL-8 may contribute to the development of a systemic disease in dogs with pyometra [17].
Overexpression of CXCL14 can contribute to inflammatory cell infiltration in pyometra due to its chemoattractive action in immune cells, like monocytes and natural killer cells [18]. Both IL8 and CXCL14 are induced by COX2, thus selective COX-2 inhibitors could be an interesting treatment option in pyometra in terms of control the inflammatory response.
Overexpression of the LTF gene was detected in pyometra of hormone-treated dogs. Lactoferrin gene and protein overexpression were previously reported in pyometra [19]. Lactoferrin interacts with LPS on the bacterial surface and activates TLR-4 on the surface of phagocytes and epithelial cells. Interestingly, LTF can also be considered an inflammatory regulator by impairing LPS ligation to TLR and subsequent activation of the inflammatory response [20]. Thus, LTF might be induced in pyometra to counterbalance the inflammatory response. In contrast, the pro-inflammatory TNF gene was identified as a central gene in pyometra of hormone-treated dogs (Fig 3C). Considering that TNF has a pivotal role in inflammation, therapies targeting TNF could be considered in pyometra. In conclusion, a more pronounced inflammatory gene signature for pyometra of treated dogs was revealed, suggesting that repetitive progesterone exposure may contribute to bacterial colonization of the endometrium and an inflammatory response. This inflammatory environment enables the exposition and recognition of DAMPs by innate immune receptors, such as TLRs and production of pro- inflammatory cytokines and other inflammatory molecules, such as COX-2, aggravating the inflammatory process. With regard to closed pyometra, the inflammatory process can progress to sepsis and SIRS, present either with or without infection [21]. Interestingly, a high frequency of closed pyometra was observed among hormone-treated dogs (7/10 cases compared to untreated dogs (2/7). Further studies are needed to confirm if repeated exogenous progesterone treatment predisposes to development of closed pyometra. Treatment with progesterone compounds to suppress heat cycles in female dogs is common in countries where spaying dogs is not a routine practice. Thus, these results are clinically relevant and neutering strategies in intact animals should be considered since treatment with progesterone compounds might ultimately produce undesired results. However, care should be taken considering that ovariohysterectomy during diestrus can lead to subtle progesterone decline and hyperprolactinemia with subsequent pseudopregnancy [22].
In this study, detection of exclusively altered genes in closed and open pyometra combined with enrichment analysis by IPA and PPI analysis were described, suggesting potential biomarkers and molecular targets for therapy. In closed pyometra, overexpression of IL1B and TNF were suggested as therapeutic targets, specifically anti-IL1 and anti-TNF antibodies. In addition, CXCL10 overexpressed in closed pyometra is important in host immunity; but its excessive activation may lead to a detrimental immune response. The CXCL10 protein and its receptor (CXCR3) have been proposed as therapeutic targets in cancer and in immune-mediated diseases. Ongoing phase II clinical trials MDX-1100-anti-CXCL10, indicated here by PPI analysis in closed pyometra, have been conducted in inflammatory bowel disease and rheumatoid arthritis [23]. High serum levels of CXCL10, IL-6, IL-10, TNF and IL-8 were reported in a canine sepsis model [24]. Thus, CXCL10 in combination with other cytokines such as TNF are potential biomarkers for sepsis as a consequence of pyometra progression and also as targets in anti-inflammatory therapy in pyometra. LBP gene was also suggested as a diagnostic marker in closed pyometra. LBP is an acute-phase protein synthesized in the liver, but also by epithelial cells of other tissues, including the reproductive tract [25]. The LBP binds to LPS and other microorganism components leading to activation of the inflammatory response [25]. LBP serum levels have been proposed as a useful diagnostic marker in both urinary infections in children and atherosclerosis, as well as a prognostic marker in acute appendicitis [26]- [28]. Future studies are warranted to confirm LBP as meaningful diagnostic marker in closed pyometra. In open pyometra, PTGS1/COX1 overexpression was detected as an exclusively altered gene. COX-1 is expressed constitutively in most cells and is also responsible for prostaglandin production during inflammation, which has the potential for being targeted by numerous antiinflammatory drugs, including ibuprofen.

Conclusions
To the best of our knowledge, this is the first report that described the expression profile of pyometra compared with other common uterine diseases, such as CEH and mucometra, in intact female dogs. In addition, a molecular signature for closed pyometra was described, which should be further explored to reveal prognostic and predictive biomarkers. Finally, distinct pyometra expression patterns from progesterone treated and untreated dogs were observed, with a more pronounced inflammatory signature in treated dogs, which suggests that previous progesterone exposure may contribute to an exacerbated inflammation in pyometra. A high frequency of closed pyometra in the hormone-treated group was observed and future studies are warranted to verify the impact of progesterone treatment on the development of closed pyometra.