Low levels of IGFBP7 expression in high-grade serous ovarian carcinoma is associated with patient outcome

In this study we have shown for the first time significant IGFBP7 under-expression in HGSC samples relative to reference NOSE cells. These results are intriguing in light of our observation of IGFBP7 protein expression in epithelial cells of distal fimbrae of the fallopian tube, as both NOSE and fallopian tubes have been proposed as the progenitor cell type for HGSC [58,65,66]. Although it is difficult to directly compare gene expression with protein expression using IHC results, only about 8% of HGSC samples exhibited strong staining of epithelial tumour cells by IHC analysis of a TMA containing tumour cores as compared with about 72.6% of samples which showed absent or low staining levels. This suggests the possibility that IGFBP7 protein levels are indeed low or absent in the vast number of epithelial tumour cells of HGSCs.

We observed a significant positive association between IGFBP7 staining intensity in the epithelial cells of HGSC samples and overall survival. In our TMA analyses, the samples with no IGFBP7 staining were from patients that exhibited the poorest outcome, suggesting a role of IGFBP7 in molecular pathways associated with favourable outcome in HGSC. To our knowledge, our study is the first showing a significant correlation between expression of the protein and a more favourable patient outcome in HGSC. An association between low IGFBP7 expression and poor outcome was also observed in pancreatic ductal, breast and hepatocellular adenocarcinoma [26,35-38,40,42,44,45]. These observations suggest a role of IGFBP7 in molecular pathways associated with favourable outcome in HGSC and other cancer types. In contrast, IGFBP7 expression has also been associated with poorest outcome of patient with oesophagial adenocarcinoma [67] or colorectal cancer [68]. It has become increasingly apparent that the levels of IGFBP7 differ in different cancer types, where relatively higher levels of expression have been described in gliobastomas multiforme [63], oesophageal squamous cell [69], and colorectal carcinoma [70]. The differences in protein expression might reflect a context dependent function of IGFBP7, although this requires further exploration.

Our findings with HGSC tissues and normal cells are largely consistent with the gene expression in EOC cell lines. With the exception of TOV81D and TOV21G, our EOC cell lines exhibited low levels or the absence of both gene and protein expression. The IGFBP7 expression in TOV81D is interesting as this cell line is not tumourigenic in immunocompromised mouse tumour xenograft models and was derived from a sample from a patient having an unusually prolonged overall survival which was in excess of 7 years [19]. TOV21G was derived from a patient with a clear cell carcinoma, a disease which exhibits distinct clinical course that differs from HGSC, HGEC and undifferentiated adenocarcinomas, cancer types from which the other EOC cell lines were derived (see Table 1) [19]. Moreover, TOV21G is an unusually rare case of EOC in that it shows evidence of methylated alleles in MLH1, a mismatch repair gene [56] and thus it is possible that molecular pathways that implicate IGFBP7 differ in this cell line as compared to other EOC cell lines studied. Consistent with this notion is a recent report showing that TOV21G exhibited a ‘hypermutator’ genotype along with a low alteration of copy number alteration, a phenotype distinct from the other 46 EOC cell lines examined in the study and from HGSC [71]. Thus, our results, suggest that underexpression is associated with tumourigenicity and that expression of IGFBP7 may be important for tumour suppressor pathway phenotype.

The absence of IGFBP7 expression in OV90, TOV112D, TOV2223, TOV1946 and OV1946 is also interesting given the observation of LOH or allelic imbalance of the IGFBP7 locus at 4q12 in these EOC cell lines as compared to TOV81D and TOV21G. These findings are consistent with a high frequency (in greater than 50% of the tumours) of LOH and copy number loss involving the 4q arm as deduced by genome wide genotyping assays in our analysis of 79 HGSC samples that were examined in the present study and reported previously by our group [10], and in the analysis of 489 HGSC samples that were examined by the TCGA group [11]. Somatic mutation and epigenetic analyses of our cell lines, however, only demonstrated the possibility of promoter epigenetic silencing occurs in TOV2223G and TOV112D. These observations are consistent with recent reports from the TCGA analysis of HGSCs [11], where there was no evidence of IGFBP7 somatic mutations by exome sequencing analysis (n = 316 samples) nor for epigenetic silencing by CpG methylation array analyses (n = 489 samples) [11]. IGFBP7 intragenic mutations were also not found in relation to down-regulation in breast and colorectal cancers [54], although there is evidence of epigenetic silencing occurring in other types of cancer cell lines [24-26,41]. Thus, when taken together, other mechanisms may be involved in regulating IGFBP7 expression in HGSC.

The absence or low levels of IGFBP7 expression observed in HGSC samples is consistent with our observation that IGFBP7 was among the list of genes up-regulated (reprogrammed) in the context of suppressing tumourigenic potential in our chromosome 3 complementation assays involving OV90 [15]. IGFBP7 has been proposed as a candidate tumour suppressor gene as suggested by experiments demonstrating suppression of tumourigenicity in murine lung, prostate and colorectal, breast and skin cancers xenograft models with rIGFBP7 [25,26,35,62,72]. Thus, our observations that rIGFBP7 affects the migration rate of cells in in vitro wound healing assays of EOC cell lines is intriguing. However, rIGFBP7 appears to have no significant impact on cell proliferation, viability or spheroid formation (though there was a modest effect of spheroid formation with TOV112D). Wajapyee et al. reported that the induction of the apoptotic process by IGFBP7 largely occurred in NCI60 human cancer cell lines that harboured an activated BRAF or RAS mutation. Among the six ovarian cancer cell lines treated with rIGFBP7 in that study, only OVCAR5, which harbours an activated RAS mutation, exhibited susceptibility to rIGFBP7 as measured by the percentage of apopototic cells 24 hours after treatment [62]. Neither OV90 nor TOV112D EOC cell lines investigated in our study harbour activating BRAF or RAS mutations [19,46]. Moreover mutations in these genes are not a feature of HGSCs [7,11]. In breast cancer, IGFBP7 treatment inhibited cell growth and induced apoptosis and senescence, in vitro and in vivo, only in cell lines that were tested negative for HER2/neu, estrogen and progesterone [72]. In the same report, IGFBP7 effects were associated with strong activation of the p38 MAPK pathway and both p53 and p21^cip1 were up-regulated implicating known senescence pathways involving these proteins [72]. In light of these observations, it is therefore interesting that both OV90 and TOV112D harbour somatic TP53 mutations, a feature of over 90% of HGSCs [8-10,73]. Thus rIGFBP7 may have interacted with factors in alternative pathways to result in the effects observed with OV90 and TOV112D in wound healing assays.

In our study, rIGFBP7 protein inhibited the phosphorylation of both ERK and AKT protein in OV90 cell line, which is consistent with reports of similar assays with different human cell lines types [23,24,34,63,64]. Biochemical analyses have revealed that IGFBP7 interacts with activin, a member of the TGFβ superfamily of signalling proteins [29], and that IGFBP7 can be activated by TGFβ proteins and retinoic acid [74-76]. A recent study revealed that IGFBP7 acts as an IGF1/2 antagonist by directly binding to IGF-1 receptor (IGF1R) and hindering its activation and internalization, which results in blocking downstream phosphatidylinositol 3-kinase (PI3K)-AKT signalling and thereby inhibiting protein synthesis, cell growth and survival [34]. These observations are intriguing given the high frequency of AKT signalling pathway activation in HGSC [77-79], whereby the TCGA study identified alteration in the PI3K/AKT and RAS pathways in approximately 45% of HGSC [11].

Recent studies have reported an important role of IGFBP7 in therapy sensitization in different types of invasive cancers [80-82]. In acute myeloid leukemia cells, IGFBP7 cooperates with chemotherapy to induce cell cycle arrest and apoptosis and this mechanism is independent of ERK and AKT activation [82]. Interestingly, IGFBP7 has been identified with IGFBP4 in the secretome of mesenchymal stem cells and promotes their senescence [83]. While the mechanism has not been elucidated it has been proposed that this interaction participates in the inhibition of stem cell renewal and cancer development [83].