HtrA1 as a promising tissue marker in cancer: a meta-analysis

The members of the widely conserved high-temperature requirement A (HtrA) family of homo-oligomeric serine proteases are involved in a number of mammalian cellular processes that include growth [1], maintenance of mitochondrial homeostasis [2], apoptosis [3], and protein quality control [4]. HtrA1 has been reported to regulate such processes through modulation of growth factor systems, like the system mediated by the extracellular protein transforming growth factor β (TGF-β) [5] as demonstrated by the association of its expression with specific tumor behaviors.

Mammalian HtrA1 is connected to tumorigenesis [6, 7]; in particular, it is down-regulated in ovarian [8], thyroid [9], endometrial [10], and breast cancer [11], in hepatocellular carcinoma [12], and colorectal cancer [13] while its up-regulation seems to induce suppression of tumor cell proliferation and migration in highly invasive melanoma [6]. Indeed, several lines of evidence indicate that HtrA1 functions as a tumor suppressor in various solid tumors, such as ovarian and lung cancer and mesothelioma [8, 14, 15]. It also participates in regulating cancer cell apoptosis, invasion, and metastasis processes [16]. Distant metastasis, the final stage of solid tumors, is involved in most cancer deaths [16].

By using different experimental techniques, Lorenzi and colleagues [17] have found that HtrA1 protein expression is downregulated in urothelial cancer tissue regardless of tumor grade and stage, and suggested that urinary HtrA1 protein may be used as an early and highly sensitive and specific biomarker for this neoplasm.

Finally, according to recent evidence HtrA1 downregulation induces the acquisition of phenotypes such as increased proliferation, delayed onset of senescence, altered centrosome number and positioning and polyploidy, all hallmarks of tumor cells [18].

In a previous review, our group explored the potential role of HtrA1 as a tumor marker and/or prognostic factor [19].

This study reports the results of a meta-analysis that was performed: i) to compare HtrA1 expression as protein and mRNA, in cancer tissue (C) versus non-cancer tissue, i.e. healthy control (HC) tissue and normal-looking (NL) tissue; and ii) to assess overall survival in relation to low or medium-high HtrA1 tissue expression in different cancers.

Literature search. Relevant studies were identified as of June 2017 using the following databases: PubMed, Embase, Web of Science, Cochrane Library, Scopus, and clinical trial registers (clinicaltrial.​gov, clinicaltrialsre​gister.​eu). A manual search was also performed. There were no limitations as to the year of publication. Only studies in English were considered. The reference lists of all studies were screened by two independent reviewers (PMA and MM); any disagreements were resolved by a methodologist (EA). The following keys words were used: HtrA1 OR PRSS11 protein, human OR L56 protein, human OR protease, serine, 11 (IGF binding) protein, human OR high-temperature requirement factor A1, human OR HtrA serine peptidase 1, human AND Neoplasm OR Tumors OR Tumor OR Neoplasia OR Cancer OR Cancers. Only studies assessing HtrA1 (mRNA or protein) as a diagnostic or prognostic tumor marker were considered. Studies were selected using the PRISMA statement (Additional file 1: Figure S1) [20]. The excluded studies and the PRISMA check list are reported in Additional file 2: Table S1 and Additional file 3: Table S2. The general characteristics of each study (tumor site, sample size, patient population) are reported in Table 1. The Newcastle-Ottawa was used for case-control studies was used (Additional file 4: Table S3).

Search of the electronic databases according to the above-listed criteria found 59 papers, whereas the manual search found none. There were no duplicates (Additional file 1: Figure S1). In the first phase, 40 papers were excluded because they did not match the inclusion criteria. In the second phase, 19 full papers were examined and 4 more were excluded because they did not match the outcomes of interest [14, 31‐33]. Eventually, 15 papers assessing HtrA1 expression at 10 tumours sites (stomach, liver, bladder, breast, esophagus, thyroid, endometrium, pleura, ovary, colorectum cancer) were included in the meta-analysis [9‐13, 15, 17, 34‐41].

In each meta-analysis, sensitivity analyses indicated stability of meta-analytic findings.

In recent years, mounting biological knowledge of disease processes and advances in molecular technologies have increased the interest in biomarkers. In fact, biomarkers are used for patient assessment in multiple clinical settings, both to distinguish an individual with disease from one without disease and to discriminate one disease from another, i.e. for differential diagnosis. Ad hoc immune-histochemical markers can also be used to try and identify the tissue where a tumor has had its origin, to assess the risk of relapse, as well as to evaluate prognosis or response to therapy.

In the present study, a meta-analysis was conducted to assess the potential role of HtrA1 as a tumor marker and/or prognostic factor in a number of tumors. Most current studies focus on the role of HtrA1 in tumor development and progression through analysis of its expression as mRNA [11, 12, 17, 34‐36, 39‐41] or protein [9, 10, 13, 15, 17, 35‐38], probably because proteins are more varied than DNA or RNA and therefore carry more information than nucleic acids. Indeed, alternative splicing and more than 100 unique post-translational modifications result in tens (and possibly hundreds) of protein species from each gene. Moreover, a number of physiological changes are mediated post-transcriptionally and are not revealed at the nucleic acid level. Proteins are also more dynamic and reflective of cellular physiology. For instance, the occurrence of a single double-strand DNA break in a cell is rapidly amplified into a protein phosphorylation cascade. Thus, protein-based markers provide a fine, specific, and accurate representation of the condition being investigated.

Based on these considerations, in this work the studies assessing HtrA1 expression as mRNA and protein were evaluated separately. An important finding of the study was that, in the absence of significant statistical heterogeneity, there was a statistically significant difference between the tumor samples analyzed and HC and NL tissue specimens; this applied both to HtrA1 measured as mRNA and as protein (Table 2, Figs. 1 and 2a,). The results regarding the non-heterogeneity of studies were supported by the analysis of the moderators used, i.e. sample size magnitude, % of females, mean age for both genders, and year of publication, which did not exhibit statistically significant differences (Additional file 5: Figure S2A, D). Publication bias was highlighted neither by the funnel plot nor by Egger’s or Begg’s test. A consistent finding of this work, that was described in all the studies that tested this aspect [9‐13, 15, 17, 34‐41], was that HtrA1 levels are higher in HC or NL tissue than in diseased tissue from patients with a variety of tumors.

The meta-analysis of the prognostic studies showed, in the absence of significant heterogeneity among the studies examined, a different survival according to HtrA1 expression. These findings are in line with the literature, since in some studies HtrA1 has been found to modulated cisplatin- and paclitaxel-induced cytotoxicity and low HtrA1 levels have been seen to correlate with a poor response to drug treatment, whereas higher levels correlated with a greater response [38, 42].

Baldi et al. [15] have demonstrated a relationship between HtrA1 expression level and survival in patients with malignant mesothelioma, suggesting that HtrA1 expression can be used as a prognostic parameter for this tumor type. Analysis of HtrA1 levels in relation to overall survival and disease free survival in breast cancer [11] indicated that patients with higher HtrA1 levels had a better prognosis.

Analysis of the moderators assessed in the works selected found no statistically significant differences and there was no publication bias (Additional file 7: Figure S4).

However, some weaknesses in the data suggest that the results of the present meta-analysis should be taken with caution. First of all, most of the studies are descriptive, preventing a causal inference between reduced HtrA1 levels and cancer. Secondly, the small number of works on the same tumor prevents an analysis by tumor in relation to individual organs. Thirdly, since parameters such as histological type and tumor grade and stage have not been addressed in all the studies, it is impossible to establish how HtrA1 expression varies in relation to these factors. Finally, the different HtrA1 expression, found in papers assessing the same cancer type, could correlate to histological grading, metastasis, and degree of cell differentiation, but were not related to patients’ age or gender, as demonstrated in hepatocellular carcinoma [12], esophageal carcinoma [35, 36] and endometrial cancer [10, 40, 41]. This means that the reduced HtrA1 expression may be closely associated to tumor development. The lack of difference in HtrA1 expression reported by Catalano et al. (2011 [38]) may be due to heterogeneous baseline characteristics of their patients, like constitutional genetic factors and other variables, which as the authors themselves stated were not investigated.

The present data may provide a contribution to future research work directed at exploring the role of HtrA1 in tumor development and progression and at establishing whether it may become a promising tissue marker for some tumors. Finally, the present work suggests that clinical investigations sharing a similar approach, especially in terms of study design, should be carried out to improve comparability across studies.