Confounding Effects of Hormone Replacement Therapy in Protein Biomarker Studies

The serum/plasma proteome is a major compartment for diagnostics that informs about the state of health of most tissues and organs through biomarker interrogation. However, there are numerous challenges associated with studying alterations in the serum/plasma proteome in relation to diseases such as cancer, in part due to exogenous factors that can alter protein levels even if issues related to sample collection, processing, and storage that can affect protein stability and levels are minimized. Recently we applied in-depth quantitative proteomics to determine the effects of postmenopausal hormone replacement therapy (HRT) with estrogen and estrogen plus progesterone on the serum proteome (1, 2). Using serum collections from the Women's Health Initiative randomized trials, large-scale proteomic analyses were performed to compare protein levels in serum collected from women at baseline and 1 year after HRT. Ten experiments were performed comparing the serum proteome of women at baseline and 1 year after the administration of oral estrogen or estrogen plus progesterone therapy (1, 2). Baseline to 1 year post-HRT ratios and associated P values were computed for 382 proteins. 144 proteins had P < 0.05, compared to approximately 19 expected by chance. Thus, 44% of quantified proteins showed evidence of change in concentration between baseline and 1 year of treatment with estrogen and/or estrogen plus progesterone. Given the profound effect of HRT on the serum proteome, we investigated the potential for a confounding effect of this single exposure on candidate cancer markers. Specifically, we examined the extent to which potential ovarian cancer markers previously described may be affected by HRT.

We compiled a list of proteins reported in the literature since 2007 that have been suggested as potential markers for ovarian cancer and that have been assayed in blood (3–19). More than 60 such proteins have been reported as potentially useful for ovarian cancer detection. However, no single candidate marker has matched the performance of CA125. Data were available for 40 candidate ovarian cancer biomarkers with respect to effects of postmenopausal hormone therapy on their circulating levels (Table 1). Ninety percent of the proteins in Table 1 are secreted, whereas ALCAM, CD14, VASN, and VCAM1 are membrane-associated proteins that may be released into the circulation through shedding. The effect of HRT was computed from 10 experiments comparing levels of circulating proteins at baseline and 1 year after the administration of HRT (1, 2). The P values in Table 1 were calculated using a t test across the 10 experiments to determine statistical significance of change in protein level with HRT. Of the 40 proteins with HRT data, 19 (52.5%) showed significant (P < 0.05) changes in their serum concentration with HRT. Nine exhibited increased levels and 10 proteins exhibited decreased levels with HRT. Interestingly, 4 proteins (GRN, LCN2, MMP2, and VCAM1) contained palindromic estrogen response elements in their gene sequence that are conserved between mouse and human (20). However, none of these 4 proteins had significantly altered serum levels with HRT, suggesting that most of the changes in protein levels observed were as a result of secondary effects of HRT as previously noted for most of the proteins affected by HRT in our initial studies (1, 2).

Remarkably, 10 members of the insulin-like growth factor pathway (IGF1, IGF2, IGFBP1, IGFBP2, IGFBP3, IGFBP4, IGFBP5, IGFBP6, IGFBP7, and NOV) have been reported as potential ovarian cancer biomarkers (3, 6, 9–13, 16, 17, 19). Of these 10 proteins, 9 are affected by HRT (P < 0.05; Table 1), with associated increased circulating levels of IGFBP1, IGFBP4, and IGFBP6; decreased levels of IGF1, IGFBP2, IGFBP3, IGFBP5, IGFBP7, and NOV; and no change in IGF2. To determine whether the effect of HRT is related to particular forms of these proteins that may result from alternative splicing or other types of processes, we searched the data for peptide sequence coverage for these proteins (Fig. 1). Intact protein fractionation prior to tryptic digestion and mass spectrometry resulted in the separation of 2 different forms of IGF1. A mature form was downregulated following HRT. Another form that represented the propeptide was also identified but lacked quantification. IGF2 was also identified as an intact form that encompassed the propeptide. No quantitative differences were observed in relation to HRT. Each of the IGFBPs (including NOV) was represented with a cleaved signal peptide, and with peptide coverage across most of the protein sequence.

Total inhibin, consisting of the dimeric and free α-subunits, has been reported as a potential ovarian cancer biomarker (15, 19). In our studies, quantitative information was available for inhibin beta E and C (INHBE, INHBC), both of which were upregulated with HRT. Figure 2 displays the peptide coverage for INHBC. Two forms of INHBC are observed, the cleaved propeptide and the mature processed protein. The propeptide exhibited increased levels with HRT, whereas the mature protein was unchanged with HRT. Figure 3 displays the individual peptide measurements for peptides corresponding to the propeptide and mature form. The average (log2) ratio is 0.70 for the propeptide and −0.13 for the mature form.

In addition to proteins that have been assayed in the blood as potential ovarian cancer biomarkers, we investigated proteins that were recently described as potential markers by Kulasingam et al. (21) on the basis of their proteomic identification in ovarian cancer cell lines (6, 22), ascites (23, 24), and tumor tissue (25), and oligonucleotide microarray experiments (26). Table 2 presents proteins that were identified in at least 3 of 6 ovarian cancer proteomic and/or transcriptomic studies for which data for the effects of HRT on their levels was available. Levels of 11 of 20 proteins presented in Table 2 were found to be affected by HRT (P < 0.05), with 8 proteins having increased levels and 3 proteins having decreased levels.

In previous studies (1, 2), we have described the substantial effect of hormone therapy on circulating protein levels. Here we examine the implications of these findings further by looking at the explicit effect of HRT on potential biomarkers reported for ovarian cancer. Hormone therapy affects a significant portion of the serum proteome and hence has the potential to confound studies aimed at assessing circulating proteins as potential biomarkers for diseases that affect older women. The analysis presented here pertains to postmenopausal HRT. Given the widespread effect of oral estrogen treatment on the serum proteome, it is feasible that premenopausal oral contraceptive use could cause significant alterations in the serum proteome. For example, circulating levels of sex hormone binding globulin, which have shown to be elevated with HRT (1, 27), are also elevated with oral contraceptive use in premenopausal women (28). Analyses presented here pertaining to potential candidate markers for ovarian cancer demonstrate that circulating levels of a substantial number of these candidates are affected by HRT. This effect and potentially effects resulting from other environmental exposures and/or drugs may confound biomarker studies and should be taken into account as part of the study design and data analysis. Subjects in discovery studies may need to be stratified on the basis of such exposures. Samples were matched by age for several studies in Table 1 (3, 4, 11, 13, 15, 16); however, only 2 studies accounted for postmenopausal hormone therapy use in study design. One study included only age-matched samples from postmenopausal women who were not using oral contraceptives or menopausal estrogens at blood draw (13), and a second study matched cases and controls by HRT use in addition to age (3). The interpretation of protein changes with disease state should be approached with caution, as confounding factors, unrelated to the disease state, as illustrated here with HRT, may affect circulating levels. Validation studies may also need to incorporate common medications and dietary supplements as covariates that may impact biomarker levels.

It should be noted that our analysis of hormonal effects is based on oral administration. In contrast to transdermal administration, orally administered hormone therapy is known to affect production of a large number of proteins by the liver, which itself is a major source of circulating proteins. In this respect, the proteins listed in Table 2 are not associated specifically with the liver. They are produced by cancer cells and are found in proximal fluids, and more than half have circulating levels that are affected by HRT. Estrogen receptors, which regulate gene expression, are widely expressed in a many tissue types. Hence, increased exposure to estrogen of a variety of tissues could result in altered levels of expression of proteins secreted into the blood.

In addition to the study presented here, growth hormones and sex steroids have been previously reported to affect circulating levels of proteins belonging to the IGF binding protein family and complement system (29–31). Previously we described the overall quantitative changes of protein levels of IGF family proteins with HRT. Here, we examine in greater depth the quantitative changes of particular protein isoforms by assessment of mass spectrometry sequence coverage, which can provide further insight into how particular protein forms may be influenced by HRT. The peptide coverage for IGF1 and IGF2 in Fig. 1 indicates 2 different isoforms of IGF1, and 1 form of IGF2. The mature form and the cleaved propeptide of IGF1 are observed, whereas the propeptide is still attached to the mature form of IGF2. The downregulation of IGF1 levels may be related to the activated form, whereas the nonprocessed form of IGF2 accounts for unchanged levels with HRT.

Inhibins are proteins known to regulate the pituitary hormone, FSH and play a role in reproduction (32). Although inhibin beta C was found to have overall increased circulating levels with HRT, closer inspection of peptides observed reveals 2 isoforms of INHBC. The separation of the propeptide and mature form, which elute in separate fractions during the intact protein separation prior to tryptic digestion and mass spectrometry analysis, suggests that the protein has undergone processing. The propeptide is upregulated with hormone therapy, whereas the mature form of the protein is not changed. Although propeptides are typically inactive, they can be turned into active forms by posttranslational modification. The propeptide of inhibin beta C has 3 known N-linked glycosylation sites. No tryptic peptides containing glycosites are identified in these studies, suggesting that the propeptide is likely glycosylated.

The serum proteome has great potential for disease diagnostics; however, the complexity caused by confounding effects, as demonstrated here for HRT, should be considered in biomarker discovery study design and validation studies.

No potential conflicts of interest were disclosed.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.