BiP

Triplex Profiling of Functionally Distinct Chaperones (ERp29/PDI/BiP) Reveals Marked Heterogeneity of the Endoplasmic Reticulum Proteome in Cancer Steven D. Shnyder,†,‡ Jonathan E. Mangum,§ and Michael J. Hubbard*,†,§,| Department of Biochemistry, University of Otago, New Zealand, Institute of Cancer Therapeutics, University of Bradford, England, and the Departments of Pharmacology and Paediatrics, University of Melbourne, Australia Received February 15, 2008

The biomedical need for streamlined approaches to monitor proteome dynamics is growing rapidly. This study examined the ability of a knowledge-based triplex-profiling strategy (i.e., three functionally distinct chaperones, ERp29/PDI/BiP) to clarify uncertainties about how cancer affects the endoplasmic reticulum (ER) proteome. Investigating a wide range of samples at the tissue and cellular levels (>114 samples from 9 tissues of origin), we obtained consistent evidence that the ER proteome undergoes a major but variable expansion in cancer. Three factors having a strong influence on the ER proteome were identified (cancer-cell type, growth rate, culture mode), and the functionally enigmatic chaperone ERp29 was linked distinctively to histogenetic aspects of tumorigenesis. These findings justify pursuit of the ER-proteome as a medical target in cancer, validate ERp29/PDI/BiP profiling as a streamlined yet powerful measure of ER-proteome dynamics, and suggest that biomarker sets based on distinct functionalities could have broader biomedical utility. Keywords: knowledge-based approaches to biomarker selection • proteome dynamics • multiplex profiling • clinical proteomics • cancer biomarkers • organellar proteome • endoplasmic reticulum biology • chaperones

Introduction The recent availability of comprehensive proteomic data sets has opened exciting opportunities to discover new biomarkers and develop predictive models with clinical value. With the onus now shifting toward clinical validation and mechanistic understanding, there is an increasing need to monitor (sub)proteome dynamics rapidly. Generally, high-throughput proteomic approaches harness the power of multivariate statistics and are hypothesis free, offering great potential to detect sample variations sensitively. However, several problems are faced, including the practical difficulties of analyzing numerous samples and the uncertain mechanistic relevance of statistically derived models.1,2 In contrast, the traditional cell biology approach typically involves univariate profiling of one or two well-characterized marker proteins whose selection is based on subcellular location. Although rapid and biologically informative, such mono/duplex approaches offer relatively low sensitivity and rigor. In this study, we sought to strengthen the traditional approach through multiplexing and proteomically informed selection of biomarkers. Our focus was on the * To whom correspondence should be addressed. Mike Hubbard, Department of Paediatrics, Royal Children’s Hospital, Flemington Rd, Melbourne, Victoria 3052, Australia. Tel: +61 3 9345 5028. Fax: +61 3 9345 6667. E-mail: [email protected]. † University of Otago. ‡ University of Bradford. § Department of Pharmacology, University of Melbourne. | Department of Paediatrics, University of Melbourne.

3364 Journal of Proteome Research 2008, 7, 3364–3372 Published on Web 07/04/2008

endoplasmic reticulum (ER), an organelle of interest to many fields including cancer. The ER is known to play several pivotal roles in cancer and thereby offers a variety of opportunities for medical exploitation.3–9 However, better understanding of the underlying protein machinery (ER proteome) is required if these practical avenues are to be pursued knowledgeably. Fundamental activities embraced by the ER include protein production, calcium storage, redox regulation, drug metabolism, and protection against stress. Thus, the ability of cancer cells to survive and grow in the face of natural and therapeutic challenges is affected broadly by functional status of the ER. Besides appealing as a universal target for functional manipulation, the ER proteome might also offer diagnostic and therapeutic selectivity based on cancer-related changes in its composition.3–7 Such expressional differences are recognized for soluble proteins residing in the ER lumen, termed reticuloplasmins. Major reticuloplasmins including BiP, endoplasmin and protein disulfide isomerase (PDI) are established as key players in cancer, but uncertainty how these and other components of the ER machinery function collectively compromises their targeting. As a chaperone of protein folding, BiP (synonyms: GRP78, immunoglobulin-binding protein) plays central roles in protein production and survival of cellular stress, making it an attractive target for cancer therapy and prognosis.4,5,8 Endoplasmin (GRP94, gp96), another stress-sensitive chaperone related to BiP, has prominence in cancer immunotherapy and as a target of the anticancer agent, geldanamycin.7,10,11 The 10.1021/pr800126n CCC: $40.75

 2008 American Chemical Society

research articles

Triplex Profiling Reveals Heterogeneity of the ER in Cancer redox sensor PDI is essential for disulfide bond formation and, due to its broad overexpression, has been proposed as a cancer biomarker.12,13 These classical ER-marker proteins are recognized to hold strong mechanistic, therapeutic, and prognostic potential given their status as multifaceted sensors of cell pathophysiology and drug exposure.5,8,9 However all three proteins are multifunctional and possess overlapping activities also shared by other ER residents (i.e., as chaperones, calcium buffers, disulfide editors, and stress sensors). The cancer field therefore faces a major challenge to understand the nature of, and reasons for, cancer-related changes in the ER proteome. Proteomic studies of cancer have suggested the ER machinery undergoes a coordinated expansion akin to that long-associated with physiological secretion.12,14 Yet the literature remains dominated by inconclusive and often contradictory reports about cancer-related variations in ER proteins.5,13,15–17 A second concern is the knowledge gap surrounding recently discovered reticuloplasmins,18,19 which in the case of ERp29 brings a new class of componentry to the ER.20,21 ERp29 was discovered as a novel reticuloplasmin through proteomics,21,22 and subsequent findings make a broad involvement in cancer seem plausible. Like the classical reticuloplasmins mentioned above, ERp29 is expressed ubiquitously and abundantly in metazoan cells, implying it has a general ”housekeeping” function. Distinctively however, ERp29 lacks classical chaperone, disulfide-editing,calcium-buffer,andstress-responseproperties.22–27 Participation of ERp29 in the production of secreted and endomembrane proteins is supported by several lines of evidence, yet its specific function remains an enigma (see reviews in refs 21, 28). ERp29 has been found highly expressed in a variety of tumors and cancer cell lines, but an apparent paucity in other cases leaves both the nature and generality of its involvement in tumorigenesis unclear.14,15,17,29–31 Elucidation of ER dynamics in cancer is currently hampered by the lack of streamlined approaches for ER-proteome profiling. Knowing the ER proteome comprises over 1,200 proteins,19 a central question is whether to pursue statistically driven proteomic approaches for biomarker selection or instead draw on knowledge of well-characterized ER proteins. In the latter case, it remains unclear how many and which ER-marker proteins might be used to best advantage. Previously we quantified expression levels of ERp29, PDI, and BiP across a range of healthy tissues and found major variations in the comparative profiles (ERp29/PDI/BiP), leading to useful insights about the distinct functionality of ERp29.23,25,26 These findings also led us to realize that comparative profiling of these three ER chaperones provided a rapid yet highly sensitive measure of ER-proteome heterogeneity. In this study, we hypothesized that profiling of three functionally distinct chaperones (ERp29/PDI/BiP) across a variety of cancers would elucidate any cancer-related variations in the ER proteome, and clarify the involvement of ERp29 in tumorigenesis. Our results have implicated ERp29 in tumor growth distinctively from PDI and BiP, and revealed other new aspects of ER-proteome heterogeneity relating to cell type and growth environment. It thus appears this streamlined triplexprofiling approach will have broad utility for elucidating dynamics of the ER proteome, in cancer and elsewhere. More generally, a useful principle may be to base selection of biomarker sets on distinct functionalities.

Experimental Procedures Antibodies. Polyclonal antibodies to ERp29 and PDI were described previously.26 Polyclonal antibodies to BiP (sc-1050), ATF6a (sc-22799), and XBP-1 (sc-7160) were purchased from Santa Cruz Biotechnology. Appropriate controls were used throughout.26 Lactogenesis Samples. Wistar-derived rats were maintained conventionally and with national regulatory approval as before.23 Inguinal mammary glands were dissected at various time points in the lactation cycle (detailed under Figure 1) and stored at -80 °C.32 For the postlactation samples, weaning was carried out prematurely (after 17 instead of the normal 21 days) to ensure uniform termination of suckling activity. Tumor Specimens. Spontaneous mammary and salivary adenocarcinomas, freshly dissected from Sprague-Dawley rats then stored at -80 °C, were donated by Dr. John Schofield (Laboratory Animal Sciences, University of Otago). Tissue microarrays of human cancer specimens (lung, colon) and tumor xenografts were obtained from the Institute of Cancer Therapeutics, University of Bradford. Microarrays were constructed from four 0.6 mm cores of each formalin-fixed, paraffin-embedded specimen as described.33 Lung and colon tumors were staged according to the International TNM system34,35 and provided by the Department of Pathology, Bradford Royal Infirmary. Ethical regulations of the Medical Research Council (UK) were met during procurement and use of these clinical specimens. Tumor xenografts were grown conventionally from human cancer cell lines maintained at the Institute of Cancer Therapeutics, as described below. Cancer Cell Lines. The human cancer cell lines COLO 205, HT29, SW-620 (colon adenocarcinoma), and MCF-7 (mammary carcinoma) were purchased from LGC Promochem (Middlesex, UK) and cultured as monolayers in RPMI 1640 cell culture medium supplemented with 1 mM sodium pyruvate, 2 mM L-glutamine, and 10% fetal bovine serum (all from Sigma, Poole, UK). Multicellular Tumor Spheroids. HT29 cells were grown on 1% agar for 2-3 days and then the resulting spheroids were transferred to spinner flasks (Techne Inc., Burlington, NJ) and cultured in suspension (50 rpm) for 31 days, as described.36 Tumor Xenografts. Female CD1-Foxn1nu immunodeficient nude mice (Charles River Laboratories, Margate, UK) aged 6-8 weeks were maintained and used under national regulatory approval and UKCCCR guidelines as before.37 Xenografts for the array studies were prepared by subcutaneous injection of 1 × 106 cells of the relevant tumor line. For the growth-rate experiments, tumors were excised from a donor animal, placed in sterile physiological saline containing antibiotics and then cut into fragments of about 2 mm3. Under brief general anesthesia by isofluorane inhalation, fragments were implanted in the left and right ventral flanks of host mice using a trocar. Once the tumors had grown to about 250 mm3, they were removed, weighed, snap frozen in liquid nitrogen and stored at -80 °C. Triplex Profiling by Quantitative Immunoblotting. Tissues and cells were extracted in buffer containing 0.5% Triton X-100 then routinely subjected to SDS-PAGE (Laemmli buffer system, reducing conditions, acrylamide ) 12.5% T, 2.7% C) followed by Coomassie Blue staining or triplex quantitative immunoblotting (ERp29/PDI/BiP) as before.23,26 Briefly, electroblots were trisected and then each section was immunoblotted in parallel with the corresponding antibody (ERp29, PDI, BiP). Antibody Journal of Proteome Research • Vol. 7, No. 8, 2008 3365

research articles

Shnyder et al.

Figure 2. Major expansion of the ER proteome in mammary and salivary carcinomas from rat. Tumor and normal-tissue extracts (Tum./Nor.) matched by histology,12 were subjected to SDSPAGE and then immunoblotted in parallel for ERp29/PDI/BiP or stained with Coomassie Blue, as indicated. This comparison of mammary and salivary tissue sets shows similar overexpression levels for ERp29, PDI, and BiP in all three tumors (all panels are from a single electroblot). Note that normal-tissue extracts were deliberately up-loaded to enable densitometric analysis, and additional loads were run to obtain the quantification data reported in the text. An apparent fragment of BiP, evident in the salivary tumor extract, was excluded from the quantification.

Figure 1. Progressive expansion of the ER proteome during lactogenesis in rat. Mammary samples were obtained from resting (V; virgin female), mature (Day 0; prelactation at parturition), lactating (Days 6, 17; lactation), and involuting (Days 2, 7; postlactation) glands as indicated. (A) Quantitative immunoblot analysis, showing parallel induction of ERp29/PDI during both the prelactation and lactation (i.e., proliferative and secretory) stages. Expression levels were normalized to the 43 kDa Coomassie-stained band (Actin; see Figure 2) to avoid interference from milk proteins, and the ERp29/PDI signals were equated at V. Data are the average of two measurements at different loads in the linear range (r2 > 0.95), and equivalent expression profiles were observed in two other sample sets (timing differences preclude data pooling). Similar profiles were obtained for BiP (see text; additional data not shown). (B) Confocal microscopy localizing ERp29 to secretory epithelial cells (arrow) in a midlactation gland. The milk-containing alveolar lumen (Alv) lacks ERp29-specific immunoreactivity. Bar ) 20 µm.

dilutions were optimized to give similar signal strengths on the reference tissue (e.g., normal mammary, Figure 2), and immunoblots were developed using avidin/biotin amplification (Vectastain ABC-peroxidase and diaminobenzidine). For densitometry, linear conditions (r2 > 0.9 for immunoblot and Coomassie) were established by varied sample loads, and expression levels were normalized based on Coomassie staining in the 30-100 kDa range except where indicated for mammary samples (i.e., Actin band at 43 kDa). Immunofluorescence Microscopy and Histology. Comparative immunofluorescence analysis of ERp29 and PDI in adjacent microarray sections was essentially as described26 except that detection was with fluorescein-avidin or TRITC-avidin (Vector Laboratories, UK) and confocal microscopy was performed using a Zeiss LSM510/Axiovert 200 M instrument. Antibodies and controls were diluted to give similar signal strengths for ERp29 and PDI on normal tissue. 23,26 Adjacent sections were stained with hematoxylin-eosin (H&E) and imaged digitally 3366

Journal of Proteome Research • Vol. 7, No. 8, 2008

(AcQuis software from Synoptics, UK). Rat tumor samples were cryosectioned, fixed with ethanol and then stained with H&E. Statistical Analysis. Immunofluorescence data from lung and colon tumors were analyzed using the chi-square (comparing levels 1-3) and Fisher’s exact (level 3 vs levels 1 + 2) tests. Pairwise comparisons of immunoblot data were made using Student’s t test (homoscedastic). All tests were two-sided.

Results ER Proteome Expands Progressively during the Proliferative and Secretory Stages of Lactation. To provide a physiological reference point for our study of cancers, we first profiled ERp29/PDI/BiP expression in the rat mammary gland where it was possible to examine epithelial proliferation and secretory stages separately. Indeed, striking parallels are recognized between the normal lactogenic cycle (i.e., epithelial proliferation, stromal invasion, secretion, apoptotic involution) and breast cancer.38 Mammary glands were obtained from before, during, and after the suckling period and then profiled by quantitative immunoblotting. As shown in Figure 1A, the specific abundances of ERp29 and PDI increased about 3-fold during the prelactation period, then nearly doubled again during lactation. After weaning, ERp29/PDI expression decreased rapidly and reached basal levels within a week, by which time the gland had undergone involution.32,38 BiP exhibited a similar pattern (3-fold induction at day 6; not shown), as expected.32,39 Confocal microscopy of lactating glands showed that, like PDI,26 ERp29 was expressed more highly in alveolar epithelial cells than in surrounding stroma, consistent with a secretory role in milk production (Figure 1B). These results implied that the ER proteome undergoes two major steps of expansion in mammary epithelial cells, in support of their proliferative and secretory activities respectively. The overall overexpression of ERp29 (5.8-fold) was considerably higher than that found previously in other physi-

Triplex Profiling Reveals Heterogeneity of the ER in Cancer

research articles

23,40–42

ological contexts, yet the secretory component was similar to that reported for enamel cells and lymphocytes (2-3-fold23,40). Consequently, these results also provided a striking new association between ERp29 and secretory protein production. ER Proteome Undergoes a Major Pathophysiological Expansion in Epithelial Tumors. Using two-dimensional electrophoresis, Bini’s group found several major reticuloplasmins consistently overexpressed in epithelial tumors (carcinomas) subject to precautions being taken to equate tumor and normal tissue samples by epithelial content.12 Their report did not include ERp29 however. As it still remains unclear whether ERp29 follows this pattern,14,15,17,29–31 we profiled ERp29/PDI/ BiP in spontaneous carcinomas using the same sampling approach.12 Mammary and salivary carcinomas were obtained from rat, together with corresponding normal tissues. As illustrated in Figure 2, ERp29 was strongly overexpressed in extracts prepared from both tumor types. Subdivision of a third mammary tumor into central, intermediate and peripheral tissue components revealed similar levels of ERp29 throughout (not shown). Densitometric analysis of the three mammary tumors revealed an average 5.2-fold ((0.6 SEM, n ) 5 specimens) overexpression of ERp29 when normalized by Coomassie staining. PDI and BiP exhibited similar profiles to ERp29 in all cases (Figure 2), consistent with overexpression values reported in the reference study (4.5 and 3.1-fold for PDI and BiP, respectively12). These results confirmed that major expansion of the ER proteome is a consistent feature of epithelial tumorigenesis12 and suggested that ERp29 is overexpressed to a similar extent as PDI and BiP. Furthermore, it was evident that the ER proteome expands to similar magnitudes during cancer and normal lactogenic activity of the mammary gland. Expansion of the ER Proteome Is Variable at the Cellular Level in Lung and Colon Tumors. Having found grossly similar expansions of the ER proteome during tumorigenesis and lactogenesis (i.e., at the tissue level, Figures 1 and 2), we next investigated whether the overexpression of reticuloplasmins is uniform at the cellular level. Immunofluorescence microscopy was used to profile ERp29/PDI in human tumors that had been staged clinically. Lung and colon carcinomas were investigated, recognizing that ERp29 and PDI are physiologically enriched in both these epithelial cell types.23,26 Lung and colon tissue microarrays, each comprising 46 tumor samples and corresponding normal tissue, were processed for immunofluorescence simultaneously to legitimize intersample comparisons (technical restrictions precluded a three-way comparison with BiP). As depicted in Figure 3, ERp29 and PDI were noticeably overexpressed in all 46 lung tumor specimens, consistent with data from other tumor origins (Figure 2).12,14 However, at the cellular level, a large variation of expression patterns was observed for both ER-marker proteins, not only across tumor stages but also between different specimens within each stage. Notwithstanding the technical limitations of cross comparison, the cellular patterns of ERp29 and PDI often diverged (Figure 3A), consistent with our observations on healthy tissues.26 The possibility of an inverse correlation between ERp29 expression and tumor prognosis was noted, with only one in the 20 mostadvanced specimens having high-level immunofluorescence versus five in the remaining 26 (Figure 3A). Further sample analysis will be required to substantiate this trend (P ) 0.1-0.2 for chi-square and Fisher’s exact tests), which was not apparent for PDI. ERp29/PDI labeling was largely restricted to epithelioid

Figure 3. Heterogeneity of ERp29 and PDI overexpression in clinical specimens of lung carcinoma. A human tissue microarray, comprising 46 tumors classified according to Tumor/Nodal (T, N) stages and with corresponding five-year survival expectations,34 was examined by confocal microscopy after immunolabeling of ERp29/ PDI as indicated. Immunofluorescence responses were designated as levels 1-3 (low, medium, high) based on the strength and amount of cellular signal.26,43 (A) Data summary depicting varied expression levels of ERp29/PDI within and between tumor stages. (B) Representative ERp29 immunofluorescence and H&E images for tumors of relatively poor, moderate, and good prognosis (T4, T2, and T1, respectively), and adjacent noncancerous lung tissue (“normal”). In all cases, the ERp29 signal corresponds with epithelial elements despite their heterogeneous morphologies. The illustrated tumors had level 2 immunofluorescence, whereas normal lung was scored 0.1), correlating with their equal growth rates in monolayer culture. Further comparative aspects of ERp29/PDI/BiP expression are discussed in the text. Illustrated data (mean ( SEM) were obtained from at least duplicate measurements of three separate samples and arbitrarily normalized to the respective values in COLO 205 xenografts (i.e., 100% expression for ERp29/PDI/BiP). Further statistical analysis appears in Figure 6B.

significant roles in tumorigenesis that differ from those established for BiP.5,44 Evidence of a Distinctive Link between ERp29 and Tumor Histogenesis. Strikingly, with all three cell lines investigated, the specific abundance of ERp29 was several-fold higher in xenograft tumors than the corresponding cell cultures (Figure 5). This observation did not hold for PDI and BiP, leading us to ask whether ERp29 is particularly important in ”threedimensional” aspects of tumorigenesis that are modeled poorly by monolayer cultures. Recent studies have shown that the complexities of three-dimensional growth (i.e., as solid tumors in vivo, or multicellular tumor spheroids in vitro) impart substantially different molecular requirements on cancer cells when compared to their growth as monolayers.45–48 However, little attention has been paid to how the ER proteome is affected by these differences in growth environment. To provide an intermediate state of complexity between tumor and monolayer, we grew HT29 colon carcinoma cells as tumor spheroids. The ER proteome was then accurately profiled in these three growth states by quantitative immunoblotting (Figure 6A). In keeping with our postulate that monolayer culture would constrain ER volume (i.e., flattened vs rounded cells), BiP levels were found somewhat increased in

Figure 6. Influence of culture mode and cell type on the ER proteome. (A) HT29 cells grown in vitro (preconfluent and confluent monolayers, tumor spheroids) and in vivo (xenograft tumors) were profiled for ERp29/PDI/BiP expression by quantitative immunoblotting as described under Figure 5. Expression levels of ERp29/PDI/BiP were at least doubled in vivo relative to in vitro, with ERp29 exhibiting a distinctively high (15-fold) increase. In contrast, relatively little difference was apparent between the three growth conditions in vitro. Illustrated data are from triplicate analysis of three separate experiments. (B) Expression ratios (xenograft vs preconfluent) for HT29 from Panel A and the three other cell lines from Figure 5, showing that ERp29 was consistently induced in xenografts unlike PDI and BiP. (C) HT29 samples from Panel A were immunoblotted in parallel for BiP and two other ERstress markers (ATF6a, XBP-1) as indicated. Expression levels of BiP, ATF6a (active 45-kDa form) and XBP-1 (only the 27-kDa unspliced variant was clearly detectable) were similar across all three growth conditions. This result dissociates the overexpression of ERp29 from a classical ER stress response, but does not preclude links with other stress mediators (e.g., PERK). Journal of Proteome Research • Vol. 7, No. 8, 2008 3369

research articles HT29 spheroids (1.7-fold vs preconfluent cells; P ) 0.1) albeit still markedly below those in the xenografts. This trend was weaker for ERp29/PDI however, implying that ER-proteome expansion was minor overall. Next we compared cell monolayers at log phase and confluence to gauge the relative demands of cell proliferation and epithelial maintenance. All three markers exhibited an insignificant increase (e1.4-fold) in confluent cells (Figure 6A). These two results suggested that factors other than cell proliferation or ER dimensionality were responsible for the relatively high expression of ERp29 in HT29 xenografts. The possibility that epithelial-stromal interactions and tumor nourishment38,49 could be influential in this regard was raised by histological analysis. Xenografts exhibited a prominent and well-vascularised stroma (Figure 4B), contrasting with only a trace of extracellular matrix in the spheroids as expected.45 An allied question concerned contributions from ER stress, and particularly the unfolded protein response.3–5 BiP, a central player in this cytoprotective action, underwent a much smaller induction than ERp29 in xenografts (Figure 6A), consistent with previous evidence dissociating ERp29 from the classical ER stress response22,24,27 (exception50). Supporting this dissociation, XBP-1 and ATF6a (two downstream effectors in the unfolded protein response pathway4,5) were each expressed similarly in xenografts, spheroids and monolayer cultures (Figure 6C). Finally, noting that ERp29 alone was highly upregulated when going from HT29 monolayer to xenograft (15fold; Figure 6A), we readdressed results obtained from the three other cell lines (cf. Figure 5). As shown by the conormalized data (Figure 6B), up-regulation in xenografts was consistently a significant feature for ERp29 but not PDI or BiP. Together, these results suggested that a distinctive requirement exists for ERp29 during the tissue-formative (histogenetic) stages of tumorigenesis. More generally, it was also evident that cancer-cell type and growth environment can have strong influences on the ER proteome.

Discussion The biomedical need for streamlined approaches to monitor proteome dynamics is growing rapidly. This study examined the ability of a knowledge-based triplex-profiling strategy (i.e., three functionally distinct chaperones, ERp29/PDI/BiP) to clarify uncertainties about how cancer affects the ER proteome. Investigating a wide range of samples at the tissue and cellular levels we obtained consistent evidence that the ER proteome undergoes a major but variable expansion in cancer. Three factors having a strong influence on the ER proteome were identified and the functionally enigmatic chaperone ERp29 was linked distinctively to histogenetic aspects of tumorigenesis. These findings justify pursuit of the ER-proteome as a medical target in cancer, validate ERp29/PDI/BiP profiling as a streamlined yet powerful measure of ER-proteome dynamics, and suggest that biomarker sets based on distinct functionalities could have broader biomedical utility. Heterogeneity of the ER Proteome in Cancer. The existing literature on ER-protein changes in cancer is generally muddled. At one extreme, numerous reports about individual ER-marker proteins give an inconsistent picture of overexpression in some cases but not others. Conversely, comprehensive proteomic ”snapshots” indicate that the ER proteome undergoes a relatively uniform expansion.5,12–17 Our wide-ranging results reconcile these two extremes by showing that the ER proteome does expand routinely, but to a variable extent depending on circumstance (Figures 2–6). Such striking heterogeneity of the 3370

Journal of Proteome Research • Vol. 7, No. 8, 2008

Shnyder et al. ER proteome has not been detailed previously, to our knowledge. However, it is unsurprising given the known microheterogeneity of cancers,49 the responsiveness of ER chaperones to multiple stimuli (protein production, redox, calcium), and thevariationsinERp29/PDI/BiPprofilesweobservedelsewhere.23,26 Through the comparison with lactogenesis, it is also apparent that the induction level of ER chaperones during tumorigenesis was in keeping with major pathophysiological demands for proliferation and secretion (Figures 1 and 2). Second, our results showed that the ER proteome can be influenced markedly by cancer-cell type, growth rate, and culture mode (i.e., xenograft vs monolayer culture; Figures 5 and 6B). It follows that many inconsistencies in the literature reflect experimental variables such as these. However, surprisingly little difference was seen between HT29 cultures grown as tumor spheroids and monolayers (Figure 6), contradicting the widespread view of spheroids being more tumor-like than monolayer cultures.45,47 Together, these findings substantiate the ER proteome as a universal medical target in cancer.3–9 Given proteomic evidence that tumorigenesis relies on multiple ER-protein activities,12,14 it is plausible other ER proteins could be targeted effectively like BiP.44 More detailed investigation into the mechanistic basis and topography of variations in the ER-proteome is now warranted, with ER-stress effects being of particular interest (Figure 6).3–5 The prospects of exploiting ER-proteome heterogeneity for diagnostic and therapeutic selectivity also appear attractive given the reproducibility and dynamic range of our immunoblotting data (Figures 5,6). However, our results also sound a warning note for such medically applied investigations. Noting the strong influence of culture mode (Figures 5,6), it is advisable to adopt in vivo analyses when querying ER-proteome responses to drugs and radiation. Presently such analyses are commonly done in vitro.8,9,48,51 Likewise, caution should be exercised in extrapolating ER-proteome findings across different cancer-cell types (Figure 6). A Key Role for ERp29 in Tumor Histogenesis? Previous studies provided inconsistent accounts about the involvement of ERp29 in cancer.14,15,17,29–31 In contrast, our results consistently implied a general role for ERp29 in epithelial cancers, with particular importance during the histogenetic stage of tumorigenesis. ERp29 was noticeably overexpressed in all epithelial cancers investigated (>114 samples from 9 tissues of origin; Figures 3–6 and data not shown), and the level of induction in carcinoma extracts matched those of PDI and BiP (Figure 2). ERp29 expression also correlated with the rates of cancer growth and lactogenesis (Figures 1 and 5), extending previous evidence for a role in the production of secretory proteins.22,23,25,52 In contrast to PDI and BiP, ERp29 was consistently expressed more highly in xenograft tumors than the corresponding monolayer cultures (Figure 6B). Our experiments dissociated this distinctive overexpression from several cell-related factors including oncogenic transformation, histotype, cell division, cell dimensionality, and classical ER stress (Figures 5 and 6). Consequently ERp29 was linked to tumor histogenesis, the stage when epithelial and mesenchymal (stromal) components interact to form tissue. One potential explanation for this association stems from the vascular nourishment of tumors, which might sustain a specific format of the ER proteome not achievable in vitro. An allied possibility is increased demand for particular secretory outputs from the epithelial cells in vivo. Besides exporting cytokines and proteases to the extracellular matrix, carcinoma cells induce numerous endomembrane proteins (e.g., cell adhesion mol-

research articles

Triplex Profiling Reveals Heterogeneity of the ER in Cancer ecules, growth factor receptors) as needed to establish the epithelial-stromal interface.38,49 Saliently, ERp29 has been implicated in membrane-protein production,43 and ERp29 expression was consistently higher in epithelial than stromal elements (Figures 1, 3, 4).26 These considerations lead us to a new hypothesis now under investigation: that ERp29 plays a key role in tumor growth and development by supporting epithelial-stromal interactions. Overall, our data suggest ERp29 could hold broad significance for cancer, both as a universal yet selective reporter on protein-production activity of the ERproteome, and as a potential clinical target in high-profile cancers including ovary,14 mammary, lung, and colon. Utility of Knowledge-Based Triplex Profiling in Biomedicine. The subproteomic-profiling approach explored in this study combines benefits from the traditional marker-protein and statistically driven proteomic approaches: (1) profiling of three biomarkers is both practicable for analyzing numerous samples rapidly and sufficient to enable multivariate analysis; (2) knowledge-based selection of biomarkers will likely enhance the biological meaningfulness of ensuing models; and (3) monitoring of multiple functional pathways can be expected to enhance discriminatory power. Indeed, our triplex profiling of ERp29/PDI/BiP has provided an informative readout of ERproteome heterogeneity in cancer, particularly when quantified stringently by immunoblotting. Together with findings from healthy tissues23,26 (Figure 1), these results support the hypothesis that quantification of three functionally distinct biomarkers (in this case ER chaperones) may provide a particularly sensitive assay of ER pathophysiology. Notably, some differences apparent with triplex analysis would have been missed in duplex analyses using the same biomarkers (Figures 5 and 6). Based on numerous inconsistencies in the literature, we anticipate the common practice of using functionally related markers (such as BiP/endoplasmin16,53,54) is a less secure approach for detecting ER heterogeneity. Being sensitive, rapid, and suitable for automation, triplex immunoprofiling offers a powerful shortcut over comprehensive proteomics (e.g., twodimensional electrophoresis and quantitative mass spectrometry). These attributes indicate that profiling of three or more functionally distinct proteins could hold broader value, both for mechanistic analyses and diagnostic applications.

Conclusions Improved understanding of ER-proteome complexity remains a challenging goal but one of major importance if this pivotal organelle is to be targeted precisely in medicine. This study makes a significant advance by using a rational subproteomic-profiling approach (i.e., three functionally distinct chaperones) to clarify existing uncertainties and reveal new aspects of ER-proteome heterogeneity in cancer. Consequently, our findings point to triplex profiling of ERp29/PDI/BiP as a streamlined yet powerful strategy for addressing ER-proteome dynamics, and justify learning more about the roles of ERp29 and other ER chaperones in cancer. More generally, it appears the idea of using biological knowledge (e.g., distinct functionalities and subcellular co-location) as a basis for selecting biomarker sets merits further attention from the proteomics field. Abbreviations: ER, endoplasmic reticulum; PDI, protein disulfide isomerase; H&E, hematoxylin-eosin.

Acknowledgment. At Otago, we thank the late Patricia Flawn for skilled technical assistance, John Schofield for the

gift of tumor specimens, and Tom Wheeler for guidance with the lactogenesis profiling. At Bradford, thanks go to Jill Seargent, Sandie Martin, Tricia Cooper, Beryl Cronin, and Roger Phillips for cancer specimens, and to Sidiq Tijani for assistance with lung tumor pathology. Other colleagues at both locations are thanked for their constructive comments and support. This work was supported by a NZ Lottery Health Research grant (to M.H.), Cancer Research UK grant C7589/A5953 (to S.S.), a Health Research Council of NZ Senior Research Fellowship (to M.H.), the Melbourne Research Unit for Facial Disorders, and the University of Melbourne (M.H., J.M.).

Note Added in Proof. While this paper was under review, Mkrtchian et al. (Mol. Carcinog. 2008 [Epub ahead of print, DOI 10.1002/mc.20444]) reported complementary evidence implicating ERp29 in mammary tumorigenesis. References (1) Zhang, Z.; Chan, D. W. Cancer proteomics: in pursuit of “true” biomarker discovery. Cancer Epidemiol. Biomarkers Prev. 2005, 14 (10), 2283–6. (2) Lai, C.; Reinders, M. J.; van’t Veer, L. J.; Wessels, L. F. A comparison of univariate and multivariate gene selection techniques for classification of cancer datasets. BMC Bioinformatics 2006, 7, 235. (3) Moenner, M.; Pluquet, O.; Bouchecareilh, M.; Chevet, E. Integrated endoplasmic reticulum stress responses in cancer. Cancer Res. 2007, 67 (22), 10631–4. (4) Ma, Y.; Hendershot, L. M. The role of the unfolded protein response in tumour development: friend or foe. Nat. Rev. Cancer 2004, 4 (12), 966–77. (5) Li, J.; Lee, A. S. Stress induction of GRP78/BiP and its role in cancer. Curr. Mol. Med. 2006, 6 (1), 45–54. (6) Denmeade, S. R.; Isaacs, J. T. The SERCA pump as a therapeutic target: making a “smart bomb” for prostate cancer. Cancer Biol. Ther. 2005, 4 (1), 14–22. (7) Sharp, S.; Workman, P. Inhibitors of the HSP90 molecular chaperone: current status. Adv. Cancer Res. 2006, 95, 323–48. (8) Tsutsumi, S.; Namba, T.; Tanaka, K. I.; Arai, Y.; Ishihara, T.; Aburaya, M.; Mima, S.; Hoshino, T.; Mizushima, T. Celecoxib upregulates endoplasmic reticulum chaperones that inhibit celecoxib-induced apoptosis in human gastric cells. Oncogene 2006, 25 (7), 1018–29. (9) Alloza, I.; Baxter, A.; Chen, Q.; Matthiesen, R.; Vandenbroeck, K. Celecoxib inhibits interleukin-12 alphabeta and beta2 folding and secretion by a novel COX2-independent mechanism involving chaperones of the endoplasmic reticulum. Mol. Pharmacol. 2006, 69 (5), 1579–87. (10) Liu, S.; Wang, H.; Yang, Z.; Kon, T.; Zhu, J.; Cao, Y.; Li, F.; Kirkpatrick, J.; Nicchitta, C. V.; Li, C. Y. Enhancement of cancer radiation therapy by use of adenovirus-mediated secretable glucose-regulated protein 94/gp96 expression. Cancer Res. 2005, 65 (20), 9126–31. (11) Ge, J.; Normant, E.; Porter, J. R.; Ali, J. A.; Dembski, M. S.; Gao, Y.; Georges, A. T.; Grenier, L.; Pak, R. H.; Patterson, J.; Sydor, J. R.; Tibbitts, T. T.; Tong, J. K.; Adams, J.; Palombella, V. J. Design, synthesis, and biological evaluation of hydroquinone derivatives of 17-amino-17-demethoxygeldanamycin as potent, water-soluble inhibitors of Hsp90. J. Med. Chem. 2006, 49 (15), 4606–15. (12) Bini, L.; Magi, B.; Marzocchi, B.; Arcuri, F.; Tripodi, S.; Cintorino, M.; Sanchez, J. C.; Frutiger, S.; Hughes, G.; Pallini, V.; Hochstrasser, D. F.; Tosi, P. Protein expression profiles in human breast ductal carcinoma and histologically normal tissue. Electrophoresis 1997, 18 (15), 2832–41. (13) Chen, G.; Gharib, T. G.; Huang, C. C.; Thomas, D. G.; Shedden, K. A.; Taylor, J. M.; Kardia, S. L.; Misek, D. E.; Giordano, T. J.; Iannettoni, M. D.; Orringer, M. B.; Hanash, S. M.; Beer, D. G. Proteomic analysis of lung adenocarcinoma: identification of a highly expressed set of proteins in tumors. Clin. Cancer Res. 2002, 8 (7), 2298–305. (14) Bengtsson, S.; Krogh, M.; Szigyarto, C. A.; Uhlen, M.; Schedvins, K.; Silfversward, C.; Linder, S.; Auer, G.; Alaiya, A.; James, P. Largescale proteomics analysis of human ovarian cancer for biomarkers. J. Proteome Res. 2007, 6 (4), 1440–50.

Journal of Proteome Research • Vol. 7, No. 8, 2008 3371

research articles (15) Myung, J. K.; Afjehi-Sadat, L.; Felizardo-Cabatic, M.; Slavc, I.; Lubec, G. Expressional patterns of chaperones in ten human tumor cell lines. Proteome Sci. 2004, 2 (1), 8. (16) Cai, J. W.; Henderson, B. W.; Shen, J. W.; Subjeck, J. R. Induction of glucose regulated proteins during growth of a murine tumor. J. Cell. Physiol. 1993, 154 (2), 229–37. (17) Lu, Z.; Hu, L.; Evers, S.; Chen, J.; Shen, Y. Differential expression profiling of human pancreatic adenocarcinoma and healthy pancreatic tissue. Proteomics 2004, 4 (12), 3975–88. (18) Ellgaard, L.; Ruddock, L. W. The human protein disulphide isomerase family: substrate interactions and functional properties. EMBO Rep. 2005, 6 (1), 28–32. (19) Gilchrist, A.; Au, C. E.; Hiding, J.; Bell, A. W.; Fernandez-Rodriguez, J.; Lesimple, S.; Nagaya, H.; Roy, L.; Gosline, S. J.; Hallett, M.; Paiement, J.; Kearney, R. E.; Nilsson, T.; Bergeron, J. J. Quantitative proteomics analysis of the secretory pathway. Cell 2006, 127 (6), 1265–81. (20) Hubbard, M. J.; McHugh, N. J. Human ERp29: isolation, primary structural characterisation and two-dimensional gel mapping. Electrophoresis 2000, 21 (17), 3785–96. (21) Hubbard, M. J. Functional proteomics: The goalposts are moving. Proteomics 2002, 2 (9), 1069–78. (22) Demmer, J.; Zhou, C.; Hubbard, M. J. Molecular cloning of ERp29, a novel and widely expressed resident of the endoplasmic reticulum. FEBS Lett. 1997, 402 (2-3), 145–50. (23) Hubbard, M. J.; McHugh, N. J.; Carne, D. L. Isolation of ERp29, a novel endoplasmic reticulum protein, from rat enamel cells evidence for a unique role in secretory-protein synthesis. Eur. J. Biochem. 2000, 267 (7), 1945–57. (24) Hubbard, M. J.; Mangum, J. E.; McHugh, N. J. Purification and biochemical characterization of native ERp29 from rat liver. Biochem. J. 2004, 383 (Pt. 3), 589–97. (25) Hermann, V. M.; Cutfield, J. F.; Hubbard, M. J. Biophysical characterization of ERp29. Evidence for a key structural role of cysteine 125. J. Biol. Chem. 2005, 280 (14), 13529–37. (26) Shnyder, S. D.; Hubbard, M. J. ERp29 is a ubiquitous resident of the endoplasmic reticulum with a distinct role in secretory protein production. J. Histochem. Cytochem. 2002, 50 (4), 557–66. (27) Sargsyan, E.; Baryshev, M.; Backlund, M.; Sharipo, A.; Mkrtchian, S. Genomic organization and promoter characterization of the gene encoding a putative endoplasmic reticulum chaperone, ERp29. Gene 2002, 285 (1-2), 127–39. (28) Mkrtchian, S.; Sandalova, T. ERp29, an unusual redox-inactive member of the thioredoxin family. Antioxid. Redox Signaling 2006, 8 (3-4), 325–37. (29) Seow, T. K.; Ong, S. E.; Liang, R. C.; Ren, E. C.; Chan, L.; Ou, K.; Chung, M. C. Two-dimensional electrophoresis map of the human hepatocellular carcinoma cell line, HCC-M, and identification of the separated proteins by mass spectrometry. Electrophoresis 2000, 21 (9), 1787–813. (30) Yoon, J. W.; Kita, Y.; Frank, D. J.; Majewski, R. R.; Konicek, B. A.; Nobrega, M. A.; Jacob, H.; Walterhouse, D.; Iannaccone, P. Gene expression profiling leads to identification of GLI1-binding elements in target genes and a role for multiple downstream pathways in GLI1-induced cell transformation. J. Biol. Chem. 2002, 277 (7), 5548–55. (31) Cheretis, C.; Dietrich, F.; Chatzistamou, I.; Politi, K.; Angelidou, E.; Kiaris, H.; Mkrtchian, S.; Koutselini, H. Expression of ERp29, an endoplasmic reticulum secretion factor in basal-cell carcinoma. Am. J. Dermatopathol. 2006, 28 (5), 410–2. (32) Ghosal, D.; Shappell, N. W.; Keenan, T. W. Endoplasmic reticulum lumenal proteins of rat mammary gland. Potential involvement in lipid droplet assembly during lactation. Biochim. Biophys. Acta 1994, 1200 (2), 175–81. (33) Seargent, J. M.; Loadman, P. M.; Martin, S. W.; Naylor, B.; Bibby, M. C.; Gill, J. H. Expression of matrix metalloproteinase-10 in human bladder transitional cell carcinoma. Urology 2005, 65 (4), 815–20. (34) Watanabe, Y. TNM classification for lung cancer. Ann. Thorac. Cardiovasc. Surg. 2003, 9 (6), 343–50.

3372

Journal of Proteome Research • Vol. 7, No. 8, 2008

Shnyder et al. (35) Wood, D. A. Clinical staging and end results classification: TNM system of clinical classification as applicable to carcinoma of the colon and rectum. Cancer 1971, 28 (1), 109–14. (36) Phillips, R. M.; de la Cruz, A.; Traver, R. D.; Gibson, N. W. Increased activity and expression of NAD(P)H:quinone acceptor oxidoreductase in confluent cell cultures and within multicellular spheroids. Cancer Res. 1994, 54 (14), 3766–71. (37) Shnyder, S. D.; Cooper, P. A.; Scally, A. J.; Bibby, M. C. Reducing the cost of screening novel agents using the hollow fibre assay. Anticancer Res. 2006, 26 (3A), 2049–52. (38) Wiseman, B. S.; Werb, Z. Stromal effects on mammary gland development and breast cancer. Science 2002, 296 (5570), 1046–9. (39) Beaton, A.; Wilkins, R. J.; Wheeler, T. T. Lactation-associated and prolactin-responsive changes in protein synthesis in mouse mammary cells. Tissue Cell 1997, 29 (5), 509–16. (40) van Anken, E.; Romijn, E. P.; Maggioni, C.; Mezghrani, A.; Sitia, R.; Braakman, I.; Heck, A. J. Sequential waves of functionally related proteins are expressed when B cells prepare for antibody secretion. Immunity 2003, 18 (2), 243–53. (41) Sargsyan, E.; Baryshev, M.; Szekely, L.; Sharipo, A.; Mkrtchian, S. Identification of ERp29, an endoplasmic reticulum lumenal protein, as a new member of the thyroglobulin folding complex. J. Biol. Chem. 2002, 277 (19), 17009–15. (42) Kwon, O. Y.; Park, S.; Lee, W.; You, K. H.; Kim, H.; Shong, M. TSH regulates a gene expression encoding ERp29, an endoplasmic reticulum stress protein, in the thyrocytes of FRTL-5 cells. FEBS Lett. 2000, 475 (1), 27–30. (43) MacLeod, J. C.; Sayer, R. J.; Lucocq, J. M.; Hubbard, M. J. ERp29, a general endoplasmic reticulum marker, is highly expressed throughout the brain. J. Comp. Neurol. 2004, 477 (1), 29–42. (44) Jamora, C.; Dennert, G.; Lee, A. S. Inhibition of tumor progression by suppression of stress protein GRP78/BiP induction in fibrosarcoma B/C10ME. Proc. Natl. Acad. Sci. U.S.A. 1996, 93 (15), 7690–4. (45) Santini, M. T.; Rainaldi, G. Three-dimensional spheroid model in tumor biology. Pathobiology 1999, 67 (3), 148–57. (46) Poland, J.; Sinha, P.; Siegert, A.; Schnolzer, M.; Korf, U.; Hauptmann, S. Comparison of protein expression profiles between monolayer and spheroid cell culture of HT-29 cells revealed fragmentation of CK18 in three-dimensional cell culture. Electrophoresis 2002, 23 (7-8), 1174–84. (47) Ghosh, S.; Spagnoli, G. C.; Martin, I.; Ploegert, S.; Demougin, P.; Heberer, M.; Reschner, A. Three-dimensional culture of melanoma cells profoundly affects gene expression profile: a high density oligonucleotide array study. J. Cell. Physiol. 2005, 204 (2), 522–31. (48) Jessani, N.; Humphrey, M.; McDonald, W. H.; Niessen, S.; Masuda, K.; Gangadharan, B.; Yates, J. R. 3rd; Mueller, B. M.; Cravatt, B. F., Carcinoma and stromal enzyme activity profiles associated with breast tumor growth in vivo. Proc. Natl. Acad. Sci. U.S.A. 2004, 101 (38), 13756–61. (49) Liotta, L. A.; Kohn, E. C. The microenvironment of the tumourhost interface. Nature 2001, 411 (6835), 375–9. (50) Park, S.; You, K. H.; Shong, M.; Goo, T. W.; Yun, E. Y.; Kang, S. W.; Kwon, O. Y. Overexpression of ERp29 in the thyrocytes of FRTL-5 cells. Mol. Biol. Rep. 2005, 32 (1), 7–13. (51) Okunaga, T.; Urata, Y.; Goto, S.; Matsuo, T.; Mizota, S.; Tsutsumi, K.; Nagata, I.; Kondo, T.; Ihara, Y. Calreticulin, a Molecular Chaperone in the Endoplasmic Reticulum, Modulates Radiosensitivity of Human Glioblastoma U251MG Cells. Cancer Res. 2006, 66 (17), 8662–71. (52) Baryshev, M.; Sargsyan, E.; Mkrtchian, S. ERp29 is an essential endoplasmic reticulum factor regulating secretion of thyroglobulin. Biochem. Biophys. Res. Commun. 2006, 340 (2), 617–24. (53) Kozutsumi, Y.; Segal, M.; Normington, K.; Gething, M. J.; Sambrook, J. The presence of malfolded proteins in the endoplasmic reticulum signals the induction of glucose-regulated proteins. Nature 1988, 332 (6163), 462–4. (54) Lee, A. S. Coordinated regulation of a set of genes by glucose and calcium ionophores in mammalian cells. Trends Biochem. Sci. 1987, 12, 20–23.

PR800126N