In this study, we used a prefractionation technique, termed heparin affinity fractionation enrichment (HAFE), for preferential enrichment of the low-abundance primary human tissue proteome of surgical specimens. Conventional 2-D PAGE and direct MS/MS-based strategies have been challenged by the very large dynamic range of tissue protein concentrations, which spans approximately 105–1012-fold differences. We demonstrated that polyanionic heparin can be used to enrich tissue proteome extracts for components that occur at relatively low abundance (). On the basis of image analysis of 2-D DIGE experiments, we showed that through HAFE a richer and deeper view of the extractable proteome was achieved with spot enrichment biases of 450% and 1540% for weakly and strongly heparin-binding fractions, respectively ().
Next, we examined matched tissue pairs of human colonic adenocarcinoma and normal mucosa from patients who had undergone surgical colectomy procedures (Supplementary Table 1
). Fifty-six protein spots were found to be differentially expressed between cancer and normal tissue (>2-fold relative change in at least 3 patients and >95% statistical significance). Of these, 42 were identified in weakly heparin-binding fractions and 14 in strongly heparin-binding fractions, respectively (, Supplementary Figure 1
, and data not shown). Importantly, we demonstrated that omission of HAFE prefractionation would have failed to identify 62% (26/42) of weakly heparin-binding proteins and 43% (6/14) of strongly heparin-binding proteins. Thus, without HAFE, a total of 57% (32/56) of differentially expressed proteins would have remained undiscovered (Supplementary Figure 2
We then focused on the 32 differentially regulated proteins identified exclusively by HAFE because we hypothesized that these proteins may have escaped previous investigation that did not employ HAFE methodology. Five particular protein spots were selected based on favorable 2-D electrophoretic separation behavior () and unambiguously identified by MS/MS sequence analysis (, Supplementary Table 2
) as proteasome subunit β
type 7 (PSB7), argininosuccinate synthase (ASSY), hemoglobin α subunit (HBA), peroxiredoxin-1 (PRDX1), and signal recognition particle 9 kDa protein (SRP9). Expression of 4 proteins (PSB7, ASSY, PRDX1, and SRP9) was increased in cancer tissue relative to normal control, while 1 protein (HBA) was decreased ().
There have been few proteomic studies of differentially expressed proteins in colorectal cancer tissue.10–12
To our knowledge, however, none has used an enrichment strategy such as HAFE to augment the depth of the investigated proteome. Furthermore, none of the 5 proteins discovered in our study have been reported previously in colon cancer. This is consistent with our observation that, without HAFE, the sensitivity of unfractionated 2-D PAGE and MS/MS alone would have failed to identify the proteins as differentially expressed. Prefractionation by heparin affinity significantly enhanced visualization of the 5 proteins on the 2-D PAGE gels (). Furthermore, it appears that heparin chromatography works not only by enriching heparin-binding proteins, but also by relatively depleting abundant proteins such as albumin (). The removal of abundant proteins increases the proportion of low-abundance proteins present in the same total quantity of sample, thus, enhancing sensitivity.
It is important to note that proteomic results embody a global look at heterogeneous intact, multifaceted tumor tissue, containing, besides neoplastic epithelial cancer cells, numerous other mucosal components, such as stroma, vessels, or smooth muscle. Therefore, increased or decreased tissue proteomic levels may not necessarily indicate altered expression within the cancer cells themselves but may, alternatively, be due to altered intra- or extracellular tissue microenvironment. We purposely chose this initial global approach to biomarker discovery over, for example, a microdissection-based approach because a whole tissue-based strategy may be expected to capture both cancer cell and tissue milieu alterations (e.g., biological tumor-stroma microenvironmental changes including extracellular compositional alterations).
Along this line of thought, the observed decrease of HBA (and thus α2β2
adult hemoglobin A) in tumor tissue is likely due to a relatively lower total blood content of tumor tissue compared to normal mucosa. Even though colon carcinomas are typically thought to possess increased microvascular density with respect to normal mucosa, they also contain an overproportionally increased solid epithelial mass component. Therefore, per mass unit of tissue, cancer tissue can be expected to contain less total hemoglobin than normal mucosa, contributing to the commonly hypoxic milieu of colon carcinoma.13,14
ASSY is a homotetrameric citrulline-aspartate ligase (EC 22.214.171.124) that catalyzes the ATP-dependent synthesis of arginine via argininosuccinate from citrulline and aspartate, a critical step in the urea cycle.15
Genetic defects in ASSY cause auto-somal recessive citrullinemia. Arginine is the direct substrate for nitric oxide (NO) synthesis.15
It is possible that in a hypoxic tumor microenvironment the rate of NO synthesis (and arginine depletion) would be increased to promote compensatory local vasodilatation. Consequently, an elevated arginine demand may cause concomitant up-regulation of ASSY under such conditions. Thus, ASSY overexpression in tumor tissue may be a biomarker of local tissue hypoxia.16
Increased levels of ASSY have been observed in acute lymphocytic and myeloid leukemia cells 17,18
and ovarian surface epithelial carcinomas.19
Our subsequent work, however, focused on PSB7, PRDX1, and SRP9 because of the interesting biological functions of these proteins, the fact that these proteins had not been previously implicated in colon carcinogenesis, and the availability of reagents to validate our proteomic findings by methodologically independent approaches.
Up-regulation of PSB7, PRDX1, and SRP9 in cancer was validated by Western blotting using a tissue collection comprising 9 matched pairs of colon adenocarcinomas and a control cohort of 15 matched pairs of lung adenocarcinomas.
For colon adenocarcinomas, comparison of unfractionated and weakly (for PSB7 and PRDX1) or strongly (for SRP9) heparin-binding samples ( and Supplementary Figure 3
, left columns) showed that, while the ERs were not exactly of the same magnitudes on a patient by patient basis, they generally displayed a concordant trend. When directly comparing the data from 1-D Western blot experiments ( and Supplementary Figure 3
) to 2-D PAGE identification (), several important experimental differences need to be considered: (i
) One band in the 1-D Western represents the superposition intensity of usually several separate spots recognized by the same antibody on a 2-D gel, while only one particular spot on the 2-D PAGE gel was used for analysis (e.g., compare (spot 25) with (bands in CA4 lanes) and (4 spots each)); (ii
) band/spot intensities in the Western blot experiments were enzymatically driven and chemiluminescence-based, whereas 2-D PAGE spot intensities were based on nonenzymatic SYPRO Ruby fluorescence dye staining, likely resulting in differences in sensitivity, response linearity, and dynamic range; and (iii
) 1-D Western blot ERs were normalized relative to β
-actin as an internal control to compensate for differences in gel loading, while 2-D PAGE analysis employed spot normalization with respect to total integrated 2-D gel fluorescence intensity. Furthermore, distinct spots on a 2-D gel originating from various isoforms of the same protein could exhibit differential behavior. For example, individual 2-D Western SIRs for 4 PRDX1 isoforms in patient CA4 were 2.5, 5.2, 9.3, and 37.0, respectively (; spot pairs from left to right). The arrowheads in indicate spots corresponding to spot 25 in , which was identified as PRDX1 by MS/MS sequencing (). For comparison, the composite PRDX1 ER for patient CA4 from the corresponding 1-D Western blot experiment was 22.2 (), which falls within the range of the 2-D Western data but does not capture the differential degree by which individual isoforms varied (range, 2.5–37.0-fold). Despite these differences, both Western blotting and 2-D PAGE results provided independent experimental evidence for robust up-regulation of PSB7, PRDX1, and SRP9 proteins in colon cancer tissue.
For lung adenocarcinomas, comparison of unfractionated and weakly heparin-binding samples on a patient by patient basis (, right column) showed a predominantly concordant trend for PRDX1 (i.e., both up or both down), while there was little concordance for PSB7. Concordance for SRP9 (Supplementary Figure 3
, right column) could not be directly assessed because different sample sets had to be analyzed due to limited sample availability. PSB7 and PRDX1 ERs in the lung cancer cohort exhibited generally greater variability, and in addition, there were several patients who had decreased expression levels of PSB7 or PRDX1 in cancer tissue (ER < 1). This finding suggests that increased levels of PSB7 and PRDX1 may play a pathophysiologic role in or are a biomarker of colonic adenocarcinoma, while pulmonary adenocarcinoma, another very common human adenocarcinoma, may not share this property. 1-D Western blot analysis of SRP9 expression, in contrast, appears to suggest that this protein may also be up-regulated in pulmonary adenocarcinoma, a finding that awaits further study.
Finally, we demonstrated by immunohistochemistry on tissue sections that PSB7, PRDX1, and SRP9 protein up-regulation are localized to the neoplastic colon carcinoma cells, not other components of heterogeneous cancer tissue, such as stroma or vessels ( and and Supplementary Figure 4
). PSB7 was increased in both cytoplasm and nucleus, while PRDX1 and SRP9 up-regulation was predominantly cytoplasmic. This is consistent with cell biological observations which have shown that PSB7 and the eukaryotic proteasome as a whole are found both in the cytoplasm and the nucleus,20,21
while PRDX1 and SRP9 expression is predominantly cytoplasmic and, to a lesser degree, nuclear.22–26
Interestingly, protein overexpression was retained in lymphatically spreading tumor cells ( and Supplementary Figure 4A
) and a hepatic metastasis ( and Supplementary Figure 4D
The biological relationship of PSB7, PRDX1, or SRP9 to heparin is currently not known, except that PRDX4, a homologous PRDX family member, can bind to heparin under reducing condition.27
Several heparin-binding consensus peptide sequences have been proposed based on sequence alignments, including XBBXBX, XBBBXXBX and TXXBXXTBXXXTBB (B, basic residue; X, hydrophobic residue; T, turn).28
The protein sequences of PSB7, PRDX1, and SRP9 were searched for these consensus motifs (Supplementary Table 2
). PSB7 possesses two regions with repeated arrays of basic amino acids at 31KR
. PRDX1 has two regions at 81H
. Furthermore, SRP9 features such putative regions at 23K
. The calculated isoelectric points of PSB7, PRDX1, and SRP9 based on primary amino acid sequence are 7.58, 8.27, and 8.28, respectively, which is in keeping with their experimentally observed interaction with polyanionic heparin at physiologic pH. As would be expected, SRP9, which is the most basic of the three proteins, required the highest ionic strength for dissociation from heparin and was thus identified in the strongly heparin-binding fraction, while the other two were identified in the weakly heparin-binding fraction.
PSB7 is a component of the proteasome, a macromolecular machine integral to cellular proteolytic degradation capability.29
The eukaryotic 20S proteasome is a cylinder-shaped assembly of four stacked rings of the general form α7β7β7
The 20S particle is composed of 2 copies each of 14 different protein subunits (7 α subunits, PSA1–7; 7 β
subunits, PSB1–7). PSB7 is part of a multicatalytic endopeptidase complex (composed of PSB5, PSB6, and PSB7). In response to interferon-γ
signaling, the three subunits can be replaced by very homologous but different gene products, LMP7, LMP2, and MECL-1, respectively, forming the so-called immunoproteasome.31
At protein level, the regulation of PSB7 and MECL-1 appears to be reciprocal, that is, when PSB7 is down-regulated, MECL-1 is up-regulated and vice versa
. Interestingly, down-regulation of LMP2, LMP10, and MECL-1 has been observed in human breast, colon, and lung cancers,32–34
implying that their counterparts including PSB7 may be up-regulated compensatorily. It may be speculated that one selective advantage for cancer cells of shifting the pendulum away from the immunoproteasome is minimization of MHC class I immune recognition.35,36
The functional role of PSB7 in cancer awaits further experimental investigation.
PRDX1 belongs to a family of thiol-specific peroxidases that reduce and detoxify a wide range of organic hydroperoxides such as H2
(37). PRDX1 has been studied in cancers and overexpression has been detected in oral squamous cell, thyroid, and pulmonary carcinomas.38–43
Interestingly, transgenic mice lacking the Prdx1
gene developed several cancers at increased frequency.44
However, PRDX1 is not a conventional tumor suppressor because re-expression of peroxiredoxins in cancer cells fails to induce cell death.45
Rather, peroxiredoxins may act as both gatekeepers of oxidative damage in normal tissue and promoters of survival of proliferating malignant cells that are exposed to increased metabolic oxidative stress.41,42
PRDX3, another member of the PRDX family, has been found to protect malignant thymoma cells against hypoxia-induced H2
accumulation and apoptosis.46
The precise functional roles and mechanisms of PRDX1 up-regulation in colon cancer will require further experimental study. Interestingly, at least 3 of the characterized differentially regulated proteins, that is, HBA, ASSY, and PRDX1, share the common theme of tissue hypoxia or hypoxia-related adaptation, a signature feature of malignancy.
SRP9 is part of the signal recognition particle (SRP), a hybrid protein-RNA complex that regulates translational targeting of membrane or secretory proteins to the endoplasmic reticulum.47
The SRP can interact with the ribosome, recognize the nascent peptide chain, recruit the ribosome to the translocon, and initiate cotranslational protein sorting.48–52
A heterodimeric complex of SRP9 and SRP14 bound to 5′ and 3′ terminal sequences of SRP RNA constitutes the SRP Alu domain which stalls translation elongation until the ribosome is properly positioned on the translocon.53
As an RNA-binding protein, it may not be surprising that SRP9 was enriched in the strongly heparin-binding fractions () because polysulfated heparin may mimic the anionic charge density of RNA. A recent gene array study identified up-regulation of SRP9 mRNA in hepatocellular carcinomas.54
It is conceivable that the greater metabolic turnover in proliferating neoplastic cells necessitates increased ribosomal protein synthesis, in particular of membrane-associated proteins. Interestingly, anti-SRP autoantibodies have been detected in patients with paraneoplastic or autoimmune necrotizing myopathy.55,56
The functional relationship between the SRP and malignancy awaits further study.
In summary, we describe the discovery of PSB7, PRDX1, and SRP9 protein up-regulation in human colon adenocarcinomas. The discovery was made possible by using a prefractionation approach for enhanced analysis of the low-abundance proteome of primary human surgical tissue specimens.