A Clinical Study of Hypoxia and Metallothionein Protein Expression in Squamous Cell Carcinomas

Immunochemical Reagents.

A supernatant from hybridoma clone 4.3.11.3 containing an anti-pimonidazole IgG1 monoclonal antibody was used to detect protein adducts of reductively activated pimonidazole (30) . A protein blocker, liquid 3,3′-diaminobenzidine (DAB), an IgG1 mouse monoclonal antibody (clone E9) that recognizes MT-I/II, a mouse antihuman IgG2a monoclonal antibody (clone PC10) that recognizes PCNA, peroxidase-conjugated streptavidin, and goat antimouse immunoglobulins conjugated to peroxidase-labeled dextran polymer were obtained from DAKO Corp. (Carpinteria, CA). A biotin-conjugated F(ab′)₂ fragment of a rabbit antimouse IgG was obtained from Accurate Chemical Scientific Corp. (Westbury, NY). An IgG1 mouse antihuman involucrin antibody clone SY5 was obtained from Sigma Chemical Co. (St. Louis, MO). Neutral buffered 10% formalin, Biomeda Pronase, enzyme grade polyoxyethylene (23) lauryl ether (Brij 35), Biomeda Crystal/Mount, ProbeOn Plus glass slides, and miscellaneous reagent-grade chemicals were obtained from Fisher Scientific Company (Norcross, GA). Clear-Rite 3, a nontoxic clearing agent for deparaffinization, was obtained from Richard-Allan Scientific (Kalamazoo, MI). Aqua Hematoxylin was obtained from Innovex Biosciences (Richmond, CA).

Pimonidazole Hydrochloride.

Pimonidazole hydrochloride (Hypoxyprobe-1) was obtained from NPI, Inc. (Belmont, MA) in the form of 1.0-g samples of lyophilized white powder contained in sealed vials. Ten ml of sterile 0.9% saline were added to the solid, and a volume of the solution equivalent to a calculated patient dose of 0.5 g/m² was withdrawn and added to 100 ml of sterile 0.9% saline. The diluted solution was infused i.v. over 20 min.

Squamous Cell Carcinomas.

University of North Carolina at Chapel Hill Institutional Review Board approval was obtained for the clinical use of pimonidazole hydrochloride, and protocol eligibility requirements were identical to those published previously (33) .

Tumor Biopsies.

Sixteen to 24 h after pimonidazole hydrochloride infusion, multiple biopsies were obtained from geographically distinct areas of the tumors. Biopsies were obtained from all quadrants of the tumor whenever possible, but regions of obvious necrosis were avoided. On average, four biopsies (range, one to nine) were obtained, and tissue sections were analyzed for each of the markers of interest. Biopsy selection and analyses were based on previous sampling error studies (34 , 35) . One section per biopsy was considered adequate for the purposes of the present study.

Fresh biopsy samples were placed in cold, 10% neutral buffered formalin, held at 4°C for 12–24 h, processed into paraffin blocks, and the blocks were sectioned. Four-μm-thick sections were placed on ProbeOn Plus slides. One tissue section per biopsy was stained with hematoxylin to confirm the presence of tumor. Separate sections were immunostained for pimonidazole adducts, MT-I/II, PCNA, and involucrin.

Immunohistochemistry.

For illustrative purposes, immunostaining for the different factors was performed on four contiguous sections from a single biopsy (Fig. 1⇓ ). However, for the qualitative and quantitative analyses in Tables 1⇓ and 2⇓ , two sets of two contiguous sections from each biopsy were prepared in which one section in each set was immunostained for pimonidazole adducts, and the second section in each set was immunostained for either MT-I/II or involucrin. There were 84 sets of contiguous sections for MT and hypoxia analysis and 83 evaluable sets for involucrin and hypoxia analysis.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Photomicrographs of tissue sections immunostained for pimonidazole adducts (Pimo; A), MT-I and MT-II (MT; B), PCNA (C), and involucrin (Invo; D). n, necrosis; arrow, periphery of tumor nest. ×100. Bar, 100 μm.

The immunostaining procedure for pimonidazole adducts was essentially identical to that reported previously (36) , except that tissue sections were deparaffinized with Clear-Rite 3, a nontoxic alternative to xylene, and antigen retrieval was achieved by incubating tissue sections with 0.01% Pronase for 25 min at 40°C in the case of cervix tumors and for 40 min at 40°C in the case of head and neck tumors. Immunohistochemical staining for pimonidazole adducts was achieved with a biotin-conjugated F(ab′)₂ secondary antibody reagent. The F(ab′)₂ strategy has become standard procedure in our laboratory because it provides low background and good cross species applicability.

Tumor tissue sections were immunostained for MT-I/II in a manner similar to that used for pimonidazole adducts. A 40°C Pronase antigen retrieval step was included, but the incubation time with Pronase was shortened to 20 min for both cervix and head and neck tumors. According to the manufacturer’s literature, the commercially available IgG1 mouse antihorse MT antibody (clone E9; used in 1:50 dilution) was raised against horse, self-polymerized MT-I and MT-II. The antibody binds to the 5-amino acid moiety (AcMet-Asp-Pro-Asn-Cyst-) at the end of the NH₂ terminus of the B domain of equine and human MT-I and MT-II (37) . The anti-MT antibody does not distinguish between MT-I and MT-II, and therefore, the designation MT-I/II is used to characterize immunostaining for MT in the present report. The NH₂ terminus of MT-IV (AcMet-Asp-Pro-Gly-Glu-Cyst-; Ref. 15 ) differs significantly from that of MT-I/II, and it has been reported that MT-I/II antibodies do not cross-react with MT-IV (13) . MT-III expression is restricted to brain tissue (12) and was not a concern for the present study.

Immunostaining for PCNA was performed in a manner similar to that for pimonidazole adducts, except that Pronase digestion was omitted and goat antimouse immunoglobulin conjugated to a peroxidase-labeled dextran polymer (DAKO EnVision+ reagent) was used as the secondary reagent. PCNA is a M_r 36,000 nonhistone nuclear protein, the expression of which is associated with late G₁, S, and early G₂ phases of the cell cycle. It is an auxiliary protein to DNA polymeraseγ and plays a critical role in the initiation of cell proliferation. PCNA expression has been correlated with bromodeoxyuridine uptake in human tumors, and the characteristics of PCNA immunostaining can be used to identify S-phase cells in tissue sections (33) . According to the manufacturer’s literature, the IgG2a monoclonal anti-PCNA antibody (used in a 1:100 dilution) was raised against rat PCNA made in the protein A expression vector pR1T2T.

Immunostaining for involucrin was carried out in a manner similar to that for pimonidazole adducts, except that Pronase antigen retrieval was omitted from the immunostaining procedure. Involucrin is a cytoplasmic M_r 92,000 protein that is cross-linked to other proteins by the action of transglutaminase during terminal differentiation of keratinocytes (38) . According to the manufacturer’s literature, the IgG1 mouse antihuman involucrin antibody clone SY5 (used in a 1:100 dilution) was raised against purified human involucrin.

Negative controls, in which primary antibodies were omitted from the protocols, showed no nonspecific binding attributable to secondary antibody reagents. Positive controls containing pimonidazole adducts, MT-I/II, PCNA, and involucrin showed staining patterns and intensities that were constant from staining session to staining session. Samples of normal human tongue served as a positive control for involucrin.

Qualitative Analysis.

Microregional comparisons of hypoxia with MT and involucrin expression were carried out on a section-by-section basis without respect to whether the squamous cell carcinoma occurred in the cervix or head and neck on the assumption that MT induction is independent of tumor site. Tissue sections from the biopsies, one per biopsy, were qualitatively assessed for hypoxia, MT-I/II, and involucrin expression. Each section was viewed at ×100, and individual microscopic fields (190 × 190μ m) were assessed with respect to: (a) no overlap; (b) overlap; or (c) a mix of overlap and no overlap with immunostaining for pimonidazole adducts.

Quantitative Analysis.

Quantitative image analysis was carried out for pimonidazole adducts and MT-I/II. Color detection threshold and default width settings were chosen for the 3,3′-diaminobenzidine chromogen on the basis of an immunostained region at ×200 in tissue sections. The settings for the chromogen staining were optimized for intensity, saturation, and hue so that all cells that were identified as labeled above background by visual inspection were also scored by the image analysis software. Cells immunostained above a threshold intensity were scored as labeled, with no distinction being made between light and heavy immunostaining. This approach provides the number of tumor cells that are labeled, which is of paramount importance, and obviates concerns about variations in marker binding intensity arising from individual differences in pharmacokinetics, tumor cell redox properties, and time to biopsy. It was assumed that all cells labeled with pimonidazole adducts irrespective of immunostaining intensity were at pO₂ < 10 mm Hg and were, therefore, of interest with respect to increased radiation resistance (35 , 36) . Multiple fields 1–60(1–60, depending on section size) from each section from each biopsy were captured at ×200 by means of an Axioskop 50 microscope (×10) and Fluar objective (×20; Carl Zeiss, Inc., Thornwood, NY) linked through a high resolution, three-chip video camera (Carl Zeiss, Inc.; model ZVS-3C750E) to a high resolution Sony Trinitron RGB color monitor (model PVM 1343 MD) and a 486/33 microprocessor workstation running Image-1 software (Universal Imaging Corp., West Chester, NY). Mean percentage of area immunostained in each field was calculated as a percentage of the total field area minus areas of acellularity, necrosis, and stroma.

Qualitative Analysis.

The photomicrographs in Fig. 1⇓ are representative of immunostaining patterns for pimonidazole adducts (Fig. 1A⇓ ), MT-I/II (Fig. 1B⇓ ), PCNA (Fig. 1C⇓ ), and involucrin (Fig. 1D⇓ ). A comparison of A and B shows little or no overlap between hypoxia and MT-I/II. This is most clearly seen in the upper right of the photomicrographs, where staining for pimonidazole adducts in A is in a zone separate from that for MT-I/II staining in B. The arrows in the photomicrographs identify the periphery of the tumor nest and help in comparing immunostaining patterns. A comparison of B and C in Fig. 1⇓ reveals that MT-I/II is expressed primarily in cells in the periphery of the tumor nests that also stain for S-phase PCNA.

A comparison of A and C in Fig. 1⇓ shows that there is little or no overlap between cells that are stained for S-phase PCNA and those stained for pimonidazole adducts. This is consistent with previous studies in which immunostaining for hypoxia markers and PCNA have been compared (32 , 33) . A comparison of A and D in Fig. 1⇓ shows that involucrin staining overlaps with that for pimonidazole adducts to a remarkable extent.

Of the total number of 84 sets of tissue sections used for the analysis of hypoxia and MT expression, 64 were found to contain hypoxia. Of these, 43 of 64 (67%) possessed little or no overlap between hypoxia and MT. That is, these sections had patterns like those shown in A and B of Fig. 1⇓ . Of the remaining sections, 7 of 64 (11%) possessed overlap, and 14 of 64 (22%) possessed a mixture of overlap and no overlap in the same section (Table 1)⇓ . When the analyses were performed on a tumor-by-tumor basis, a complete absence of overlap between MT and hypoxia occurred in 5 of 7 head and neck tumors and in 7 of 13 cervix tumors.

Of the total number of 83 sets of tissue sections that were available for the analysis of hypoxia and involucrin, 72 contained measurable hypoxia. Of these, 59 of 72 (82%) showed overlap between immunostaining for pimonidazole adducts and involucrin. Of the remaining sections, 10 of 72 (14%) showed a mixture of fields with overlap and no overlap and 3 of 72 (4%) showed no overlap at all (Table 1)⇓ . The lack of overlap in individual fields was attributable entirely to microregions of hypoxia that did not express involucrin. However, when all 83 sections were analyzed including those that had no detectable hypoxia, it was found that 11 of 83 (13%) sections showed involucrin expression in the absence of hypoxia, as defined by the absence of pimonidazole binding. When examined on a tumor-by-tumor basis, 4 of 7 head and neck and 7 of 13 cervix tumors tissue sections possessed involucrin and hypoxia overlap as the only patterns of immunostaining. In the remaining tumors, sections showed mixed patterns of overlap and no overlap.

Quantitative Analysis.

Patient stage and grade are summarized in Table 2⇓ . The data are organized in ascending order of tumor hypoxia. No trends between hypoxia and stage or grade were observed. Hypoxia was measurable in all except two tumors [tumors nos. 1 and 2 in Table 1⇓ (<0.1% area hypoxic)]. The range of the percentage of area hypoxia was wide for both head and neck and cervix tumors, 0.9% to 17% and <0.1 to 14%, respectively. Areas immunostained for MT-I/II were measurable in all but one tumor (tumor 10 in Table 1⇓ ) and ranged widely from 0.0 to 36.6%. No correlation between overall hypoxia and MT was discernible (Fig. 2)⇓ .

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Quantitative comparison between MT-I/II expression and pimonidazole labeling for hypoxic cells. Data are means; bars, SE. The absence of error bars indicates tumors for which fewer than three biopsies were available. The data are organized for presentation in order of increasing hypoxia, from left to right. The tumor numbers refer to tumors described in Table 2⇓ .

Pimonidazole Hydrochloride as Hypoxia Marker.

Pimonidazole hydrochloride was found to have advantages as a marker for human tumor hypoxia. The solid is stable for at least 2 years at room temperature in subdued light and has a maximum solubility of 400 mm or 116 g/100 ml of 0.9% saline. The solutions used for infusion contained 34 mm pimonidazole hydrochloride in 0.9% saline and were stable for at least 1.5 years at 4°C in subdued light, as determined by high performance liquid chromatography and UV spectroscopy. The pH of the 34 mm infusion solution was 3.9 ± 0.1. When exposed to laboratory light, solid pimonidazole hydrochloride and its solutions slowly turned yellow. At a dose of 0.5 g/m², pimonidazole hydrochloride caused neither central nervous system toxicity nor sensation (e.g., flushing) in any of the 20 patients studied. Central nervous system toxicity was of particular interest because this was the dose-limiting toxicity for pimonidazole hydrochloride at the higher, multiple doses used in radiosensitizer trials (39 , 40) . In addition to the absence of central nervous system effects, the overall procedure from pimonidazole hydrochloride infusion to tumor biopsy was well tolerated in both inpatient and outpatient settings. The 4.3.11.3 hybridoma supernatant containing the monoclonal antibody to pimonidazole adducts was stable for at least 4 months at 4°C when supplemented with 10 mg/ml of BSA and 10 mm sodium azide.

Hypoxia and MT Expression.

Hypoxic inducibility and pluripotential biological activity made metallothionein an interesting possibility as a link between hypoxia and poor prognosis in human cancer. It was known that hypoxia induces MT in human cell lines, and there was clinical data linking MT to poor prognosis. However, the present results show that MT-I/II is not overexpressed in the majority of hypoxic cells in established squamous cell carcinomas. Furthermore, there is no quantitative correlation between overall hypoxia and MT expression at the time of clinical presentation. This result was unexpected on the basis of preclinical studies, but it is consistent with clinical data for VEGF expression in squamous cell carcinomas (11) . It does not appear to be consistent for VEGF mRNA expression in gliomas, where the microregional distribution of VEGF mRNA and HIF-1 protein is generally consistent with the expected location of hypoxic cells (8, 9, 10) . Nevertheless, hypoxic cells were not positively identified in the gliomas, and it is known that necrosis can develop in glioma xenografts in the absence of microregional hypoxia (41 , 42) . Under these circumstances, necrosis might not be a reliable marker for hypoxia in gliomas. In the present study, hypoxic cells were positively identified, and it is clear that MT is not expressed in the majority of these cells in squamous cell carcinomas.

Transcriptional control for both VEGF and MT is complex and, in the case of hypoxic induction, subject to different hypoxia-sensitive transcription factors. The fact that neither VEGF nor MT is expressed in the majority of hypoxic cells in squamous cell carcinomas indicates that a generalized mechanism for the suppression of oxygen-regulated protein expression might be operating. One possibility is that hypoxic regions defined by pimonidazole binding are ischemic. Shweiki et al. (43) have shown, for example, that combined severe glucose and oxygen depletion suppresses the expression of VEGF in rat C6 glioma cells in monolayer culture. Although this mechanism cannot be ruled out, Arteel et al. (44) have shown that the binding of hypoxia markers requires energy, and it seems unlikely that pimonidazole-labeled cells are severely nutrient depleted. A second possibility is that pimonidazole binds to cysteine residues in oxygen-regulated proteins, thereby inhibiting their detection by antibodies (45) . The 20-amino acid epitope at the NH₂ terminus of VEGF is cysteine free, but the 5-amino acid epitope at the NH₂ terminus of MT-I/II contains a cysteine moiety. However, this does not appear to be a problem in practice because MT-I/II protein was easily detectable in pimonidazole-labeled cells of rodent tumors (24) with the antibody used in the present study. Consideration of other possible global mechanisms led us to reports that keratinocyte differentiation suppresses the expression of VEGF (46) and MT (15) and that squamous carcinomas express markers for terminal differentiation (47) .

During normal stratified epithelial maturation, keratinocytes move from the proliferating, basal layer into suprabasal layers, where synthesis of specific cytoskeletal proteins occurs. As keratinocyte differentiation proceeds, cells are pushed through the stratum spinosum and stratum granulosum toward the outermost stratum corneum, where they become flattened cytoskeleton-filled scales or squames. Morphological changes are accompanied by the expression of molecular markers for terminal differentiation such as involucrin, transglutaminase, and a variety of cytokeratins (48) . Details of the molecular biology of these changes in squamous cells is under active investigation (49) . Interestingly, molecular markers for terminal differentiation are also expressed in squamous cell carcinomas (Fig. 1)⇓ , and it is generally accepted that the markers are associated with differentiation processes in the tumors (50, 51, 52, 53, 54, 55, 56, 57, 58) .

The remarkable colocalization of involucrin and pimonidazole adducts raises questions of whether involucrin is an oxygen-regulated protein and whether oxygen gradients regulate differentiation. Although there are no published data on this point, there is evidence that differentiation events are initiated in the well-oxygenated basal cell layer of normal stratified epithelia and that cell migration to suprabasal layers is a consequence, and not a cause, of differentiation (48) . If this is correct, then the expression of involucrin in hypoxic cells of squamous cell carcinomas is coincidental. Whether oxygen gradients play a role in intensifying the involucrin signal seen in Fig. 1D⇓ is not known. It is generally accepted that patients with hypoxic and poorly differentiated tumors, as defined pathologically, have poorer prognoses. The tight association between hypoxia and involucrin expression would seem paradoxical in this regard. However, pathological assessment of differentiation makes use of morphological clues in tissue sections, and it is possible that tumor cell differentiation can, in many cases, proceed to a point that falls short of producing changes that are visible under the microscope (50) .

As reported by others and shown in Fig. 1⇓ , MT-I/II immunostaining occurs in proliferating microregions in the periphery of tumor nests (59) , as might be expected of a protein that is associated with cell cycle progression (26) . On the other hand, MT-I/II immunostaining is largely absent from hypoxic regions in the center of tumor nests where involucrin is strongly expressed (Fig. 1⇓ ; Refs. 47 , 54, and 55 ). Quaife et al. (15) have found that MT gene expression switches from highly inducible MT-I to noninducible MT-IV when cells transition from the basal layer to more differentiated layers in normal stratified epithelia, and it is tempting to speculate that the absence of MT-I/II protein expression in hypoxic cells in squamous cell carcinomas is attributable to the suppression of the inducible forms of MT by cell differentiation in hypoxic zones. It has been shown in studies of explanted cell lines from human cervical carcinomas that cells expressing involucrin are not necessarily undergoing a process of normal differentiation (50) , and it remains to be seen whether changes in the expression of MT isoforms are correlated with the microregional expression of involucrin in squamous cell carcinomas. In situ hybridization studies of the microregional distribution of human MT-I, MT-IIa, and MT-IV mRNA expression are under way to test this.

Relevance to Cancer Therapy.

The expectation that MT expression in the hypoxic cells of human tumors might add to the radioresistance (60, 61, 62, 63) or chemoresistance (64, 65, 66) of these cells is not supported by the present results. However, MT in proliferating compartments of tumors could account for the inverse relationship between overall MT expression and chemotherapy response (21 , 67 , 68) by both protecting cells near blood vessels and limiting drug diffusion beyond perivascular regions (69) . With respect to radiation therapy, it is interesting to note that differentiation sensitizes human tumor cells in vitro (70 , 71) . Both oxic and hypoxic cells are sensitized, and their relative radiosensitivities are, therefore, unchanged. However, if hypoxic cells in squamous cell carcinomas were more differentiated than oxic cells, then hypoxic cells might be selectively radiosensitized and their therapeutic importance diminished. The importance of hypoxic cells in squamous cell carcinomas would be further diminished if they were differentiated to the extent that they were incapable of reentering the cell cycle. Given these observations and the conventional wisdom that differentiated tumor cells belong to the cell loss compartment in tumors (72) , further investigations of the extent to which hypoxic cells in squamous cell carcinomas are differentiated appear to be warranted.

Summary.

The present results combined with earlier data for VEGF indicate that the expression of oxygen-regulated genes in human squamous cell carcinomas at the time of clinical presentation is not predicted by in vitro studies with tumor cell lines. The reason for this is not clearly understood, but a possible scenario is that differentiation of the hypoxic cells exerts collateral control on gene expression whereby oxygen-regulated protein expression is suppressed. The lack of MT expression in hypoxic tumor cells indicates that this radio- and chemoprotective protein will not contribute directly to the therapy resistance of hypoxic cells in squamous cell carcinomas.