Inhibitors of the mammalian target of rapamycin (mTOR) are emerging as promising therapies for metastatic renal cell carcinoma (RCC). Because rational treatment strategies require understanding the activation status of the underlying signaling pathway being targeted at the desired stage of disease, the authors examined the activation status of different components of the mTOR pathway in RCC metastases and matched primary tumors.
The authors immunostained metastatic RCC samples from 132 patients and a subset of 25 matched primary RCCs with antibodies against phosphatidylinositol 3′-kinase, PTEN, phospho-Akt, phospho-mTOR, and p70S6. PTEN genomic status was assessed by fluorescent in situ hybridization. Marker expression was correlated to clinicopathologic variables and to survival.
The mTOR pathway showed widespread activation in RCC metastases of various sites with strong correlation between different components of this signaling cascade (P<.0001), but without significant PTEN genomic deletion. Only cytoplasmic phospho-mTOR showed independent prognostic significance (P = .029) and fidelity between primary RCCs and their matched metastases (P = .004).
Renal cell carcinoma (RCC) represents a significant cause of cancer-specific mortality, accounting for approximately 13,000 deaths in the United States for 2008.1 It is the metastatic phase of this disease that is largely incurable, with a 5-year mortality of >90%. Recently, development of antiangiogenic targeted therapies have shown promise in prolonging life expectancy in metastatic RCC. The mammalian target of rapamycin (mTOR) pathway, which regulates cell growth2 and acts as an intermediary in a variety of cell signaling events that control cellular proliferation, angiogenesis, and survival, is another signaling cascade that has been successfully targeted in this disease.
Inhibition of the mTOR pathway is a rational strategy to block the translation of cell-cycle regulatory proteins and angiogenic growth factors that contribute to neoplastic cell growth and survival. Across most tumors, activation of this pathway would be required for optimal sensitivity to mTOR inhibitors.3 Preliminary evidence for pathway activation in RCC exists in studies comparing RCC tumors to non-neoplastic kidney,4 and in RCC of the common, clear cell type.5 In primary RCC, mTOR pathway activation was shown to correlate with survival and poor pathologic prognostic features.6 At present, the mTOR inhibitors temsirolimus (CCI-779) and everolimus (RAD001) have been characterized in both preclinical models and clinical trials as anticancer agents that induce G1 growth arrest and apoptosis, with temsirolimus approved by the US Food and Drug Administration (FDA) for advanced RCC7 and everolimus FDA approved for the treatment of patients with advanced RCC after treatment failure with sunitinib or sorafenib.8
Therapeutic benefit from mTOR inhibitors has been variable among patients who have, heretofore, largely been unselected with respect to tumor-related factors and in whom treatment efficacy has been correlated with patient-related clinical factors. In a small cohort, primary RCC tumor tissue has been used as a surrogate for pathway activation in the metastatic phase of this cancer.9 The activation status of this pathway has not, however, been comprehensively studied in the metastatic RCC lesions themselves, which are the putative targets of these drugs. Our aim was thus to interrogate the activation status of the mTOR pathway, including PTEN gene status, on a large cohort of metastatic RCC samples to determine which components of this pathway are up-regulated, whether they are affected by tumor differentiation or microenvironment/site of metastasis, and whether any pathway proteins are prognostic. Moreover, we evaluated a subgroup of matched primary RCC to determine the fidelity of primary RCC sampling in predicting mTOR pathway activation in metastatic disease.
MATERIALS AND METHODS
Patient and Tumor Characteristics
Two tissue microarrays were constructed; the first array consisted of only metastatic RCC samples (n = 132) while the second array was a subset that included both primary RCC and their matched metastases (n = 25). Also, a normal tissue array was constructed using 10 non-neoplastic kidneys. The tumor and normal tissues were obtained from the McGill University Health Center and The University of Texas M. D. Anderson Cancer Center, respectively, using institutional review board-approved protocols with informed consent. An hematoxylin & eosin (H&E)-stained section of each tissue microarray was used to establish the adequacy of sampling, histologic type of RCC, and Fuhrman nuclear grade. All cases were reviewed by at least 2 pathologists, including 1 genitourinary pathologist (K.S.). Conventional, clear cell RCC was the predominant histology (126 specimens, 95.5%); other subtypes were papillary RCC in 5 specimens and 1 chromophobe RCC. Among metastatic RCC, 46.4% of specimens showed low nuclear grade (Fuhrman grades I-II), whereas 53.6% showed high nuclear grade (Fuhrman grades III-IV). The sites of metastases were as follows: adrenal (n = 11), bone (n = 24), brain (n = 21), liver (n = 6), lymph node (n = 34), lung (n = 20), gastrointestinal tract (n = 9), thyroid (n = 3), and skin (n = 4).
Clinical follow-up data was obtainable for 50 patients who underwent nephrectomy. Their clinicopathologic characteristics are summarized in Table 1. Table 2 shows characteristics of those patients who were included in the matched primary and metastatic RCC tissue array. Follow-up duration was calculated as the interval in months between initial presentation and either last visit or RCC-related death. The length of the patients' follow-up time ranged from 1 to 176 months. The tumor sizes ranged from 3 to 17 cm (mean, 9.7 cm; median, 10 cm). At the end of follow-up, 23 patients died from disease progression, with median cancer specific survival of 54 months from time of nephrectomy.
Table 1. Clinicopathologic Features of Nephrectomized Patients
Min indicates minimum; Max, maximum; IL, interleukin; IFN, interferon; NED, no evidence of disease; AWD, alive with disease; DOD, dead of disease.
Average 52.6/median 52.1
Male, n = 36
Female, n = 14
Total follow-up duration
Average 43.3/median 37
Average 10.8/median 1.5
Immunotherapy, n = 25; IL-2, n = 11; IFN, n = 14
Targeted therapy, n = 3; sorafenib, n = 1; sunitinib, n = 2
None, n = 22
NED, n = 8
AWD, n = 19
DOD, n = 23
Primary Tumor Characteristics
Mean 9.7 cm
Median 10 cm
Range 3-17 cm
Clear cell, n = 44
Papillary, n = 5
Chromophobe, n = 1
GI, n = 3; GII, n = 18
GIII, n = 24
GIV, n = 5
T1b, n = 4; T2, n = 14
T3a/b, n = 28
T4, n = 4
Nx, n = 43
N2, n = 7
M0, n = 26
M1, n = 24
Table 2. Clinicopathologic Features of Matched Primary and Metastatic Renal Cell Carcinoma Patients
Min indicates minimum; Max, maximum; IL, interleukin; IFN, interferon; NED, no evidence of disease; AWD, alive with disease; DOD, dead of disease.
Male, n = 17
Female, n = 6
Total follow-up duration, mo
Disease free interval, mo
Immunotherapy, n = 10; IL-2, n = 5; IFN, n = 5
Targeted therapy, n = 0
None, n = 13
NED, n = 6
AWD, n = 9
DOD, n = 8
Primary Tumor Characteristics
Mean 11 cm
Median 10.5 cm
Range 7-17 cm
Clear cell, n = 19
Papillary, n = 3
Chromophobe, n = 1
GII, n = 8
GIII, n = 13
GIV, n = 2
T1b, n = 1; T2: n = 7
T3a/b, n = 11
T4, n = 4
Nx, n = 16
N2, n = 7
M0, n = 10
M1, n = 13
Immunohistochemistry: Methods, Scoring, and Interpretation
Tissue microarray sections were subjected to immunohistochemistry (IHC) with antibodies against phosphatidylinositol 3′-kinase (PI3K), PTEN, phospho-Akt (p-Akt), phospho-mTOR (p-mTOR), and p70S6. Immunostaining on the tissue microarray (TMA) was done using the LSAB 2 peroxidase system from Dako Diagnostics (Carpentaria, Calif). Briefly, TMAs were deparaffinized by 2 washes, 5 minutes each, with toluene. Tissue sample rehydration was done by a series of 3 minute washes in 100%, 90%, and 80% ethanol and distilled water. To eliminate endogenous peroxidase activity, TMAs were treated with 0.3% H2O2/methanol for 30 minutes. An antigen retrieval step was performed next in boiling ethylenediaminetetraacetic acid buffer (1 mM, pH 8.0) for 15 minutes. All steps were performed at room temperature. After blocking with a protein-blocking serum-free reagent (Dako) and incubating with primary antibody for 90 minutes, 2 washes with phosphate-buffered saline (PBS), 5 minutes each, were performed before 15 minutes treatment with the secondary biotinylated antibody (Dako). After 2 PBS washes, TMAs were incubated for 15 minutes with streptavidin-peroxidase (Dako). Additional PBS washes were performed and reaction products were developed with diaminobenzidine (Dako). Nuclei were counterstained with Harris hematoxylin (Sigma-Aldrich, St Louis, Mo). Negative controls involved using PBS instead of the primary antibody.
The optimal dilution was determined for each antibody by serial dilutions. The primary antibody incubation was completed with the following antibodies: mouse monoclonal antibody PI3K-p85 at 1/20 dilution (Sc-1637, Santa Cruz Biotechnology, Santa Cruz, Calif), mouse monoclonal PTEN at 1/10 dilution (# 1809-500, Abcam, Cambridge, Mass), rabbit monoclonal p-Akt antibody at 1/25 dilution (Ser 473, IHC specific) (#3787, Cell Signaling, Danvers, Mass), rabbit monoclonal p-mTOR at 1/100 dilution (Ser 2448, IHC specific) (#2976, Cell Signaling), and rabbit monoclonal p70 S6 kinase at 1/60 dilution (#2708, Cell Signaling).
Each array was independently analyzed in a blind manner by at least 2 independent observers. Cytoplasmic expression of PI3K, PTEN, p-Akt, p-mTOR, and p70S6 kinase were scored by 2 independent observers according to the staining intensity (1-3) multiplied by the proportion of immunoreactive cells in the areas of interest (1-4).10 As such, the overall score range was from 1 to 12. Immunoreactivity scores were then dichotomized using a cutoff value of 6, with a score of ≥6 representing positive staining and <6 representing negative staining. An immunoreactivity score of 6 to 8 was counted as 1+ staining, and a score of 9 to 12 was counted as 2+ staining. Nuclear p-Akt staining was scored as the product of staining intensity (1-2) multiplied by the proportion of immunoreactive cells in the areas of interest (1-4), with overall scores ranging from 1 to 8. Immunoreactivity scores were then dichotomized using a cutoff value of 4, with a score of ≥4 representing positive staining and <4 representing negative staining.
Fluorescence In Situ Hybridization: Methods, Scoring, and Interpretation
Fluorescence in situ hybridization (FISH) was performed on a formalin-fixed, paraffin-embedded TMA containing samples of primary and matched metastatic RCC. A sequential 4-color method was used to map the PTEN locus and flanking genomic regions associated with PTEN deletions in tumors along with chromosome 10p11.1-q11.1 (SpectrumAqua CEP 10, Abbott Molecular, North Chicago, Ill). The following regions were spanned using the indicated bacterial artificial chromosome (BAC) clones: 1) 10q23.2 (88.2-88.7Mb, 900 kb upstream from the 5′ PTEN gene locus): RP11-141D8, RP11-52G13, and RP11-420K10; 2) 10q23.31 (89.6-89.7 Mb, PTEN gene locus): RP11-846G17; 3) 10q23.3 (90.6-90.8 Mb, 1 Mb downstream from the 3′PTEN gene locus): RP11-339O19 and RP11-360H20. The positions and names of the BAC clones were taken from the Human March 2008 assembly of the UCSC Genome Browser.11 BAC DNA was extracted and labeled with SpectrumGreen-dUTP, SpectrumOrange-dUTP (Abbott Molecular), or SpectrumRed-dUTP (PerkinElmer Life and Analytical Sciences, Boston, Mass) using the Vysis Nick Translation Kit (Abbott Molecular). Labeling of probes was done as described previously,12, 13 with the locus specificity of all BAC clones confirmed by both normal metaphase FISH and 4-color FISH analysis.
PTEN copy number was evaluated counting spots in 100 nonoverlapped, intact, interphase nuclei per tumor tissue core. Areas of carcinoma were identified by 4,6-diamidino-2-phenylindole dihydrochloride (DAPI; Vector Laboratories, Burlingame, Calif) staining of nuclei with reference to the corresponding H&E-stained tissue. On the basis of hybridization in 10 control cores (data not shown), hemizygous deletion of PTEN was defined as >40% (mean + 3 standard deviations in non-neoplastic controls) of tumor nuclei containing 1 PTEN locus signal and by the presence of CEP 10 signals. Homozygous deletion of PTEN was exhibited by the simultaneous lack of both PTEN locus signals and by the presence of control signals in >30% of cells.14-17
We studied the correlation of different clinical, pathological, and molecular variables using the nonparametric Spearman correlation test. For survival analysis, Kaplan-Meier (log-rank test) and univariate Cox proportional hazard tests were used. Statistical analysis was performed using the SPSS 16 statistical package (SPSS, Chicago, Ill).
Overall mTOR Pathway Expression in RCC Metastases
The mTOR pathway proteins were strongly expressed in the studied metastatic lesions, with significant up-regulation in metastatic RCC compared with normal tissue controls, as illustrated in Figure 1. Immunoreactivity scores of non-neoplastic proximal tubular epithelial cells, which are the cells of origin of clear cell RCC, were significantly lower than metastatic RCC for PI3K (P < .00001), p-mTOR (P = .0056), p-Akt (P = .10), and p70S6 (P < .00001).
The PI3K showed cytoplasmic staining in 92% (120/130) of metastatic clear cell RCCs and 80% (4/5) of metastatic papillary RCCs. PTEN showed cytoplasmic staining in 93% (114/123) of metastatic clear cell RCCs and 75% (3/4) of metastatic papillary RCCs. Activated or p-Akt staining was seen in both cytoplasmic and nuclear cellular compartments, with metastatic clear cell RCC showing cytoplasmic expression in 56% (74/130) and nuclear expression in 3% (4/130). Cytoplasmic p-Akt was expressed in 60% (3/5) of metastatic papillary RCCs, with all metastatic papillary RCCs negative for nuclear p-Akt. The p-mTOR showed cytoplasmic immunoreactivity in 66% (86/130) of metastatic clear cell RCCs and in 60% (3/5) of papillary RCC tumors. The p70S6 showed cytoplasmic staining in 84% (109/130) of metastatic clear cell RCCs and 80% (4/5) of papillary RCCs. Our sole metastatic chromophobe RCC tumor stained positively only for p70S6.
PTEN Gene Status in Primary and Metastatic RCC
FISH for PTEN was informative in 23/24 primary RCC cases and in 20/24 matched metastases. The PTEN gene showed hemizygous deletion in only 1 (4.3%) primary RCC and 1 (5.0%) metastatic RCC. In no case was there biallelic or homozygous PTEN deletion. The majority of cases showed no change in PTEN copy number in primary RCCs, 18/23 (78%), or metastatic RCCs, 15/20 (75%). Gain of genomic material at the 10q23.2-q23.31 region was noted in 2/23 (8.7%) of primary RCCs and 4/20 (20%) of corresponding metastases, with the remaining cases showing chromosomal polysomy or ploidy. Representative FISH images are shown in Figure 2.
Correlation Between mTOR Pathway Markers in RCC Metastases
The correlation of different PI3K/Akt/mTOR pathway proteins with each other is summarized in Table 3. The mean expression of cytoplasmic markers PI3K, p-mTOR, p-Akt, and p70/S6 kinase was significantly correlated to each other (P<.0001). Nuclear p-Akt correlated only with cytoplasmic p-Akt (P = .005), but not with cytoplasmic expression of PI3K, p-mTOR, and p70S6. IHC expression of PTEN did not show a significant correlation with any other pathway proteins or with PTEN gene status.
Table 3. Correlations Between mTOR Pathway Markers in Renal Cell Carcinoma Metastases
Expression of mTOR Pathway Markers Across Different Metastatic Sites and Tumor Grades
Site of metastasis is a relevant feature in RCC, as some sites appear to be more resistant to targeted therapy (eg, bone). It is therefore important to determine the presence and extent of mTOR pathway activation at specific metastatic sites. Our results, illustrated in Figure 3, show mTOR pathway markers to be commonly expressed across various metastatic sites.
Tumor differentiation in RCC is a significant prognosticator in organ-confined tumors and is generally expressed as the Fuhrman nuclear grade. We stratified nuclear grade as low grade (grades I and II) and high grade (grades III and IV), with 47% of the studied specimens classified as low grade and 53% classified as high grade. Among low-grade tumors, 94% had at least 1 component of the mTOR pathway expressed, whereas 98% of high-grade tumors had at least 1 component expressed. In general, mTOR pathway proteins were expressed across all nuclear grades among metastatic RCC without significant differences among low- or high-grade tumors for expression of any component. One exception was nuclear p-Akt, which showed a trend toward expression in low-grade compared with high-grade tumors (P = .029), although this was an infrequent occurrence, with only 4 tumors showing positive staining.
Correlation of Pathway Markers With Cancer-Specific Survival
We performed survival analysis with regard to expression of mTOR pathway markers and tumor nuclear grade, dichotomized as low grade (Fuhrman I-II) versus high grade (Fuhrman III-IV), with results shown in Table 4. Mean and median disease-free intervals were 10.8 months and 1.5 months, respectively. Patients with shorter disease-free intervals showed significantly higher expression of PI3K (P = .035).
Table 4. Univariate Proportional Hazard Analysis of mTOR Pathway Markers in Renal Cell Carcinoma Metastases With Cancer-Specific Mortality
Cytoplasmic p-mTOR expression in metastatic RCC was the only pathway marker that was significantly correlated to poorer disease-specific survival on univariate analysis. Cytoplasmic p-mTOR retained its significance on multivariate analysis even after adjustment for nuclear grade (P = .021). A survival curve for cytoplasmic p-mTOR expression is shown in Figure 4.
Concordance Between Metastatic RCC and Matched Primary RCC
We correlated overall staining of the primary RCC and their matched metastases, using our scoring scheme that incorporates both IHC staining intensity and percentage of tumor cells stained. A statistically significant correlation between the primary and metastatic RCC was only seen with cytoplasmic p-mTOR expression (P = .004). As shown in Table 5, expression of PI3K, PTEN, p-Akt, and p70S6 in primary RCC did not correlate with their matched metastases. PTEN gene status was likewise nonconcordant for gains and losses of chromosomal material at 10q23.2-q23.3.
Table 5. Correlation of mTOR Pathway Markers in Matched Primary and Metastatic Renal Cell Carcinoma
The mTOR pathway, up-regulated in many human cancers, involves downstream signaling from PI3K/Akt that leads to phosphorylation of mTOR and to activation of its substrate, p70S6 kinase (p70S6K), in turn promoting mRNA translation, cell cycle progression, and angiogenesis. Agents targeting this pathway have shown promising results in terms of disease progression and overall survival in RCC.18, 19 Because rational treatment strategies require understanding the activation status of the underlying signaling pathway being targeted at the desired stage of disease, we report herein an analysis of the mTOR pathway in RCC metastatic lesions and matched primary tumors. Our results demonstrate that the mTOR pathway is activated in metastatic lesions of RCC with significant correlation between pathway markers, underscoring the internal cohesiveness of this signaling cascade in RCC metastases. Moreover, we found activation at various nodes in this pathway, across different histologies, nuclear grades, and metastatic sites, including relatively therapy-resistant sites such as bone. The foregoing data suggests that various mTOR pathway inhibitors, including newer generation multitarget inhibitors, can be rationally applied to metastatic RCC in the hopes of bypassing the feedback activation that plagues the successful application of targeted therapy in advanced cancers.
PI3K is a lipid kinase that converts phosphatidylinositol bisphosphate to phosphatidylinositol 3,4,5-triphosphate. PI3K further recruits phosphoinositide-dependent kinase 1 and Akt to the cell membrane, where phosphoinositide-dependent kinase 1 activates Akt.20 In our study, PI3K was the most frequently expressed marker and showed correlation with shorter disease-free interval (P = .035). Inhibition at the upstream PI3K is an attractive target with the PI3K inhibitors Wortmannin and LY294002, having demonstrated marked antitumor cell activity in vitro, particularly in PTEN-null or PI3K-overexpressing renal carcinoma cells.21 PI3K inhibitors were shown to also increase radiosensitivity in prostatic and cervical cancer cell lines.22, 23 Given that PI3K is highly expressed in RCC metastases, which are themselves radioresistant, newer generation PI3K inhibitors such as PX-866, with better bioavailability and less toxicity, may show utility as radiosensitizers in RCC metastases.
Akt is stimulated by a second messenger generated from PI3K, phosphatidylinositol (3,4,5)-trisphosphate, or by decreased expression of the inhibitory PTEN.24 Akt regulates cell growth and survival mechanisms by phosphorylating a wide spectrum of cellular substrates, including mTOR, and through inactivation of the tuberous sclerosis complex.2 Recently, p-Akt expression was shown to be correlated with pathologic variables and survival, with higher levels of cytoplasmic p-Akt expression compared with nuclear p-Akt in primary RCC.6 We found cytoplasmic p-Akt to be significantly correlated to other pathway markers and to nuclear p-Akt in RCC metastases. Unlike primary RCC, p-Akt staining was not prognostic in our cohort of RCC metastases.
There have been conflicting reports in the literature regarding the genomic status of PTEN, with investigators reporting a frequency of PTEN deletions ranging between 4% and 42% in primary RCC.25-27 To our knowledge, PTEN gene status has not been reported in metastatic RCC lesions. Our data showing a ∼5% rate of monoallelic deletion in primary and metastatic RCC would suggest that PTEN deletion is not an important event in either RCC tumor initiation or in driving the progression of RCC.
mTOR is a large, polypeptide, serine/threonine-specific kinase of the PI3K-related kinase family.28 The mTOR acts downstream of PI3K/Akt and functions as an important intermediary in a variety of cell signaling events to regulate cell growth, proliferation, angiogenesis, and survival.28, 29 Its biologic importance has been demonstrated in many other tumors, including breast cancers, where tumors expressing activated mTOR had a greater risk for recurrence and poorer survival.30 In our cohort, cytoplasmic mTOR expression in RCC metastases was the only pathway marker that was found to be independently associated with poorer cancer-specific survival. Furthermore, cytoplasmic p-mTOR expression in primary RCC was correlated with that in their matched metastases, suggesting its potential use as a surrogate to predict mTOR activation status in metastatic tumors.
When mTOR is activated, it phosphorylates 2 proteins, 4E-BP1 and S6 kinase, which start the cell cycle protein translation process.28, 31 In primary RCC, p70S6 expression has been associated with T stage, nuclear grade, incidence of metastasis, and cancer-specific survival (P = .0001).6 Furthermore, p70S6 expression in primary RCC was correlated with clinical response9 in a cohort of 20 patients receiving the mTOR inhibitor temsirolimus (P = .02). We found p70S6 expression to be widely expressed and significantly correlated with cytoplasmic p-mTOR (P = .0001) in RCC metastases. There was, however, no correlation of p70S6 staining with survival and no concordance between primary and metastatic RCC.
The association of activated mTOR with survival suggests that it may be prognostic in the metastatic setting. The predictive utility of mTOR expression is not strictly known, as our cohort consisted of patients who were not treated by molecularly targeted therapy with an mTOR inhibitor. Clinical application of a biomarker on metastatic RCC tissue samples may be unrealistic, as such tissue is generally not sampled before initiation of treatment. Rather, immunostaining of primary pretreatment tumor with phospho-specific antibodies has been used as a predictive marker with limited success in RCC.9, 32 The concordance between primary and metastatic RCC has varied depending on the techniques used and the markers that were evaluated.33, 34 Our IHC and FISH results did not show correlation between metastatic and paired primary RCC, with the exception of mTOR, suggesting that the metastatic process may be associated with clonal selection leading to differential activation of this pathway that sampling of primary RCC may not accurately reflect. Such differential staining of primary and metastatic tumor pairs has also been reported to occur in lung carcinomas.35 Ideally, tumor tissues should be collected and immunostained before and after treatment to assess true response versus off-target effects. In a study of neuroendocrine carcinoma where such paired tumor tissues were collected before and after treatment with temsirolimus, it was found that higher levels of p-mTOR that predicted response to therapy.36 In this context, the fidelity between mTOR staining of primary RCCs and their matched metastases in our cohort is a promising finding. It is also notable that p-mTOR staining was not reported by Cho et al9 or Figlin et al32 in their analyses of primary RCC before treatment with the mTOR inhibitor temsirolimus.
In summary, we found activation at various nodes of the mTOR pathway, without significant PTEN genomic deletion, in metastatic RCC lesions of various histologies and nuclear grades and across different metastatic sites. The foregoing suggests the potential for vertical blockade of components upstream and downstream of mTOR even in unselected metastatic RCC patients, as they are likely to show overall pathway activation. Although the data reported herein are promising, they require external validation, ideally with tumor tissues or other surrogate biospecimens collected in a prospective clinical trial setting that includes a cohort of patients treated with mTOR pathway inhibitors.
We thank Kim-Anh Vu, Mireya Guerrero, Linda Corley, and Mannie Steglich for secretarial and technical help.
CONFLICT OF INTEREST DISCLOSURES
This work was partly supported by the Canadian Cancer Society (to J.S.) and institutional funds from the McGill University Health Center and The University of Texas M. D. Anderson Cancer Center (K.S.). Jeremy Squire is a consultant with CymoGen Dx LLB.