Phase I Trial of Induction Histone Deacetylase and Proteasome Inhibition Followed by Surgery in Non-small Cell Lung Cancer

Abstract

Introduction

Despite complete surgical resection survival in early stage non-small cell lung cancer (NSCLC) remains poor. Based on prior pre-clinical evaluations, we hypothesized that combined induction proteasome and histone deacetylase inhibitor therapy, followed by tumor resection, is feasible.

Methods

A phase I clinical trial using a two-staged multiple agent design of bortezomib and vorinostat as induction therapy followed by consolidative surgery in patients with NSCLC was performed. Standard toxicity and MTD were examined. Pre- and post-treatment tumor gene expression arrays were performed and analyzed. Pre- and post-treatment FDG-PET imaging was used to assess tumor metabolism. Finally, serum 20S proteasome levels were analyzed with ELISA, and selected intratumoral proteins were assessed via immunohistochemistry.

Results

Thirty-four patients were consented with 21 patients enrolling in the trial. One patient withdrew early secondary to disease progression. The MTD was bortezomib 1.3 mg/m² and vorinostat 300 mg BID given. There were (2) grade III dose-limiting toxicities of fatigue and hypophosphatemia that were self-limited. There was no mortality. Thirty percent (6/20) of patients had greater than 60% histologic necrosis of their tumor following treatment, with two having ≥90% tumor necrosis. Tumor metabolism, 20S proteasome activity, and specific protein expression did not demonstrate consistent results. Gene expression arrays comparing pre- and post-therapy NSCLC specimens revealed robust intratumoral changes in specific genes.

Conclusions

Induction bortezomib and vorinostat therapy followed by surgery in patients with operable NSCLC is feasible. Correlative gene expression studies suggest new targets and cell signaling pathways that may be important in modulating this combined therapy.

Introduction

Lung cancer accounts for 13% (1.6 million) of the global total cancer cases and 18% (1.4 million) of the global cancer deaths (1). The majority of lung cancer is loco-regionally advanced or metastatic at the time of presentation and as such the overall 5-year survival for all cases remains a dismal 17%. Surprisingly, even non-small cell lung cancer (NSCLC) that is surgically removed has a five-year survival of 22–56% for larger tumors and those with nodal involvement (2, 3). Thus, there is unmet need for novel therapies that will improve survival in patients with NSCLC.

Increased mRNA and protein levels of class I and II histone deacetylases have been shown to be independent predictors of survival in patients with operable NSCLC (4). Vorinostat (suberoylanilide hydroxamic acid, SAHA) is a hydroxamic acid derivative that inhibits both class I and II histone deacetylases (HDAC) and has shown clinical benefit in hematologic malignancies such as hairy cell leukemia and mantle cell lymphoma (5). Based on these promising results we initiated experiments to determine the ability of HDAC inhibitors, such as vorinostat, to enhance apoptotic cell death in NSCLC. Unfortunately, as reported by our group, both HDAC inhibitors vorinostat and sodium butyrate demonstrated little efficacy in NSCLC cell lines, primarily secondary to an increase in the transactivation potential of the anti-apoptotic NF-κB subunit, RelA/p65 (6, 7). This process is mediated by HDAC inhibitor induced Akt-mediated phosphorylation of the transcriptional co-activator p300 (S1834) which results in increased p300 acetyltransferase activity. The transcriptional activity of RelA/p65 is then robustly increased through p300 dependent acetylation of lysine 310 on RelA/p65 which in turn drives anti-apoptotic gene (BclXL, Blf1, cIAP) transcription and promotes cancer cell survival (7).

While these in vitro experiments suggested a limited role at best for HDAC inhibitors in NSCLC, we next made the observation that vorinostat, when combined with the proteasome inhibitor bortezomib, was synergetic in causing a robust NSCLC cell death (8, 9). Further studies suggested that either the inhibition of NF-κB-dependent transcription via proteasome inhibition or the use of a dominant-negative inhibitor to IκB, the cytosolic inhibitor of NF-κB, was the primary mechanism through which the cytotoxic benefits of HDAC inhibitor therapy in NSCLC cells could be realized (6, 8). Finally, additional studies using NSCLC cell lines in a xenograft mouse model of lung cancer confirmed our in vitro observations and supported moving this combined therapeutic approach forward to clinical investigation (10).

While the safety and efficacy of bortezomib and vorinostat are well established in hematologic malignancies (11), their utility in solid tumor malignancies is not well known. In addition, there is an absence of important correlative studies, including profiling changes in intratumoral gene expression before and after combined vorinostat and bortezomib therapy. Finally, the effects of this combined therapy on specific proteins involved in HDAC and NF-κB signaling pathways, as well as on tumor metabolism, are not well described.

In this study we report the first phase I clinical trial using combined vorinostat and bortezomib as an induction strategy in surgically resected NSCLC patients. We chose this cohort of patients given their relatively poor prognosis, the ability to obtain tumor tissue for correlative studies before and after vorinostat and bortezomib therapy, and to examine this strategy of induction therapy followed by surgery as a durable platform to critically examine the effects of novel molecularly-targeted therapies on tumor biology.

The primary objective of the trial was to determine the maximum tolerated doses as well as the safety and toxicity profile of combined vorinostat and bortezomib therapy as an induction strategy in patients with operable NSCLC. The secondary aims were a) to assess changes in gene expression in the tumors before and after vorinostat and bortezomib therapy, b) to ascertain if tumor metabolism as measured by FDG-PET-CT changed following therapy, and c) to examine changes in the relative expression of specific proteins associated with vorinostat and bortezomib therapy.

Patients and Methods

Patient eligibility

This study was an open-label, phase I, single institution, investigator-initiated trial. The trial was approved by the Institutional Review Board for Health Sciences Research at the University of Virginia (UVA IRB-HSR # 12472) and was conducted in accordance with the International Conference on Harmonization (ICH) for Good Clinical Practice (GCP) regulations and the principles of the Declaration of Helsinki.

Patients were eligible provided they had biopsy-proven NSCLC and had no clinical or pathologic evidence of N2, N3, or M1 disease. In addition, the patient had to have a Zubrod performance status ≤2 and have no significant hepatic, cardiac, renal, or hematologic disease. Predicted post-operative pulmonary function had to be adequate following a complete resection of the tumor.

Study Design

A two-staged multiple agent phase I design was used to estimate the combined maximum tolerated dose (MTD) of the combination of the two agents (12). The first stage of the trial allowed for rapid escalation through treatment combinations with low adverse event rates and accrual of an additional subject to a given dose level if it was known that an inadequate biopsy sample was obtained from the prior subject treated at that dose level. Accrual to the second stage began after observation of a dose limiting toxicity (DLT) at which time the choice of treatment combination was estimated as the treatment combination with the estimated adverse event probability closest to the target probability of 33% within the constraints of the study defined stopping criteria. DLT was defined as any treatment-related grade 4 hematologic toxicity (except lymphopenia) that lasts more than 7 days, any treatment-related grade 3 or higher non-hematologic toxicity, any treatment-related febrile neutropenic event, grade 2 or higher treatment-related peripheral sensory neuropathy with pain occurring during the 21-day cycle of treatment, or any treatment delay due to toxicity lasting > 2 weeks. At study conclusion the MTD was defined as the first dose combination where a 7^th subject was recommended for treatment at that dose combination.

Study Conduct

Prior to inclusion in the study, subjects provided written, informed consent to participate in the study and underwent standardized evaluations. These included serum chemistries, CT scan, FDG-PET-CT scan, pulmonary function studies, and other studies as deemed necessary by the investigators to establish operability and confirm clinical staging. While patients may be initially enrolled in the trial without a histologic diagnosis, all patients had histologic diagnosis of NSCLC prior to cohort assignment and initiation of study therapy. Following a CT-guided fine-needle aspiration of the tumor for diagnosis, all patients had a core needle biopsy (maximum 3 passes per protocol) for tumor tissue for subsequent gene expression and protein analysis.

Patients were accrued to dose levels in cohorts of size 1–2 patients. A second patient was enrolled if the pre-treatment biopsy specimen was inadequate for at least immunohistochemical analysis or the patient experienced a DLT. Bortezomib was administered at escalating doses, with a starting dose of 1.0 mg/m² as an i.v. bolus once weekly for 3 weeks (on days 1, 8, and 15 of a 21-day cycle). Vorinostat was administered orally at escalating doses, with a starting dose of 100 mg qd 3 days per week for 3 weeks (on days 1, 2, 3, 8, 9, 10, 15, 16, and 17 of a 21-day cycle). Surgical resection of the lung cancer was performed no sooner than 7 days and no later than 14 days after completion of the last dose of combined therapy. The median time from consent to surgery was 36 days (range 26–52).

Serum was obtained prior to each dosing and immediately pre-operatively. All patients had a FDG-PET-CT scan obtained before initiation of induction therapy and within 5 days of completing their induction regimen.

Evaluation of safety

Adverse events were recorded for patients for all patients who received induction therapy. Severity was assessed according to the National Cancer Institute Common Terminology Criteria for Adverse Events (NCI CTCAE), version 3.0. In addition to vital signs, hematologic and clinical chemistry blood analyses and physical examinations were performed at each dosing and pre-operatively. Patients were monitored for safety events for 30 days following surgical resection.

The University of Virginia Cancer Center Data and Safety Monitoring Committee provided safety monitoring for this phase I clinical trial.

Evaluation of efficacy

Treatment efficacy was evaluated by a pre-operative FDG-PET-CT performed within 7 days of the last V2 treatment using the Response Evaluation Criteria in Solid Tumors (RECIST) (13). Briefly, complete response was disappearance of all lesions, partial response was 30% or more reduction in the sum of the longest diameters of the lesions, stable disease was denoted in patients whose sum of longest diameters was not decreased more than 30% and not increased more than 20%, and progressive disease was a 20% or more increase in the sum of the longest diameters of the lesions. In addition, resected NSCLC specimens were analyzed histologically for evidence of treatment effect.

Paraffin embedded tissue samples and immunohistochemistry

For the lung resection specimens, zinc formalin-fixed paraffin-embedded tissue blocks for all cases were retrieved from the archives of the UVA Pathology Department. A tissue microarray (TMA) was constructed using a TMArrayer (Pathology Devices, Westminster, MD) using sampling of each tumor with four 0.6 mm tissue cores. For the pre-therapy tumor sampling, available paraffin blocks of core needle biopsy specimens were obtained from the archives, and whole block sections were used for immunohistochemistry (IHC). IHC was performed using antibodies at dilutions and antigen retrieval methods listed in Table 1 (see Appendix) on 4-micron histologic sections of the TMA and biopsy specimens. Antigen retrieval was performed using Envision Flex Target Retrieval Solution and low pH or high pH buffers in a PT Link instrument (Dako, Carpinterio, CA). The avidin-biotin immunoperoxidase technique was used, utilizing reagents present in the Envision + Dual-Link System-HRP (DAB+) kit (Dako, Carpinterio, CA). Slides were counterstained with hematoxylin. Immunostains were scored in a semiquantitative fashion with a score assigned to staining intensity (1 = weak, 2 = moderate, 3 = strong) and a quartile score for percent of tumor cells stained (0=no staining, 1=1–25%, 2=26–50%, 3=51–75% and 4=76–100%). The two scores for each case were multiplied to arrive at an index score (range 1–12).

Table 1.

Patient demographics and tumor characteristics

Variable	N
Age (median, range)	64 (50 – 84)
Gender (M:F)	14:7
Histology
Squamous cell	11
Adenocarcinoma	8
Large Cell	2
Clinical stage
T1N1	3
T2N0	14
T2N1	2
T3N0	2
Operation
Lobectomy	18 (4 VATS*)
Pneumonectomy	2
Tumor size (cm; median, range)	3.0 (1.8 – 7.4)
Pathologic stage
T1N0	1
T1N1	2
T2N0	6
T2N1	8
T2N2	3
Adjuvant therapy
Chemotherapy	12
Radiation and chemotherapy	1
Post-operative complications (%)	10
Perioperative mortality (%)	0

^*VATS = video-assisted thoracic surgery

20S proteasome assay

Serum for 20S proteasome analysis was obtained at four time points during the study. These included prior to each induction treatment and the day of surgery. To determine 20S proteasome concentrations in the patient’s serum, we utilized the proteasome ELISA kit according to the manufacturer’s instructions (Enzo Life Sciences, Farmingdale, NY) (14).

RNA preparation for microarray experiments

Total RNA was extracted from patient tumor samples and further purified through RNeasy columns according to the manufacturer’s instructions (Qiagen, Valencia, CA). RNA was quantified by UV absorbance at 260 nm, and RNA quality was assessed using an Agilent Bioanalyzer (Palo Alto, CA) and spectrophotometric analysis. This measure of quality takes into account both the ratio of ribosomal bands 18S and 28S as well as the presence of degradation fragments.

Microarray processing and quality control

Processing of the total RNA for hybridization was performed, in general, according to the manufacturer’s instruction and their supplied reagents. Briefly, an amount of 1–2 ug of total RNA were used to generate biotin-labeled cRNA using an oligonucleotide (T7) primer in a reverse transcription reaction mixture, followed by in vitro transcription with biotin-labeled uridine triphosphate and cytosine triphosphate. cRNA (10μg) was fragmented and hybridized to a Human Genome U133_plus 2 array (Affymetrix, Santa Clara, CA), which contained >54,000 transcripts. The hybridized arrays were stained in a Fluidics Station 450 and scanned with an Affymetrix Scanner 3000 (7G). All of the array images were visually inspected for defects and quality. Signal values for each array were determined using GeneChip operating system version 1.0 software (Affymetrix, Santa Clara, CA). Even though the microarray experiments were performed in batches over a period of time when the patient tissues were collected, the pre-treatment and post-treatment pair from each patient were always processed together.

Extensive quality control analyses were performed to examine array quality. First, the images of the chips were examined for fraud or other anomaly. When passed, the image derived.cel files were analyzed using R-Bioconductor (version 12.1) and AffQCReport package to detect any arrays with excessive background, low signal intensity, or major defects, or any severe RNA degradation as revealed by the 3′/5′ intensity ratios for some control genes.

Microarray analysis and generation of differentially expressed gene lists

Probe-level raw signal values were transformed into transcript level intensity values using another Bioconductor package, RMA, or Robust Multichip Average (15). Statistically significant differential expressions between the mRNA abundance levels of the pre- and post-pairs were determined using the empirical Bayes moderated t-test in the Bioconductor version 2.4 package, linear models for microarray data (16). The top-table function with this package provided the most significant genes chosen from a parametric combination of p-values and log ratios.

Gene annotation and data mining

The transcripts identified in the microarray analysis of interest were inputted to Ingenuity Pathway Analysis (version 8.0; www.ingenuity.com) for data mining. Specifically, the gene list was queried in the knowledge-based database for enriched gene ontology and biological pathways. The significance of the findings was measured using the Fisher exact test.

Results

Patient demographics

Thirty-four patients consented for participation in the study at the University of Virginia. Eight consenting patients failed screening procedures (i.e. diagnosis of small cell carcinoma, benign disease, N2 proven disease). An additional five patients elected to withdraw from the trial prior to initial treatment but following initial consent to participate in the study. Thus, 21 patients were treated on study. Of these, one patient was withdrawn from the study after the initial treatment secondary to disease progression (developed stage IV disease). Thus, 20 patients completed the induction regimen, had a complete resection of their NSCLC, and are the focus of this report. Patient demographics and clinicopathologic characteristics are shown in Table 1.

All patients underwent surgical resection of their NSCLC within 14 days (median 11, range 7 – 13) of their last treatment. All patients with node-positive disease received adjuvant therapy within 8 weeks of their surgery. This is now considered standard of care for this group (17). There was no perioperative mortality. The 30 day hospital readmission rate was 1/20 (5%).

Toxicity

There were two dose limiting toxicities (DLT) in two separate patients. These occurred in the bortezomib (1.6 mg/m²) and vorinostat (100 mg BID) (N=1) and vorinostat (200 mg BID) (N=1) dosing schemas (Table 2). The DLTs were grade III fatigue that resulted in no V2 treatment on day 8, and grade III hypophosphatemia on day 15 that resulted in no dose or cycle change. The most common toxicities included grade I fatigue (14/20, 70%), grade I nausea (8/20, 40%), grade I neuropathy (4/20, 20%), and grade I diarrhea (4/20, 20%). Because the criteria for DLT were exceeded at the bortezomib (1.6 mg/m²) and vorinostat (200 mg BID) dosing, the maximum tolerated dose (MTD) in this study was the dosing schema consisting of bortezomib 1.3 mg/m² and vorinostat 300 mg BID.

Table 2.

Bortezomib and vorinostat dosing schema

Bortezomib(ma/m2)			Vorinostat (mg)
	Level		S1	S2	S3	S4
		Dose	100 mg QD × 3d/week	100 mg BID × 3d/week	200 mg BID × 3d/week	300 mg BID × 3d/week
	V1	1.0	X*, X	X*	X*	X*
	V2	1.3	X*, X	X*	X, X	X*,X, X, X, X, X
	V3	1.6	X*, X	X*, Y*	Y*

X = patient with no DLT; Y = patient with DLT;

^*patients with adequate pre- and post-induction therapy tissue for paired gene expression analysis

Tumor metabolism

FDG-PET-CT scans were obtained in all patients immediately prior to initiating induction therapy and immediately prior to surgery. The median time interval between FDG-PET-CT scans was 40 days (range 24 –76). The median pre-treatment SUVmax for all tumors was 10.3 (range 2.1 – 35.8). The median post-treatment SUVmax for all tumors was 12.3 (range 2.1 – 25.0). The primary tumor SUVmax decreased following induction therapy in 7 patients. The average decrease was 28% (range 10–45%), suggesting less tumor metabolic activity post-induction therapy. The remainder of patients had either no change or increased SUVmax activity between scans.

Tumor size and histopathology

The majority of patients (19/21) had stable disease and two patients had disease progression of the primary tumor as determined by RECIST criteria. Pre- and post-therapy tumor tissue was examined histologically to assess for evidence of treatment effect. In the majority of patients there was no appreciable difference. However, in eight patients there was significant evidence of treatment effect as measured by tumor necrosis and fibrosis in the surgically resected specimen (Fig. 1). This ranged from 40–95% necrosis/fibrosis in the primary tumor. In 30% of all cases (6/20) there was greater than a 60% necrosis. In these 6 patients, 2 patients had a decrease in their post-V2 therapy FDG-PET SUVmax and 4 had a slight increase (Table 3). Interestingly, in additional patient, there was complete histologic necrosis of a pre-treatment FDG-PET avid N1 lymph node.

Figure 1.

Histomicrographs (H & E) demonstrating tumor pre- and post therapy. These two patients had 90% and 95% necrosis, respectively following induction therapy.

Table 3.

Characteristics of histologic responders to bortezomib and vorinostat

Subject #	Necrosis (%)	Histology	Δ in SUVmax	Bortezomib/Vorinostat Dose*
10	95	Adenocarcinoma	decrease	V2/S1
12	90	Squamous	decrease	V3/S1
20	75	Squamous	increase	V1/S4
24	70	Squamous	increase	V3/S2
29	70	Adenocarcinoma	increase	V2/S4
34	60	Adenocarcinoma	increase	V2/S4

^*see Table 2 for key

Immunohistochemistry and 20S proteasome analysis

There were no consistent changes in protein levels of p21, CD44, BclxL, Met, RAD23b, CFLAR, RelA/p65 phospho-S⁵³⁶, or BIRC3 between pre- and post-tumor specimens. Several patients had obvious increased expression levels of these proteins while others had dramatic decreases in the same proteins. There was no correlation between changes in protein expression and tumor histology, dosing schema, or pathologic stage.

Serum 20S proteasome activity as measured by ELISA was decreased between pre- and post-treatment levels in 10/20 (50%) of patients. 20S proteasome activity levels varied from a 10% to over 600% reduction between initiation of V2 therapy and levels on the day of surgery. All robust reductions (N = 4) in proteasome activity occurred in patients who received higher dose Bortezomib (1.3 to 1.6 mg/m²) although not all patients receiving these doses had a dramatic change in 20S proteasome activity. Given the small number of patients in our study, changes in proteasome activity did not correlate with any clinicopathologic variable.

Gene expression arrays

Adequate pre-treatment tumor tissue was obtained in 11 patients. Reasons for failure to obtain adequate tissue for analysis included: insufficient quantity or quality of RNA (N = 6) and no viable tumor cells (N = 3). All patients had adequate post-resectional tumor tissue. Therefore, there were 11 paired specimens available for paired gene expression analysis. This included adequate RNA from all dosing schemas except the vorinostat 200 mg BID and the bortezomib 1.3 mg/m² cohort (see Table 2).

Following the removal of unpaired specimens and normalizing to minimize batch effects, there were 174 genes that were upregulated and 116 that were downregulated in the resection specimens compared to the pre-treatment tumor tissue (Table 4). The consistency of the significantly varied genes in our study cohort is appreciated as 92% of the upregulated genes and 84% of the downregulated genes were consistent in at least 9/11 (82%) of the paired samples. This suggests that differential gene expression profiles were similar across the study cohort.

Table 4.

Most common up- and down-regulated genes post-induction therapy

Up-regulated genes
Gene Symbol	Gene Name	Mean fold change	p-value
GAS5	growth arrest-specific 5	7.2	0.0003
RBM6	RNA binding motif protein 6	5.5	0.001
RGS1	regulator of G-protein signaling 1	5.0	0.00005
SFPQ	splicing factor proline/glutamine-rich	4.9	0.0004
CXCL2	chemokine (C-X-C motif) ligand 2	4.8	0.008
HNRNPA2B1	heterogeneous nuclear ribonucleoprotein A2/B1	4.7	0.00009
HERC2P2	hect domain-RLD 2 pseudogene 2	4.7	0.001
PILRB	paired IG-like type 2 receptor β	4.2	0.001
NCRNA00201	non-protein coding RNA 201	3.8	0.0007
RBM15	RNA binding motif protein 15	3.7	0.0006
Down-regulated genes
Gene Symbol	Gene Name	Mean fold change	p-value
HBB	hemoglobin, beta	5.3	0.0002
IGJ	immunoglobulin J polypeptide	3.9	0.002
DCN	decorin	3.7	0.0004
CYP1B1	cytochrome P450, family 1, subfamily B	3.7	0.004
MMP1	matrix metallopeptidase 1	3.3	0.003
HBA1	hemoglobin, alpha 1	3.0	0.0005
CXCL13	chemokine (C-X-C motif) ligand 13	2.9	0.01
SYNPO2	synaptopodin 2	2.8	0.02
ZEB1	zinc finger E-box binding homeobox 1	2.8	0.002
FCRL5	Fc receptor-like 5	2.7	0.01

Gene ontogeny analysis reveals that the varied genes have significant associated network functions in RNA post-transcriptional modification, immunologic and inflammatory processes, cell signaling, cellular movement and immune trafficking. With respect to biologic functions, the top diseases involved are cancer, reproductive, and skeletal diseases. Finally, the most common molecular and cellular functions for these genes involve cell death, cell signaling, and molecular transport.

Discussion

This is the first clinical trial examining the safety and toxicity of combined vorinostat and bortezomib as an induction strategy in patients with operable NSCLC. The MTD for the study was bortezomib 1.3 mg/m² given on days 1, 8, and 15 and vorinostat 300 mg BID given days 1–3, 8–10, 15–17 over a 21 day cycle. Our MTD is similar to that of Badros et al. who recently completed a phase I trial of vorinostat and bortezomib in refractory or relapsed multiple myeloma (11). In that study the MTD for vorinostat was 400 mg daily for 8 days every 21 days, and bortezomib 1.3 mg/m² every third day for a total of 4 doses. No difference in toxicity or clinical response was observed with BID compared to QD vorinostat dosing in that study.

The toxicity profile of this combined induction regimen in our study followed by consolidative surgical resection was minimal. All patients except one received full dose and full cycle of both bortezomib and vorinostat. The DLTs of grade III fatigue and hypophosphatemia in two patients only were well tolerated and self-limited. The other toxicities of nausea, gastrointestinal and neuropathy were mild and transient. Important to this specific study protocol all patients were able to undergo complete surgical resection of their NSCLC within two weeks of completing the induction regimen. There were no major perioperative complications and no mortality. Finally, given the proven survival benefit of adjuvant cisplatin-based doublet chemotherapy for node-positive NSCLC (17), all patients in whom it was indicated were able to receive this treatment without delay.

Given the relatively short duration of induction treatment, it is not surprising that there was no change in measureable tumor size as determined by the RECIST criteria in our study. Interestingly, one-third of patients had a decrease in tumor metabolic activity as measured serial FDG-PET-CT SUVmax activity. While some authors have suggested that decreases in SUVmax following conventional induction therapy with radiation and chemotherapy in NSCLC can predict operability and survival (18), this remains controversial (19) and is completely unexplored when examining more novel molecularly targeted therapies in NSCLC.

Further promising evidence of a potential benefit of induction bortezomib and vorinostat therapy was the histologic evidence of measureable tumor necrosis that was present in 6/20 (30%) of tumors. While necrosis is a relatively common histopathologic finding in NSCLC, particularly squamous cell cancer, it is uncommon to observe the extent of necrosis that we observed in these patients. In addition, 50% (3/6) of the NSCLC that demonstrated significant necrosis were adenocarcinoma. Our data does not allow us to unambiguously ascribe all of the observed tumor death to the therapy; however, it is reasonable to assert that tumors with the highest levels of necrosis are a reflection of their sensitivity to the induction regimen of bortezomib and vorinostat.

A unique aspect of this study was the ability to perform gene expression arrays on pre- and post-therapy tumors specimens to ascertain which genes were potentially affected. While several studies have examined gene expression profiles of vorinostat (20–22) and bortezomib in vitro (23), this is the first study to examine gene expression arrays following combined HDAC and proteasome inhibition therapy in a human clinical trial. Among others, we identified Decorin, MMP1, and Zeb1 to be significantly downregulated following combined treatment with bortezomib and vorinostat (see Table 4). Decorin, a prototypical small leucine rich proteoglycan involved in the extracellular matrix, is upregulated in lung cancer and has been shown to correlate with lung cancer progression in a cDNA array study of early stage NSCLC specimens (24, 25). Interestingly, Decorin is regulated by the proteasome, perhaps accounting for the significant reduction observed in our study following bortezomib and vorinostat therapy (26).

Examples of genes significantly upregulated following combined therapy include GAS5, CXCL2, and RBM6. The noncoding growth arrest specific transcript 5 gene (GAS5), encodes multiple small nucleolar RNAs, induces growth arrest and apoptosis in breast cancer cell lines, and is significantly downregulated in breast cancer (27). CXCL2 levels are typically downregulated in NSCLC compared to non-cancerous adjacent tissue (28). Finally, both RBM5 and RBM6 map to 3p21.3, a tumor suppressor region that experiences loss of heterozygosity in the majority of lung cancers (29).

This study adds to the gene expression analysis we have recently published that utilizes an in silico methodology of a refined “coexpression extrapolation (COXEN)” algorithm using a continuous spectrum of drug activity (30). Multivariate regression modeling of the refined COXEN scores in over 40 NSCLC cell lines significantly predicted the activity of combined bortezomib and vorinostat therapy. We are currently exploring the relationship between our refined COXEN algorithm and the results of the current study.

One of the unique features of this trial is the two-stage trial design with the first stage designed for rapid escalation through treatments with low toxicity. Allocation rules were specifically designed to limit the number of patients exposed to highly toxic levels of the agents and to minimize the number of patients treated below potentially effective doses. Finally, stopping rules were incorporated in the design, allowing the investigators to stop the trial early when there was sufficient evidence about the MTD. To our knowledge this is the first clinical application of this novel trial design (12). While we chose a relatively standard phase I study design, it may be that future early clinical trials involving surgery that involve molecularly targeted therapies should not be focused on safety and toxicity, instead being tasked with identification of biomarkers. This would necessitate that MTD and safety assessments be done a priori in other trials involving likely advanced stage (IIIB/IV) patients, but it would allow a more focused approach on biomarker assessment and validation.

Limitations of this study include the small number of paired NSCLC specimens secondary to a limited number of adequate pre-treatment samples. Certainly, if an adequate amount of pre-treatment tumor RNA was obtained this would have permitted a more robust paired gene expression and protein analysis. To potentially address this it may be possible to increase the number of core needle passes permitted by protocol or alternatively, subject patients to a repeat biopsy if initial assessment suggests inadequate or poor quality tissue acquisition. A second potential confounder is the varied tumor histologies and stages treated with this induction regimen. NSCLC is a heterogeneous disease and this combined therapy is likely to be most beneficial for a select subset of NSCLC.

In conclusion, an induction strategy of combined bortezomib and vorinostat followed by surgery for early stage NSCLC is safe and well tolerated. Correlative studies suggest a number of new genes and pathways that may merit further investigation regarding their ability to modulate the effects of this combined therapy in NSCLC. Finally, while not a primary or secondary goal of this phase I study, there is a group of patients with histologic evidence of tumor necrosis following their induction therapy. Based on these results we believe a phase II study is indicated to identify which patients with NSCLC should be considered for combined bortezomib and vorinostat treatment.

Appendix

Table 1.

Antibodies used for immunohistochemistry

Antigen	Antibody supplier	Catalog #	Dilution	Antigen Retrieval
p21	Santa Cruz Bio	SC-6246	1:30	High pH
CD44	Epitomics	1998–1	1:400	High pH
Bcl-xL	Cell Signaling	2764	1:400	Low pH
MET	Santa Cruz Bio	SC-8307	1:50	High pH
RAD23b	Sigma-Aldrich	HPA029718	1:50	Low pH
CFLAR	Sigma-Aldrich	HPA019044	1:250	Low pH
RelA/p65 pS536	Abcam	Ab28856	1:50	High pH
20S proteasome	Calbiochem	ST 1053	1:100	Low pH
BIRC3	Sigma-Aldrich	HPA002317	1:150	High pH