Tumor Hypoxia Response After Targeted Therapy in EGFR-Mutant Non-Small Cell Lung Cancer: Proof of Concept for FMISO-PET

Abstract

Hypoxia is associated with resistance to radiotherapy and chemotherapy. Functional imaging of hypoxia in non-small cell lung cancer (NSCLC) could allow early assessment of tumor response and guide subsequent therapies. Epidermal growth factor receptor (EGFR) inhibition with erlotinib reduces hypoxia in vivo. [18F]-Fluoromisonidazole (FMISO) is a radiolabeled tracer that selectively accumulates in hypoxic cells. We sought to determine whether FMISO positron emission tomography (FMISO-PET) could detect changes in hypoxia in vivo in response to EGFR-targeted therapy. In a preclinical investigation, nude mice with human EGFR-mutant lung adenocarcinoma xenografts underwent FMISO-PET scans before and 5 days after erlotinib or empty vehicle initiation. Descriptive statistics and analysis of variance (ANOVA) tests were used to analyze changes in standardized uptake value (SUV), with pooled analyses for the mice in each group (baseline, postvehicle, and posterlotinib). In a small correlative pilot human study, patients with EGFR-mutant metastatic NSCLC underwent FMISO-PET scans before and 10 to 12 days after erlotinib initiation. Changes in SUV were compared to standard chest computed tomography (CT) scans performed 6 weeks after erlotinib initiation. The mean (±standard error of the mean; SUVmean) of the xenografts was 0.17 ± 0.014, 0.14 ± 0.008, and 0.06 ± 0.004 for baseline, postvehicle, and posterlotinib groups, respectively, with lower SUVmean among the posterlotinib group compared to other groups (P < .05). Changes on preclinical PET imaging were striking, with near-complete disappearance of FMISO uptake after erlotinib initiation. Two patients were enrolled on the pilot study. In the first patient, SUVmean increased by 21% after erlotinib, with progression on 6-week chest CT followed by death after 4.8 months. In the second patient, SUVmean decreased by 7% after erlotinib, with regression on 6-week chest CT accompanied by clinical improvement; the patient had stable disease at 14.5 months. In conclusion, we observed that FMISO-PET can detect changes in hypoxia levels after EGFR-directed therapy in EGFR-mutant NSCLC. Further study is warranted to determine its utility as an imaging biomarker of early response to EGFR-directed therapy.

Introduction

Multimodality treatment of non-small cell lung cancer (NSCLC) cures relatively few patients, and NSCLC is the leading cause of death worldwide.¹ Resistance of NSCLC therapy is mediated by several factors, but there is evidence that hypoxia may play a central role. Malignant tumor progression depends heavily on neoangiogenesis, yet the new tumor vasculature is irregular and leads to areas of acidosis and hypoxia with increasing distance between tumor cells and blood vessels.^2,3 Hypoxia itself may confer adaptive survival responses, including stimulation of angiogenesis, conversion to anaerobic metabolism, and selection for genetic alterations that confer resistance to apoptosis.⁴ One promising therapeutic strategy is reversing tumor hypoxia, since hypoxic tumors are resistant to chemotherapy and radiation and carry poorer prognoses.

Epidermal growth factor receptor (EGFR) has emerged as a potent target for molecularly directed therapies in NSCLC and have been shown to reduce tumor hypoxia. Activating somatic EGFR mutations are present in about 10% of unselected US patients with NSCLC but in up to 50% of never-smoking patients with NSCLC.^5–7 Autophosphorylation of the receptors through their tyrosine kinase domains leads to recruitment of downstream effectors and activation of proliferative and cell survival signals.⁸ Small molecule, orally bioavailable tyrosine kinase inhibitors (TKIs) of the EGFR tyrosine kinase domain, such as erlotinib and gefitinib, have been associated with rapid, profound responses to these EGFR-TKIs among some patients with EGFR-mutant NSCLC,^5,9,10 and it is now a standard of care in this patient population. Inhibition of EGFR by TKIs is associated with decreased expression of downstream markers of hypoxia including hypoxia-inducible factor (HIF) 1α and vascular endothelial growth factor (VEGF) and with normalized tumor microvasculature and decreased hypoxia in vivo in mouse xenografts.^11,12 However, because 25% to 50% of patients with EGFR-mutant NSCLC do not respond clinically to EGFR-TKIs,^5,9,10,13 early identification of these patients could allow nonresponders to quickly switch to more effective or intensive therapies.

Given the role of TKIs in eliciting an EGFR-mediated decrease in tumor hypoxia, there is growing interest in the ability to visualize tumor hypoxia in vivo to assess response to TKIs and other hypoxia-reducing therapies.¹⁴ [18F]-Fluoromisonidazole (FMISO) is a radiolabeled nitroimidazole tracer, a class of compounds that selectively accumulates in hypoxic cells by forming stable adducts with intracellular macromolecules, which is prohibited at higher oxygen levels.¹⁵ The FMISO positron emission tomography (FMISO-PET) imaging has been validated against direct electrode quantification of tissue oxygenation which some consider the gold standard of in vivo hypoxia assessment and allows accurate, invasive, in vivo quantification of tumor hypoxia levels in a variety of tumors, including NSCLC and head and neck cancers.^16,17 We sought to use FMISO-PET to quantify the extent of tumor hypoxia in treatment-naive EGFR-mutant advanced NSCLC and to test the hypothesis that EGFR-targeted therapy in this population can lead to tumor hypoxia changes in vivo that are detectable on FMISO-PET.

Materials and Methods

Preclinical Protocol

A tumor model of human EGFR-mutant NSCLC was established as xenografts in nude (nu/nu) mice, as follows. EGFR-mutant (del E746_A750) human lung adenocarcinoma HCC827 cells (ATCC, Manassas, Virginia) were cultured in RPMI 1640 (ATCC) medium supplemented by 10% fetal bovine serum (Lonza BioWhittaker, Walkersville, Maryland). After reaching 80% confluence, the cells were harvested using 0.25% trypsin (Cellgro, Grand Island, New York), mixed with Matrigel (BD Biosciences [San Jose, California]), and 5 × 10⁵ cells were injected subcutaneously in the right upper flank of each Nu/Nu mice. Once tumors grew to a diameter of ≥5 mm, baseline FMISO-PET/CT scans were acquired as follows. The mice were fasted for 4 hours before image acquisition and were kept warm using a heating pad.A dose of 800 to 900 µCi of 18F-FMISO was injected intravenously via tail vein. Anesthesia was induced using 5% isoflurane and maintained with 2% isoflurane. The PET images were acquired using an Argus microPET/CT scanner (Sedecal, Madrid, Spain) for 10minutes on 1 bed position (energy window 250–700) 4 hours after FMISO administration. Images were reconstructed using 2-dimensional OSEM algorithm (2 iterations, 16 subsets) and corrected for randoms and scatter. A CT scan was acquired in standard resolution. Image data were reconstructed using FeldKemp algorithm.

Mice were assigned to either erlotinib or vehicle (dimethyl sulfoxide diluted in phosphate-buffered saline) treatment groups and received a daily dose of 50mg/kg erlotinib or vehicle, respectively, via oral route. Mice were treated for 5 consecutive days following the baseline scan before they were imaged again with FMISO-PET/CT. The FMISO-PET image analysis was done using ASIPro VM. Three-dimensional (3D) regions of interest were drawn around the tumor using CT guidance, and mean standardized uptake values (SUVmean) were measured in this area. One-way ANOVA with Tukey posttest was utilized to compare the SUVmean and SUVmax of tumors at baseline and posttreatment in vehicle and erlotinib treatment groups, respectively. Descriptive statistics are presented as mean ± standard error of the mean. Experiments were performed in triplicate such that 3 mice with established xenografts underwent baseline FMISO-PET/CT and then a follow-up FMISO-PET/CT after receiving treatment with empty vehicle, and another 3 mice with established xenografts underwent baseline FMISO-PET/CT and then a follow-up FMISO-PET/CT after receiving treatment with erlotinib.

Patient Protocol

Eligibility for the patient pilot study included pathologically confirmed diagnosis of NSCLC, presence of an EGFR-sensitizing mutation with plans to initiate erlotinib, stage IV disease, minimum lung tumor diameter of 2 cm on chest CT, Eastern Cooperative Oncology Group (ECOG) performance status 0–2, no prior EGFR-TKI treatment, and no prior radiation to the region of measurable disease. Patients unable to fast for 4 hours or with uncontrolled diabetes and/or patients with uncontrolled intercurrent illness were excluded from participation in our study. Informed consent was obtained from all patients. Demographic information of patients was recorded including gender, age at diagnosis, clinical presentation, smoking status/history, primary tumor size on CT scan, specific EGFR mutation, and metastatic sites of disease.

Study patients underwent a baseline FMISO-PET/CT scan of the thorax within 7 days before the start of therapy with erlotinib. The [18F]-FMISO-PET/CT was performed on a 64- slice PET/CT scanner, with administration of 9.5 to 10.5 mCi of [18F]-FMISO intravenously. Static emission scans were obtained centered at 2 hours after injection of radiotracer. Imaging was performed for 30 minutes (15 minutes × 2 bed positions to cover the region of interest within the thorax). A low-dose native CT scan was also performed. The PET data were coregistered with CT based on 3D coordinates of the integrated PET-CT and corrected using internal anatomical landmarks.

Patients began oral dosing of erlotinib at 150 mg daily on day 1 of treatment and continued daily dosing without interruption. The FMISO-PET/CT imaging was repeated 10 to 14 days after initiation of erlotinib therapy, using the identical steps outlined earlier.

Tracer uptake on FMISO-PET scans was assessed using SUV, normalizing the radioactivity measured in tissue by the injected dose and the body weight of the patient. Mean and maximum SUV and thresholded volume of FMISO uptake were measured to quantify the extent of hypoxia in the primary tumor. Absolute percentage change in SUVmean and SUVmax between pretreatment and posttreatment FMISO-PET scans was calculated for each patient. Patients also underwent separate diagnostic spiral chest CT scans 6 weeks after starting erlotinib.

This study was approved by the Dana-Farber/Harvard Cancer Center Institutional Review Board. Patients provided informed written consent to participate in this study, and all animal care was in accordance with institutional guidelines.

Results

Preclinical Data

In mouse xenografts of EGFR-mutant NSCLC, SUVmean and SUVmax decreased significantly on FMISO-PET scans after initiation of erlotinib but not after initiation of empty vehicle alone. The average SUVmean of the tumors were 0.17 ± 0.014, 0.14 ± 0.008, and 0.06 ± 0.004 in the baseline, postvehicle, and posterlotinib FMISO-PET scans, respectively (Figure 1A). The SUVmean was significantly lower in the posterlotinib scans compared to baseline (P = .001) and to postvehicle scans (P = .004). There was no significant difference in SUVmean between baseline and postvehicle scans (P = .17).

Figure 1.

Pooled analyses among epidermal growth factor receptor (EGFR)-mutant non-small cell lung cancer flank xenografts, with 3 mice in the baseline group, 3 mice in the vehicle group, and 3 mice in the erlotinib group. A star (⋆) indicates a statistically significant difference at the P < .05 level. A, The average standardized uptake value (SUV)mean of the tumors was 0.17±0.014, 0.14±0.008, and 0.06±0.004 in the baseline, postvehicle, and posterlotinib [18F]-fluoromisonidazole (FMISO) positron emission tomography (PET) scans, respectively. B, The average SUVmax of the tumors were 0.23 ± 0.015, 0.19 ± 0.013, and 0.09 ± 0.083 in the baseline, postvehicle, and posterlotinib FMISO-PET scans, respectively.

The average SUVmax of the tumors were 0.23 ± 0.015, 0.19 ± 0.013, and 0.09 ± 0.083 in the baseline, postvehicle, and posterlotinib FMISO-PET scans, respectively (Figure 1B). The SUVmax was significantly lower in the posterlotinib scans compared to baseline (P < .001) and to postvehicle scans (P = .003). There was no significant difference in SUVmax between baseline and postvehicle scans (P = .10).

Representative images of the mouse flank xenografts on FMISO-PET/CT scans are shown in Figure 2, according to baseline, posterlotinib, and postvehicle groups. Substantial reductions in FMISO uptake were visually apparent after initiation of erlotinib.

Figure 2.

Axial (first column), sagittal (second column), coronal (third column), and three-dimensional (3D) maximal intensity projection (MIP; fourth column) from [18F]-fluoromisonidazole (FMISO) positron emission tomography (PET)/computed tomography (CT) scans among epidermal growth factor receptor (EGFR)-mutant non-small cell lung cancer flank xenografts. Red arrows indicate primary tumor location; all xenografts were established in the same location in the right upper flank in each mouse. A, Baseline FMISO-PET/CT on a representative mouse before any treatment. B, Follow-up FMISO-PET/CT after that same mouse received erlotinib. C, Follow-up FMISO-PET/CT on another representative mouse after receiving empty vehicle alone.

Patient Data

Baseline characteristics of the 2 patients enrolled on the pilot study are shown in Table 1. Both patients were elderly women with no or remote smoking history, presenting with metastatic EGFR-mutant NSCLC. Both patients underwent baseline FMISO-PET/CT scans and began erlotinib within 7 days. Follow-up FMISO-PET/CT scans were performed 10 to 12 days after initiation of erlotinib.

Table 1.

Patient Characteristics in FMISO-PET Pilot Study.

Clinical Characteristic	Patient # 1	Patient # 2
Gender	Female	Female
Age at NSCLC diagnosis	72	81
Presentation	Cough	Bony pain
Smoking status	10 pack-years; remote	Never
Primary tumor size on chest CT	5.5 cm RUL mass	2.4 cm RUL mass
EGFR mutation	Exon 20 insertion	L858R
Metastatic sites	Pleural effusion	Bone, adrenals

Abbreviations: NSCLC, non-small cell lung cancer; CT, computed tomography; EGFR, epidermal growth factor receptor; RUL, right upper lobe; FMISO-PET, fluoromisonidazole positron emission tomography.

The FMISO-PET SUVmean had an absolute increase of 21% (1.9 to 2.3) in patient 1’s primary tumor, while SUVmean decreased by 7% (1.4 to 1.3) in patient 2’s primary tumor (Table 2 and Figure 3). The SUVmax decreased in both patients’ tumors, by 3% (3.1 to 3.0) in the first patient and 13% (1.6 to 1.4) in the second patient. Restaging diagnostic chest CT at 6 weeks revealed marked progression of the primary tumor in patient 1 plus new contralateral lung metastases, when compared to marked regression of the primary tumor in patient 2, as interpreted by the radiologist as part of routine clinical care, blinded to our study (Figure 4). At last clinical follow-up, patient 1 had died 4.8 months after initiation of erlotinib, whereas patient 2 was alive with stable disease and ongoing clinical benefit 14.5 months after initiation of erlotinib.

Table 2.

Imaging Characteristics and Clinical Follow-Up in FMISO-PET Pilot Study.

Characteristics	Patient # 1	Patient # 2
Interval from baseline FMISO-PET to erlotinib initiation, days	2	6
Interval from erlotinib initiation to follow-up FMISO-PET, days	10	12
Baseline FMISO-PET (pre-erlotinib)
SUVmax	3.1	1.6
SUVmean	1.9	1.4
Follow-up FMISO-PET (post-erlotinib initiation)
SUVmax	3.0	1.4
SUVmean	2.3	1.3
Absolute % change between FMISO-PET scans
SUVmax	3% decrease	13% decrease
SUVmean	21% increase	7% decrease
Interval from follow-up FMISO-PET to first re-staging chest CT, weeks	6.0	6.0
Restaging chest CT findingsa	Marked progression, including new contralateral metastases	Marked regression
Clinical status at last follow-up	Death from progressive intrathoracic disease, 4.8 months after erlotinib initiation	Alive with stable disease/ongoing clinical benefit on erlotinib, 14.5 months after erlotinib initiation

Abbreviations: CT, computed tomography; FMISO-PET, [18F]-fluoromisonidazole positron emission tomography; SUV, standardized uptake value.

^aPer blinded radiologist as part of routine clinical care.

Figure 3.

Axial fluoromisonidazole (FMISO) positron emission tomography (PET)/computed tomography (CT) fusion with lung windowing (first column), axial FMISO-PET (second column), and coronal FMISO-PET (third column) in patients with epidermal growth factor receptor (EGFR)-mutant non-small cell lung cancer. Red arrows indicate primary tumor location. A, Patient #1, before erlotinib. B, Patient #1, after starting erlotinib. C, Patient #2, before erlotinib. D, Patient #2, after starting erlotinib.

Figure 4.

Chest computed tomography (CT) scans with lung windowing (axial) in patients with epidermal growth factor receptor (EGFR)-mutant non-small cell lung cancer. Red arrows indicate primary tumor location. A, Patient #1, before erlotinib (top) and 1.5 months after erlotinib initiation (bottom) obtained as part of routine clinical care. This patient experienced radiographic progression and clinical decline and died from disease 4.8 months after erlotinib initiation. B, Patient #2, before erlotinib (top) and 1.5 months after erlotinib initiation (bottom) obtained as part of routine clinical care. This patient experienced radiographic regression and clinical improvement and was alive with stable disease at last follow-up 14.5 months since erlotinib initiation.

Discussion

In this investigation, we have demonstrated that molecular imaging using FMISO-PET is able to detect changes in tumor hypoxia after hypoxia-directed targeted therapy in EGFR-mutant NSCLC, both in mouse xenografts and in patients. After EGFR-directed TKI therapy, visually apparent and quantitative decreases in xenografts occurred, with both SUVmean and SUVmax being statistically significantly lower after erlotinib as opposed to empty vehicle in xenografts. In a limited correlative patient pilot study, changes in FMISO-PET SUVmean early after erlotinib therapy initiation were observable. One patient with a clinical response to erlotinib had a modest decrease in FMISO uptake, while another patient with clinical progression on erlotinib had increase in FMISO uptake. These data suggest that FMISO-PET can detect hypoxic responses to EGFR-directed therapy in mice and in humans as a proof of concept and is hypothesis generating. Whether FMISO-PET could be used as an early imaging biomarker of treatment response after hypoxia-reducing therapies in EGFR-mutant NSCLC remains to be tested in a larger cohort.

A prior feasibility study of FMISO-PET evaluated hypoxia and tumor response after nonspecific induction chemotherapy with gemcitabine and vinorelbine for locally advanced NSCLC.¹⁸ The FMISO-PET was performed within 3 days prior to initiation of induction chemotherapy and again at 14 days after completion of induction chemotherapy. After chemotherapy, all patients who had stable to decreased FMISO uptake in the primary tumor went on to have a partial or complete tumor response by completion of all treatment. The one patient who had increase in FMISO uptake in the primary tumor after induction chemotherapy was the only patient to have stable disease at the primary site after all treatment. Our small pilot investigation echoes these findings, but in the EGFR-mutant patient population with NSCLC, with EGFR-directed therapy. In another study using mouse xenografts only, the EGFR-TKI gefitinib significantly reduced uptake of the hypoxia marker [18F]-fluoroazomycin arabinoside when assessed by PET scan before and after TKI therapy, and this correlated with immunohistochemical findings of hypoxia-induced adducts in the tumors.¹⁹ A recent PET study assessed the correlation between changes in [18F]-fluorodeoxyglucose (FDG) after 1 week of erlotinib treatment and eventual response on subsequent chest CT, yet it was focused on metabolic as opposed to hypoxic changes and was conducted among patients who were mostly EGFR-wild type, representing a distinct patient population.²⁰

Tumor oxygenation responses to EGFR-TKI therapy appear to be mediated via effects on the hypoxia-mediated transcription factor HIF-1α and downstream expression of VEGF. Overexpression of EGFR in tumor cells can increase expression of VEGF directly as well as via the PI3K/Akt pathway through regulation of HIF-1α.²¹ Inhibition of EGFR by TKIs has been shown to decrease expression of HIF-1α and VEGF in vitro.^11,12,22,23 In the in vivo setting, erlotinib given to mice bearing NSCLC xenografts resulted in almost completely abolished HIF-1α and phosphorylated Akt levels on Western blotting of excised tumor lysates and a 50% reduction in expression of VEGF.²⁴ In that study, immunohistochemical staining for EF5, a 2-nitroimidazole similar to FMISO, showed significantly less staining 5 days after erlotinib treatment, consistent with decreased hypoxia. In addition, EGFR-TKI therapy led to normalized vasculature by altering blood vessel morphology and decreasing vessel permeability and increased tumor blood flow. Moreover, lower HIF1αexpression has been associated with improved response to platinum chemotherapy and overall survival in multiple reports.^25,26 Therefore, EGFR inhibition with agents such as erlotinib may improve NSCLC outcomes directly through hypoxia pathways. Whether hypoxia levels decrease after EGFR-TKI therapy through direct or indirect mechanisms and the ability to detect these changing levels and predict tumor response substantially earlier than conventional imaging would serve a substantial unmet clinical need, as there are no commonly used biomarkers of early response in NSCLC.

Although there have been significant advances in understanding of and therapies for NSCLC, 5-year overall survival rates for all patients with lung cancer remains at 15%.²⁷ Given that the subset of patients having NSCLC with activating somatic mutations in EGFR often have dramatic responses to EGFR-TKIs and associated improved survival, it is important to not only accurately identify these patients for receipt of EGFR-TKIs but also to switch them expediently to second-line therapies if they fall within the one-quarter to one-half of patients who unfortunately do not have robust responses to EGFR-TKIs.^5,9,10,13 Our pilot data are hypothesis generating, and ongoing study is warranted to explore whether pretreatment FMISO-PET followed by FMISO-PET early after EGFR-TKI initiation might serve as an imaging biomarker of response that could help identify responders at an earlier time point compared with anatomic CT imaging alone. In this investigation, one patient experienced an increase in SUVmean after erlotinib and went on to have clinical decline and death in less than 5 months, in contrast to a second patient who had decrease in SUVmean after erlotinib initiation, who was still doing well clinically with stable disease well over 1 year after starting erlotinib. Changes in FMISO-PET uptake were observable after only 10 days of erlotinib, whereas standard follow-up chest CT scans did not show responses until 1.5 months after therapy for both patients had started. Knowledge of hypoxic responses to therapy could theoretically prove clinically valuable, as patients showing no reduction in hypoxia levels after EGFR-directed therapy could be switched to second-line or more intensive treatment; similarly, the timing of chest radiotherapy among patients with NSCLC could be timed more optimally since a less hypoxic tumor may be more radiosensitive.²⁸ Further investigation is required to determine the value of FMISO-PET in this role.

Limitations of our pilot study include uncertainty regarding the optimal timing of performing FMISO-PET scans after EGFR-TKI initiation, in order to best capture hypoxic response within the tumor while still remaining an early imaging biomarker shortly after initiation of treatment. A prior FMISOPET study in NSCLC-gauging treatment response used a time point of 14 days after chemotherapy completion,²⁰ but we were interested in an earlier indication of treatment response and saw visually striking disappearance of FMISO uptake in our preclinical model with xenografts only 5 days after erlotinib initiation. This echoes the metabolic changes seen on FDGPET after only 7 days of erlotinib.²⁰ In addition, we did not observe substantial differences in SUVmax for patients before and after treatment initiation as we did for the xenografts. This may be due to maximal values having less relevance than SUVmean in reflecting the overall state of hypoxia within the tumor but requires an expanded patient cohort to further evaluate our preliminary findings, similar to other proof-of-concept pilot imaging studies.^18,29,30 In addition, although it is possible that the modest differences in SUVmax we observed in the human data only reflect inherent variability between FMISOPET scans, 2 previous studies on FMISO-PET scans in patients with head and neck cancer conducted 2 or 3 days apart showed high reproducibility of hypoxia levels on FMISO-PET within the same patient on different days,^31,32 although the precise spatial distribution within the tumor can vary.³⁰

Conclusion

In summary, we have demonstrated that FMISO-PET is able to detect changes in tumor hypoxia after hypoxia-directed targeted therapy in EGFR-mutant NSCLC, both preclinically and clinically. Ongoing study of FMISO-PET is warranted to evaluate its potential role as an early biomarker of EGFR-directed treatment response in NSCLC, which remains a widespread and challenging disease.