Intratumoral drug-releasing microdevices allow in situ high-throughput pharmaco phenotyping in patients with gliomas (original) (raw)

• Journal List
• HHS Author Manuscripts
• PMC10754230

Sci Transl Med. Author manuscript; available in PMC 2024 Mar 6.

Published in final edited form as:

PMCID: PMC10754230

NIHMSID: NIHMS1945207

Pierpaolo Peruzzi,1,*† Christine Dominas,2 Geoffrey Fell,3 Joshua D. Bernstock,1 Sarah Blitz,4 Debora Mazzetti,1 Mykola Zdioruk,1 Hassan Y. Dawood,1 Daniel V. Triggs,1 Sebastian W. Ahn,2 Sharath K. Bhagavatula,2 Shawn M. Davidson,5 Zuzana Tatarova,2 Michael Pannell,1 Kyla Truman,1 Anna Ball,1 Maxwell P. Gold,6 Veronika Pister,6 Ernest Fraenkel,6,7 E. Antonio Chiocca,1 Keith L. Ligon,8 Patrick Y. Wen,9 and Oliver Jonas2,†*

Pierpaolo Peruzzi

1Department of Neurosurgery, Brigham and Women’s Hospital, 60 Fenwood Road, Boston, MA 02115, USA

Christine Dominas

2Department of Radiology, Brigham and Women’s Hospital, Harvard Medical School, 221 Longwood Ave, Boston, MA 02115, USA

Geoffrey Fell

3Department of Data Science, Dana Farber Cancer Institute, 450 Brookline Avenue, Boston, MA 02215, USA

Joshua D. Bernstock

1Department of Neurosurgery, Brigham and Women’s Hospital, 60 Fenwood Road, Boston, MA 02115, USA

Sarah Blitz

4Harvard Medical School, 25 Shattuck Street, Boston, MA 02115, USA

Debora Mazzetti

1Department of Neurosurgery, Brigham and Women’s Hospital, 60 Fenwood Road, Boston, MA 02115, USA

Mykola Zdioruk

1Department of Neurosurgery, Brigham and Women’s Hospital, 60 Fenwood Road, Boston, MA 02115, USA

Hassan Y. Dawood

1Department of Neurosurgery, Brigham and Women’s Hospital, 60 Fenwood Road, Boston, MA 02115, USA

Daniel V. Triggs

1Department of Neurosurgery, Brigham and Women’s Hospital, 60 Fenwood Road, Boston, MA 02115, USA

Sebastian W. Ahn

2Department of Radiology, Brigham and Women’s Hospital, Harvard Medical School, 221 Longwood Ave, Boston, MA 02115, USA

Sharath K. Bhagavatula

2Department of Radiology, Brigham and Women’s Hospital, Harvard Medical School, 221 Longwood Ave, Boston, MA 02115, USA

Shawn M. Davidson

5Lewis-Sigler Institute of Integrative Genomics, Princeton University, Princeton, NJ 08540, USA

Zuzana Tatarova

2Department of Radiology, Brigham and Women’s Hospital, Harvard Medical School, 221 Longwood Ave, Boston, MA 02115, USA

Michael Pannell

1Department of Neurosurgery, Brigham and Women’s Hospital, 60 Fenwood Road, Boston, MA 02115, USA

Kyla Truman

1Department of Neurosurgery, Brigham and Women’s Hospital, 60 Fenwood Road, Boston, MA 02115, USA

Anna Ball

1Department of Neurosurgery, Brigham and Women’s Hospital, 60 Fenwood Road, Boston, MA 02115, USA

Maxwell P. Gold

6Department of Biological Engineering, Massachusetts Institute of Technology, Cambridge, MA 02139, USA

Veronika Pister

6Department of Biological Engineering, Massachusetts Institute of Technology, Cambridge, MA 02139, USA

Ernest Fraenkel

6Department of Biological Engineering, Massachusetts Institute of Technology, Cambridge, MA 02139, USA

7Broad Institute of MIT and Harvard, Cambridge, MA 02139, USA

E. Antonio Chiocca

1Department of Neurosurgery, Brigham and Women’s Hospital, 60 Fenwood Road, Boston, MA 02115, USA

Keith L. Ligon

8Department of Pathology, Brigham and Women’s Hospital, 75 Francis Street, Boston, MA 02115, USA

Patrick Y. Wen

9Division of Neuro-Oncology, Dana Farber Cancer Institute, 450 Brookline Avenue, Boston, MA 02115, USA

Oliver Jonas

2Department of Radiology, Brigham and Women’s Hospital, Harvard Medical School, 221 Longwood Ave, Boston, MA 02115, USA

1Department of Neurosurgery, Brigham and Women’s Hospital, 60 Fenwood Road, Boston, MA 02115, USA

2Department of Radiology, Brigham and Women’s Hospital, Harvard Medical School, 221 Longwood Ave, Boston, MA 02115, USA

3Department of Data Science, Dana Farber Cancer Institute, 450 Brookline Avenue, Boston, MA 02215, USA

4Harvard Medical School, 25 Shattuck Street, Boston, MA 02115, USA

5Lewis-Sigler Institute of Integrative Genomics, Princeton University, Princeton, NJ 08540, USA

6Department of Biological Engineering, Massachusetts Institute of Technology, Cambridge, MA 02139, USA

7Broad Institute of MIT and Harvard, Cambridge, MA 02139, USA

8Department of Pathology, Brigham and Women’s Hospital, 75 Francis Street, Boston, MA 02115, USA

9Division of Neuro-Oncology, Dana Farber Cancer Institute, 450 Brookline Avenue, Boston, MA 02115, USA

†These authors contributed equally to this work.

Author contributions: P.P., O.J. and P.Y.W. conceptualized the study and developed the methodology. P.P. and O.J. supervised the study. P.P., O.J., C.D., S.W.A., S.K.B., S.M.D., and Z.T. performed tissue response measurements and data analysis. C.D., S.W.A., and O.J. prepared microdevices for the study.G.F., P.P., and O.J. performed statistical analysis. P.P., O.J., S.B., D.M., J.D.B., M.Z., and H.Y.D. performed data visualization and figure generation. M.P.G, V.P., and E.F. performed experiments and data analysis related to mass spectrometry. D.V.T., M.P., K.T., A.B., and C.D. were responsible for project administration. P.P. and O.J. wrote the original manuscript draft, and P.P., O.J., E.A.C., P.Y.W., G.F., and K.L.L. were primarily responsible for reviewing and editing the manuscript. O.J., P.P., E.A.C., and P.Y.W. acquired the funding for this study.

Supplementary Materials

Supplementary information.

GUID: 45F13BB7-3870-47CB-85CF-D5B9A6131711

Data Availability Statement

All data associated with this study are present in the paper or the Supplementary Materials. Additional data files related to transcriptomics and mass spectrometry that are not presented in the manuscript or in the supplementary materials section are available at 10.5281/zenodo.8226867. Microdevices were fabricated in the Jonas laboratory at Brigham and Women’s Hospital. Access to the IMD can be shared through a standard licensing agreement with MassGeneralBrigham.

Abstract

The lack of reliable predictive biomarkers to guide effective therapy is a major obstacle to the advancement of therapy for high-grade gliomas, particularly glioblastoma (GBM), one of the few cancers whose prognosis has not improved over the past several decades. With this pilot clinical trial (number NCT04135807), we provide first-in-human evidence that drug-releasing intratumoral microdevices (IMDs) can be safely and effectively used to obtain patient-specific, high-throughput molecular and histopathological drug response profiling. These data can complement other strategies to inform the selection of drugs based on their observed antitumor effect in situ. IMDs are integrated into surgical practice during tumor resection and remain in situ only for the duration of the otherwise standard operation (2 to 3 hours). None of the six enrolled patients experienced adverse events related to the IMD, and the exposed tissue was usable for downstream analysis for 11 out of 12 retrieved specimens. Analysis of the specimens provided preliminary evidence of the robustness of the readout, compatibility with a wide array of techniques for molecular tissue interrogation, and promising similarities with the available observed clinical-radiological responses to temozolomide. From an investigational aspect, the amount of information obtained with IMDs allows characterization of tissue effects of any drugs of interest, within the physiological context of the intact tumor, and without affecting the standard surgical workflow.

INTRODUCTION

Glioblastoma, one of the most aggressive human malignancies, was the first tumor to be dissected using genomic analysis (1), initiating the modern era of oncologic medicine. This molecular approach has led to the identification of several key driver genes, including epidermal growth factor receptor (EGFR), platelet-derived growth factor receptor alpha (PDGFRA), phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha (PIK3CA), phosphatase and tensin homolog (PTEN), neurofibromin 1 (NF1), RB transcriptional corepressor 1 (RB1), tumor protein p53 (TP53), isocitrate dehydrogenase [NADP(+)] 1 (IDH1), and pathways, including frequent alterations in chromatin remodeling (2). Although this has resulted in advances in diagnosis (3) and prognosis (4), it has not meaningfully affected treatment outcomes (5–7). Clinical trials investigating therapies specifically targeted against major oncogenic pathways like EGFR (8) or cyclin-dependent kinase 4/6 (CDK4/6) (9) have shown no benefit to patient survival. Presently, there are only two clinically relevant molecular biomarkers for predicting therapy response in high-grade gliomas (HGG). The first is the R132H mutation in the IDH1 gene, which defines a subfamily of tumors and correlates to a better prognosis (2) and increased response to IDH1 inhibitors (10). The second is the expression status of the O6 methyl guanine methyl transferase (MGMT) gene, which is used as a predictor of response to DNA alkylating agents such as temozolomide (TMZ) (11) and lomustine (12). The use of the MGMT promoter methylation status for predicting therapy efficacy is fraught with limitations, such as unreliability and inconsistency of current clinical assays as well as interobserver variability (13). Also, for the large group of patients with partial MGMT methylation, cutoff thresholds are not well established, leading to a gray zone in which the readout is generally inconclusive (14). In addition, no clinically validated biomarkers exist for the prediction of tumor sensitivity for the range of other therapies in HGG.

The disconnect between the abundance of molecular data available from each tumor and its lack of practical therapeutic value is due to many factors. First, in vitro and in vivo models, which are used to test drug effects, are often suboptimal and yield results that are not recapitulated in patients (15). Second, the notorious heterogeneity of GBM cell populations (16, 17) makes it difficult to generalize biological responses across all the different cellular subtypes of the tumor, let alone among different patients. Third, redundant oncogenic pathways (2, 18) and pronounced epigenetic plasticity characteristics of glioma cells (19) make these tumors adaptable and insensitive to isolated molecular hits.

For these reasons, lab-in-a-patient approaches have been gaining traction in recent years as a potentially more effective way to establish the benefit of experimental treatments, in a personalized manner. Modern phase 0 window of opportunity clinical studies, where experimental drugs are given systemically before tumor resection, have demonstrated their value in providing important information, including tissue concentrations, cell responses, and molecular biomarkers (20–22). However, they still suffer from limitations, in particular, the fact that each patient is generally exposed to only one drug at a time, making this design unsuitable for high-throughput efficacy screening. In addition, they cannot provide a comparison of effectiveness among different drugs and remain significantly resource intensive.

To fill this gap, and facilitate a high-throughput approach toward personalized drug screening on a patient-by-patient basis, we developed an intrasurgical approach that takes advantage of the operational window provided by the standard-of-care craniotomies for tumor resection. This approach probes a patient’s glioma with different pharmacological perturbations directly within its native microenvironment to obtain data on the personalized comparative drug responses that to date have been elusive to the field.

Our approach is based on tiny (6 mm by 0.7 mm) biocompatible intratumoral microdevices (IMDs) (Fig. 1A) (23) that are inserted into the tumor at the beginning of surgery and remain in place while the bulk of the tumor is resected to finally be removed together with the last piece of tumor at the end of surgery. This allows an exposure time of 2 to 3 hours within the intact tumor environment. During this time, they release nanodoses of drugs in a spatially confined manner such that they do not overlap (Fig. 1, ​B and ​C). Nanodoses are defined as amounts of drugs that result in negligible plasma concentrations (in our approach, each reservoir is loaded with, and releases directly into, the tumor ~1/100,000th of the amount of drug that is administered during normal systemic dosing) but can provide appropriate concentrations in the tissue immediately adjacent (~0.5 to 1 mm) to the point of release (24). After incubation, the exposed tissue is collected, and the effect of each drug on the tumor is assessed independently and in parallel, allowing multiplexed pharmacological measurements (Fig. 1D).

Intratumoral microdevices.

(A) Photography of microdevice in real dimensions, compared with a pencil tip. The computer-aided design next to the picture shows an enlargement of the device to display the configuration of each independent drug reservoir outlet. (B) List of drugs contained in the IMD and their mechanism of action. (C) Cartoon representing the workflow of IMD use. (D) Trial schema.

Here, we report the results of a first-in-human pilot clinical trial (NCT04135807) in patients with HGG that provides evidence of safety, efficient integration into the standard surgical workflow, and technical feasibility. Robust drug phenotypes, across a variety of different analytical techniques and ranging from standard immune-histopathology to modern assays like spatial transcriptomics and metabolomics, were obtained for a wide range of anticancer agents within the time of incubation afforded by standard surgical resection. The workflow allowed full integration and virtually no interference with standard surgical and clinical practice. We found early evidence that IMD readouts of intratumor nanodose drug effect parallel the extent of tumor response observed with systemic chemotherapy in patients with GBM. These results support the potential usefulness of integrating the data obtained from IMDs in the decisional algorithm of an effective, fully personalized adjuvant pharmacologic therapy for these patients.

RESULTS

Patient characteristics

A total of six patients were enrolled in this pilot study between April 2020 and August 2021. There was an equal frequency of female and male individuals (50% each). The median age was 76 years, with a range between 27 and 86. Five patients were diagnosed with glioblastoma, whereas the remaining patient had grade 4 astrocytoma (due to the presence of IDH1 mutation), according to the most recent World Health Organization (WHO) classification (3). Five out of six (83%) were newly diagnosed tumors and naïve to prior chemoradiation, whereas the remaining patient had tumor recurrence after radiation and TMZ, before trial enrollment. Tumor size averaged 81 cm3 (measured by the ellipsoid formula ½ × length × width × depth), with a range between 26.8 and 129 cm3. Five out of six tumors (83%) had the wild-type IDH1 gene, whereas one had the R132H mutation. Three tumors (50%) were partially methylated in the MGMT promoter, two (33%) were non-methylated, and one (16%) was methylated. Four out of six patients (66%) underwent craniotomy under general anesthesia, whereas two (34%) were operated under conscious sedation, or “awake surgery”, because of tumor involvement of eloquent regions (table S1).

Primary endpoint 1: The use of IMD in patients with glioma is safe

Postoperative follow-up for each patient was performed daily for the first 3 days after surgery, then at 12 ± 2 days, and lastly at 30 ± 4 days. There were no immediate (<48 hours after surgery) nor delayed (<30 days) adverse events (AEs). Twelve out of 12 inserted IMDs [100%; 90% confidence interval (CI) (61 to 100%)] were successfully retrieved, and none was lost or abandoned in the patient (Table 1). All postoperative bloodwork remained stable compared to preoperative values. Postoperative brain magnetic resonance imaging (MRI) with and without IV gadolinium was obtained within 48 hours after surgery: Gross total resection, meaning the removal of all contrast-enhancing tissue, was achieved in five out of six (83%) patients, whereas subtotal resection [residual contrast-enhancing nodule ≤5 cm3 (25)] was obtained in the remaining patient (patient 3) (table S1).

Table 1.

Trial results–primary endpoints.

CI, confidence interval. Column reporting feasibility endpoint is highlighted with an asterisk. Columns reporting safety endpoint are marked with daggers.

Primary endpoint 2: Feasibility and integration with neurosurgical practice

Eleven out of 12 [92%; 90% CI (66 to 100%)] total implanted IMDs provided specimens that could be successfully processed for the downstream molecular analysis. The only exception was due to inadvertent microdevice dislodgement of one IMD during tumor resection in patient 4. Each specimen was successfully aliquoted into multiple samples, which allowed different molecular analysis protocols, such as multiplexed immunofluorescence, transcriptional analysis, and mass spectrometry analysis, to be carried out simultaneously from the same tissue. Microdevices remained in situ in living tumor tissue for an average of 136 min during tumor resection (range, 122 to 155 min; SD, 11 min). The time between specimen removal and freezing was <1 min in all cases (Table 1).

The use of microdevices had a very low footprint on the surgery performance and in all other aspects of clinical care. We compared the application of IMD with a control cohort of nine similar patients with gliomas who underwent surgery but who were not included in the trial for not meeting inclusion criteria. IMD application did not result in significant changes in the surgical procedure and its aftermath: an average of 32.1 min to the full surgery (fig. S1). The length of postoperative intensive care length of surgery (skin incision to skin closure) was slightly increased in the patients with IMDs (300 min versus 230 min) (Fig. 2A). This difference, although not statistically significant, resulted from the trial Institutional Review Board (IRB) requirement to wait for intraoperative frozen section results (average of 19.6 min; range, 16 to 24 min) before an IMD could be implanted into each patient. This wait did not apply to standard craniotomies. Furthermore, there was an additional time (average 12.5 min; range, 5 to 20 min) required for the preparation and implantation of the devices into the tumor (IMD handling) before the actual tumor resection could begin. Thus, in total, the use of IMD-added unit (ICU, intensive care unit) stay (Fig. 2B) or total hospital stay (Fig. 2C) was not different.

Effects of IMD integration in the surgical care of patients with gliomas.

Comparison of common healthcare metrics between the group of patients receiving IMD implantation (red, n = 6) and a cohort of patients receiving standard surgery for HGG operated during the same period of time (gray, n = 9). (A) Surgical time (in minutes) from skin incision to skin closure during the surgical operation for tumor resection. (B) Duration of patient stays in the intensive care unit after surgery. (C) Duration of total hospital stays for each patient, from the day of surgery to the day of discharge from the hospital. (D) Cumulative expenses related to the surgical procedure. Reported are means and SD for each group. Comparisons use an unpaired t test, with two-tailed P values shown per each comparison.

In terms of surgical costs, the additional 32 min related to IMD use equated to an extra $7800 (7.5% of total surgery cost), calculated at the operating room utilization rate billed by our institution. This did not appear to significantly affect the overall surgical costs (Fig. 2D), although it is possible that this lack of significance might be due to the small cohort size and the high variability in several factors specific to each case, such as tumor size, anatomical complexity, equipment used, technical delays, and many more. Regarding laboratory analysis, all patients in both cohorts received the same standard-of-care diagnostic tests performed at the Brigham and Women’s Hospital central pathology core (hematoxylin and eosin stain, immunohistochemistry for established markers, MGMT promoter methylation analysis, and next-generation sequencing). The only additional costs in the IMD cohort were related to the exploratory analysis of the specimens, such as dedicated immunofluorescence, mass spectrometry, and spatial transcriptomics, itemized in table S2.

Reproducible measurements of localized intratumor drug release

Each of the pharmaceutical agents loaded into the IMD reservoirs was released upon implantation into a confined region of the tumor directly adjacent to its reservoir. The local concentration was determined by the ratio of drug versus polyethylene glycol (PEG) polymer in the formulation, and the release kinetics and diffusion distance were controlled by the molecular weight of the polymer being used. We demonstrated uniformity of release and tissue transport for two agents with opposite solubility properties: doxorubicin, which is water soluble, and lapatinib, which is insoluble. These drugs were chosen because they are readily detected and quantitated by autofluorescence. We observed a distance-dependent concentration gradient where higher concentrations were present at the device-tissue interface and decreased gradually with increasing radial distance from the reservoir (Fig. 3, ​A and ​B). The presence of the drug in the tissue at the correct reservoir site also confirmed that the IMD did not move during the implantation time in the tumor or during excision and processing. We observed only moderate variability in the diffusion curves of <20% from the mean in the maximum exposure concentrations, and <15% in the average drug exposure, across all six patients (Fig. 3, ​C and ​D). For non-autofluorescing drugs, matrix-assisted laser desorption/ionization–time-of-flight (MALDI-TOF) analysis showed a consistent pattern of drug penetration into tumor tissue after release from the IMD, also establishing a strong colocalization between PEG, the vehicle used for solubilizing and controllably releasing the drugs, and the drug moieties themselves (fig. S2).

Characterization of intratumoral drug diffusion.

(A) Drug release profiles from each patient for doxorubicin. (B) Drug release profiles from each patient for lapatinib. The inset shows the typical two-dimensional spatial profile of drug distribution. Inset scale bar, 200 μm. (C) Variation in maximum and average dose for doxorubicin between patients. (D) Variation in maximum and average dose for lapatinib between patients. Each dot on the graph represents the mean of triplicate measurements of drug release profile values (maximum or average dose) from biologically distinct regions of each patient’s tumor. Error bars represent SD.

Measurement of tumor drug sensitivity and identification of response biomarkers

In each patient sample, we determined the tumor sensitivity to drugs by measuring the expression of cleaved caspase-3 (CC3), a marker for apoptosis, and phospho-H2A histone family member X (pH2AX), a marker for DNA damage, in each drug-exposed tumor section. Our first analysis focused on the tumor sensitivity to TMZ, because this is the most widely used agent in this patient population, and offered the opportunity to compare the IMD readout with clinical-radiological response to the drug. TMZ is a DNA alkylating agent that causes apoptosis by inducing DNA damage (26). Thus, DNA damage, measured by pH2AX, is an early marker of drug effect for agents such as TMZ (27). Accordingly, we quantified the amount of pH2AX and CC3 induced by TMZ in each patient tumor across multiple spatially distant tumor regions from different microdevice reservoirs implanted in the same patient (Fig. 4 and fig. S3).

Differential tumor response to temozolomide.

(A) Quantification of immunohistochemistry stains for pH2AX and CC3 in IMD tissue from each numbered patient. Each dot on the graph represents a unique biological replicate measurement from a tumor region within the area of the tumor exposed to TMZ for a given patient. Statistical comparisons were made using the repeated-measures ANOVA test, with P values shown for each comparison (in parentheses). *P < 0.05; ** P < 0.01. (B) Distance and concentration-dependent analysis of pH2AX and CC3 stains across the six patients. Graphs are shown as means (black) and SD (gray) where available. PAT, patient. ns, not significant.

The highest amount of DNA damage was observed in patient 3 (65.8% of cells in the drug-exposed region), and the lowest amount was observed in patients 5 and 6 (9.8 and 17.7%, respectively). A repeated-measures analysis of variance (ANOVA) test indicated that the means of the comparisons of patient 3 versus patient 5 and patient 3 versus patient 6 were highly significantly different with a P value of 0.0037 and 0.0051, respectively, whereas the comparison of patient 5 versus patient 6 was not significant (P value of 0.4948).

Apoptosis induction as measured by CC3 expression was generally low (<5%) except in patient 3, where it was expressed in 13.1% of cells. A repeated-measures ANOVA test showed that the means of the comparisons of patient 3 versus patient 5 and patient 3 versus patient 6 were significantly different, with a P value of 0.028 and 0.029, respectively, whereas the comparison of patient 5 versus patient 6 was not significant (P value of 0.336) (Fig. 4A). Because the time of drug exposure was about 2 hours, CC3, a marker of apoptosis, was not expected to be highly expressed yet in the samples, in keeping with prior observations made by us and others that CC3 becomes best detectable in a time window between 3 and 9 hours after drug incubation (23, 28, 29).

Determination of concentration dependence of antitumor effect for temozolomide

We exploited the distance-dependent concentration gradient of drugs eluting from the IMD (shown in Fig. 3 and fig. S2) to determine the dose dependence of the antitumor effect for TMZ (Fig. 4B). We generally observed high DNA damage scores at the immediate vicinity of the drug reservoir, which corresponded to the highest concentration, sharply declining sensitivity at lower doses. Patient 3 exhibited the highest IMD response across the entire concentration range and maintained >40% of cells with confirmed pH2AX down to 0.1 μM. Patient 5, the only patient in the cohort who had already failed systemic TMZ, showed the lowest response across all TMZ concentrations (<10%). Patient 4 exhibited an unexpected response pattern, with a peak in DNA damage further away from the microdevice (Fig. 4B).

As we investigated this apparent anomaly, MALDI analysis of TMZ concentration of the same tissue of patient 4 demonstrated a very similar pattern (fig. S4). This was due to the presence of necrotic/hemorrhagic tissue directly adjacent to the microdevice, which biased the molecular readout in that region and which could not be averaged with other specimens, because of the lack of a duplicate IMD in this patient. Although the result of a specimen aberration, this further confirmed the strong correlation between measured TMZ concentrations and cellular response.

Comparisons between IMD drug responses and clinical responses to systemic treatment

TMZ is the most widely used drug in the treatment of GBM, and several patients in our trial received it systemically, either before or after IMD insertion, as part of the standard of care. Thus, although our trial was not designed to choose chemotherapy agents based on IMD data, we still could compare the observed clinical-radiological response to systemic TMZ with the patient-specific response to TMZ in the IMD-exposed tissue. Patients 1, 2 (MGMT partially methylated), 4, and 6 (MGMT unmethylated) all showed no to minimal response to TMZ in the IMD, as measured by pH2AX expression in the drug-treated region (Fig. 5A). Of these, only patient 6 received adjuvant TMZ, with no observed benefit (Fig. 5, ​B and ​C), in concordance with the poor tissue response observed in the IMD analysis. None of the other three patients received adjuvant TMZ, and therefore, no direct connection could be established between IMD readout and clinical response.

Clinical-molecular comparisons.

(A) Quantification of specific in situ response to TMZ (by pH2AX immunostaining) for each patient in the study as determined by IMD analysis. Each point represents a measurement from a distinct tumor region comprising 800 mm by 400 mm exposed to the drug. Error bars display means and SD. Comparisons use a repeated-measures ANOVA test with P values shown per each comparison (in parentheses). **P < 0.01; ns, not significant. (B) Survival data for each patient in the study, including the type and timing of adjuvant therapy administered. Specific patients are color-coded to better visualize the alignment among radiologic data, IMD response, and survival. (C) Time-course MRIs of three representative patients who received systemic therapy after surgery and IMD analysis.

On the contrary, patient 3, a 72-year-old female with a partially methylated MGMT promoter, whose IMD analysis showed a marked response to TMZ (Fig. 5A), had an overall survival of 20 months after receiving systemic TMZ (Fig. 5B), with radiological evidence of tumor response (Fig. 5C). This was despite having a subtotal tumor resection, in itself an unfavorable prognostic factor (30). The patient expired because of an unrelated cardiovascular event, although she had remained neurologically stable. In this specific case, the patient’s MGMT promoter methylation status, assessed with the gold standard bisulfite sequencing analysis (31), would not have predicted the positive clinical response that was observed, which instead was better recapitulated by the IMD analysis. Conversely, patient 6, an 81-year-old female with unmethylated MGMT promoter, and whose IMD analysis showed no response to TMZ, experienced clinical and radiological evidence of tumor progression within 6 months after treatment and proceeded to palliative Avastin. This progression occurred despite a gross total tumor resection and the patient receiving the same adjuvant treatment protocol as patient 3.

Patient 5, a 27-year-old male with recurrent grade 4 astrocytoma and a methylated MGMT promoter, underwent surgery after confirmed radiological progression while on TMZ. In this patient, IMD analysis showed no response to TMZ, confirming the lack of efficacy, which was already observed clinically and radiologically, despite the favorable MGMT promoter methylation status (Fig. 5A).

Measurement of drug sensitivity for other agents to generate treatment hypotheses

Although several drugs were tested in the IMD, only TMZ was clinically administered to the patients. Therefore, as part of the exploratory endpoints of this trial, we quantified the sensitivity of each tumor to several other agents contained in the IMD. Measurement of pH2AX was initially used as a validated biomarker across the multiple drugs because each drug has been previously shown to induce some degree of DNA damage and induction of pH2AX foci (fig. S5) (32–33). We observed that patient 5, the only patient with a recurrent tumor, was generally resistant to all tested drugs, possibly confirming the mounting evidence that recurrent tumors are generally less responsive to any subsequent pharmacotherapy (Fig. 6) (34). Patient 6 (unmethylated tumor), although unresponsive to TMZ, appeared to be potentially sensitive to several other drugs (Fig. 6), which supports the strategy to avoid TMZ as a first-line treatment in these particularly difficult-to-treat patients, in lieu for alternative, more effective drugs.

Personalized tumor responses to different drugs.

Comparison of tumor response to several agents by DNA damage (pH2AX). Values are expressed as a normalized marker index for all cells in drug-exposed tumor regions. Each dot on the graph represents a unique biological replicate measurement from a tumor region within the area of the tumor exposed to the listed drug, for a given patient. Mean and SE are shown. Patient 5 (cyan) and patient 6 (magenta) are highlighted. Comparisons use a one-way ANOVA test, with P values shown for each comparison between patients 5 and 6. *P < 0.05; **P < 0.01.

To offer further proof of principle for the relevance of the IMD approach to providing accurate response biomarkers, we also interrogated the tumor tissue exposed to targeted therapies to characterize tumor response by different molecular means. As an example, using MALDI-based tumor metabolomics, we observed specific changes in the glutathione detox pathway in response to lapatinib. This was in keeping with prior evidence that EGFR inhibition is associated with changes in glutathione metabolism in tumor tissue (35–37). In addition to suggesting a drug-specific response by a previously validated biomarker, this analysis also offered the opportunity to discover metabolites signatures associated to drugs of interest. We next applied spatial transcriptomics to determine pathway signatures of drug response. Gene expression changes in response to abemaciclib showed that the exposed tumor up-regulated genes involved in interferon and cytokine pathways and down-regulated genes involved with cell cycle and DNA repair (Fig. 7B). These responses have already been described by others in association with this drug (38, 39) and are used here to demonstrate drug-biomarker specificity. Similar drug-specific responses were evident with the tissue exposed to everolimus, an mTOR inhibitor, and osimertinib and lapatinib, both EGFR pathway inhibitors. On one hand, these results confirm that each drug affects its expected targets, strongly suggesting that the IMD tissue analysis is specific. In addition, they also underscore the discovery potential of this approach, highlighting pathways/genes not previously described either as biomarkers or functional molecular underpinnings of drug pharmacodynamics. The specificity of the observed transcriptional changes related to each drug was further supported not only by the fact that osimertinib and lapatinib produced two similar patterns (both being EGFR-targeting drugs), as shown in Fig. 7B, but also by the reproducibility of the patterns observed across different patients (fig. S6).

In situ phenotyping of targeted drug responses in patient 2.

(A) Biomarker discovery using metabolomics: MALDI images of metabolite changes in tumor in response to lapatinib exposure. (i) Tumor cross section shows three elevated metabolite concentrations in region of drug exposure. (ii) Glutathione concentrations are increased at lapatinib drug reservoir release zone. (iii) Lapatinib distribution measured by autofluorescence shows spatial overlap with elevated metabolite concentrations. Scale bar, 500 μm. (B) Volcano plots of spatial transcriptomics and pathway analysis from tumor specimens exposed to each targeted therapy. On-target, drug-specific effects are confirmed for four targeted agents used on IMD. The most down-regulated (blue) and up-regulated (red) pathways are shown for each drug. Control is tumor tissue adjacent to the microdevice but not exposed to any drugs. P values were generated from unpaired t tests based on four distinct regions of interest per condition. GPCR, G protein–coupled receptor; MAPK, mitogen-activated protein kinase.

DISCUSSION

Obtaining phenotypic information on tumor responses to drugs to enable precision medicine remains an unmet need in the treatment of gliomas. With this first-in-human pilot trial, we provide evidence of safety and feasibility for the use of intratumoral, drug-releasing microdevices as an approach to characterize and compare the efficacy of different pharmacologic therapies in patients with gliomas, in a personalized manner.

The main goals of this study were to demonstrate that microdevices can be easily incorporated into standard neurosurgical practice, with minimal impact on the operative protocols, no significant burden on healthcare costs, and no evidence of adverse effects and to provide valuable biological data that can be integrated with, and potentially be superior to, other currently used biomarkers. The amount of information obtained with this approach, which directly integrates surgery with bioengineering, pharmacology, and multiomics analysis, provides relevant arguments for their possible use in the clinical setting for patients with glioma, as well as for their potential as investigative tools when they are combined with high-throughput platforms like MALDI and spatial transcriptomics.

One potentially limiting aspect of this study is the relatively short indwelling time of the microdevices, which was dictated by the need to minimize changes to the current standard of patient care and the decision to not submit patients to an additional invasive IMD implantation procedure several days before surgery. During the available ~2.3-hour incubation period, we demonstrated the detection of early markers of drug effects by inducing cellular stress response in a drug- and concentration-dependent manner. We observed robust activation of early markers of DNA damage (phosphorylation of histone gamma) and low to moderate activation of molecular cascades that lead to cell death (CC3). We found that the amount of pH2AX expression in response to TMZ treatment is congruent with the molecular characterization of the patient’s tumor and aligned with the clinical responses observed across the patients who received systemic TMZ treatment. This is particularly evident in the case of patients 3 and 5, where MGMT promoter methylation status by itself was not predictive of clinical response, which was instead correctly identified by the IMD measurement.

Another limitation was that the longitudinal clinical-radiological correlates were available only for three patients because the other three patients did not receive adjuvant therapy after surgery. This was due to patient/caregiver decisions, dictated by coronavirus disease 19–related contingencies in two cases and changes in goals of care in the remaining patient, but did not prevent the achievement of the study objectives.

Intratumor TMZ dose from systemic administration was not measured, and there is still generally a lack of available data on intratumor drug concentrations for most agents. As more such data are obtained, the measurements of a dose-dependent effect provided by the IMD may provide insight into minimum required intratumor concentrations to obtain threshold amounts of DNA damage or apoptosis that need to be reached for durable effects, which may, in turn, inform systemic dosing regimens. These thresholds for localized drug efficacy will be defined in subsequent larger studies.

Larger clinical studies will be needed to confirm the predictive capability of the IMD to identify systemic responders and to quantitatively define exact thresholds of IMD response correlating with favorable clinical outcomes. We have focused the current study on agents that had been previously US Food and Drug Administration (FDA) approved and have been used in the treatment of gliomas, either as a standard of care or in clinical trials. For agents that do not penetrate the blood-brain barrier, IMD readouts of intratumor effect may help determine the minimum effective intratumor concentrations required, and this could guide the decision to implement different delivery techniques, such as convection-enhanced or nanoparticle-mediated delivery to achieve sufficient intratumor drug concentrations.

Although the current study focused on rapidly acting cytotoxic and targeted agents, the length of exposure is likely not enough to detect changes in adaptive immune response, which have been shown to occur over the course of 2 days or longer (40). Supported by the evidence of safety and non-futility provided with this first study iteration, the feasibility and safety of a two-staged procedure, consisting of insertion by a minimally invasive surgery, and retrieval 72 hours later by craniotomy, are currently underway. This will provide data to compare biological readouts between short and long exposures and define whether a two-surgery approach is necessary to maximize data or whether the predictive values obtained with a single surgery and shorter exposure are sufficient to reliably inform therapy.

In addition to providing the ability to directly test a range of drugs in a patient, the use of IMDs in gliomas offers opportunities to answer questions that so far have been elusive: First, by mixing different drugs into each reservoir, this technology will allow in the future to safely and rapidly test the efficacy of drug combinations, as we have already demonstrated in a preclinical setting (40). The synergistic use of different drugs is a diffuse practice in several cancers but is underused in glioblastomas, despite convincing preclinical evidence of its efficacy (41–43).

Second, the analysis of microdevice-exposed specimens allows a realistic vantage point into the tumor microenvironment, particularly how drugs also affect non-neoplastic cells, such as immune cells, astrocytes, and neurons. For example, it is still not clear how drugs modulate the antitumor immune response: Chemotherapy is generally believed to be immunosuppressive (44). However, although some have confirmed a detrimental effect of TMZ against T and B cells in mouse models of GBM, with resultant further impairment of an already weak antitumor response (45), others have shown that TMZ might preferentially deplete immune-suppressive CD4 regulatory T cells (46, 47). In theory, any drugs might display unexpected effects against nontumor cells, which, in turn, can affect clinical outcomes. Last, by providing a measurable drug gradient within the specimen, which is achievable through detection by autofluorescence (fig. 2) or using MALDI mass spectrometry (fig. S2) [and previously published evidence (48)], the analysis of microdevice specimens allows quantification of tissue concentrations at which each drug is biologically effective against the tumor. The intrasurgical use of IMDs in patients with gliomas represents a feasible and promising approach that addresses the need to maximize the efficacy of pharmacotherapy and to understand drug mechanisms of action in the most representative and predictive model.

MATERIALS AND METHODS

Study design

This is an investigator-initiated, nonrandomized, nonblinded, single-center phase 1 study (NCT04135807). The two co-primary endpoints were to establish the safety and feasibility of IMD use in patients with gliomas undergoing resective surgery.

The limiting toxicities were defined as either a grade 3 or higher AE associated with the IMD or situations where the IMD becomes lost or unretrievable. Stipulations for early termination of the study were based on interim analysis after the sixth enrolled patient. The device would be considered unsafe if ≥3 limiting toxicities were observed in the first six patients or ≥4 in a total of 12 patients. IMD futility was measured with regard to individual devices, defined as the successful extraction of the implanted device containing viable tissue for histopathologic analysis. We considered the procedure successful if the estimate for retrievable success rate had a lower bound that exceeded 50%. Because this was a first-in-human study, enrollment of 12 patients was initially calculated as necessary to reach a power of 90% using a beta-binomial distribution. For safety, we assumed a dose-limiting toxicity rate = 5%. For feasibility, we assumed a device failure rate = 5%. All available data from all patients were included, with no omissions. An interim analysis after the sixth patient revealed that both study endpoints were met, and because this was not an interventional study with a therapeutic goal, accrual to this stage of the study was terminated. All consecutive patients meeting inclusion criteria were offered to participate in the study, without randomization or blinding.

Each patient received two IMDs to maximize data collection and replication. The number of replicates for each reported experiment is specified in the respective figure legends.

Patient selection and enrollment

The trial was open to any patients older than 18 years of age, with known or suspected supratentorial glioma (WHO, grades 2 to 4) observed in a brain MRI with and without intravenous gadolinium, and for which a craniotomy for tumor resection was indicated. The required lesion volume was greater than 5 cm3. A Karnofsky performance score ≥60 was also required. Exclusion criteria were enrollment in concomitant trials. Patients with coagulopathies, platelet counts <100,000 per ml, or deep-seated tumors (in the brainstem and/or thalamus) were also excluded. Eligibility was assessed in the clinic, and details were explained before informed consent was obtained. All patients underwent surgery and follow-up at the Brigham and Women’s Hospital and Dana Farber Cancer Institute, both affiliated with Harvard Medical School, Boston, USA. After obtaining Investigational New Drug approval for the use of IMDs, all aspects of the trial were approved on 23 October 2019 by the Institutional Review Board (IRB) at the Dana Farber Cancer Institute under protocol number 18-623. The study was registered at https://clinicaltrials.gov under the identifier NCT04135807. The purpose of this study was explorative, to investigate the safety and feasibility of integrating IMD use during otherwise standard brain surgery for tumor resection. Consequently, as specified in the trial consent, no information obtained from the IMD was used to make medical decisions regarding the postsurgical care of patients. The accrual diagram is presented in fig. S7.

IMD development

IMDs were manufactured from implant-grade radiopaque polyetherketoneketone with 20% barium sulfate (Oxford Performance Materials) on a five-axis computer numerical control micromachining station using subtractive machining techniques and inspected in accordance with quality control guidelines, as previously described (49). A rigid nitinol guidewire of 0.25-mm diameter, designed to increase visualization of the devices in the operatory field and within the specimen, was attached to the IMD body using medical-grade epoxy (EPO-TEK MED-301) and a curing step. IMDs were rinsed in United States Pharmacopeia (USP)–grade sodium hydroxide and endotoxin-free water (49, 50).

All pharmaceutical agents used are FDA approved and were purchased commercially. The list of drugs and their mechanism of action is provided in Fig. 1B. The drugs were prepared, mixed with USP-grade PEG matrix, and loaded. IMDs were singularly placed into 15-ml polypropylene tubes and into a sterilization pouch. Pouches were sent for gamma irradiation, followed by endotoxin and sterility testing, before they were stored in the operating room pharmacy for off-the-shelf use.

Intraoperative IMD insertion

For every patient, surgery proceeded as per standard neurosurgical practice. All surgeries were performed by the first author (P.P.) for the benefit of procedural consistency. After exposure of the brain surface and localization of the lesion, either by direct visualization or through image-guided means (neuronavigation or ultrasound), an intraoperative biopsy was obtained (fig. S8A), because the confirmation of the nature of the lesion by frozen histopathology analysis was required before proceeding with IMD implantation. The two microdevices per patient were then placed free-handedly, about 10 to 15 mm apart from each other so that one IMD did not affect the other yet still close enough to be in a circumscribed part of the tumor and not impede the resection of the rest of the tumor, whereas the IMDs remained indwelled within the tissue (fig. S8B). The IMDs were inserted into tissue for their entire length, so that their terminal bevel was anchored just underneath the pia mater, increasing their stability. The nitinol tail remained visible during the entire time (fig. S8, C to F) to minimize the involuntary displacement of the IMD and to facilitate its retrieval. Two IMDs were implanted in each patient.

IMD retrieval

At the end of resection, the small part of the tumor containing the IMD was removed under operating microscope visualization, ensuring that at least 1 cm of untouched tissue around the IMD was recovered. Immediately upon removal, the specimen was placed in liquid nitrogen or dry ice and transported to the laboratory for downstream analysis (fig. S8, E and F).

Specimen analysis

For every patient, a fragment of the tumor was sent to the pathology laboratory for standard diagnostic immunohistochemistry, MGMT promoter methylation analysis, and genetic profiling by next-generation sequencing analysis. The remainder of the tissue was used for additional exploratory correlative studies.

The tumor specimen containing the IMD was snap-frozen immediately upon surgical resection. The tumor-device specimen was sectioned on a standard cryotome, and several serial tissue sections of 8-μm thickness were collected at each drug reservoir position of the IMD, as previously described (51). Imaging of drug autofluorescence and quantitation was performed as previously described (52) and further detailed in Supplementary Materials and Methods.

Statistical analysis

Continuous measures are reported as means with SD, and categorical measures as counts and percentages. All laboratory experiments and data analysis were performed at least in triplicates (except for patient 4, the only instance where only one IMD could be used), reporting both means and SD for each dataset. The data points were plotted and analyzed using GraphPad Prism version 9.0. In Figs. 4 and ​5, to account for interdependence of individual measurements within each patient, we performed a repeated-measures ANOVA to examine statistical significance of differences in ph-H2AX expression in TMZ-exposed tumor regions between patient 3 and patients 5 and 6. Post hoc paired t tests were performed using a Bonferroni-corrected a = 0.017. In Fig. 6, a one-way ANOVA test was applied to compare patients 5 and 6.

Supplementary Material

Supplementary information

Acknowledgments:

We thank J. Jakubik for helping with CAD and figure design.

Funding:

This work was supported by NIH R37CA224144 (to O.J.) and P41EB028741 (to O.J.) and National Institute of Neurological Disorders and Stroke 1R01NS116144 (to P.P.).

Competing interests:

P.P. is a cofounder and is on the Board of Directors of Ternalys Therapeutics. He is a named inventor on microRNA-related patent PCT/US2019/029988. J.D.B. has an equity position in Treovir Inc., an oHSV clinical-stage company, and is a member of the UpFront Diagnostics, Centile Bioscience, and NeuroX1 Boards of Scientific Advisors. E.A.C. is currently (within the last year) an adviser to Amacathera, Bionaut Labs, Candel Therapeutics Inc., Genenta Inc., Insightec Inc., DNAtrix Inc., Seneca Therapeutics, and Theriva. He has equity options in Bionaut Laboratories, DNAtrix, Immunomic Therapeutics, Seneca Therapeutics, and Ternalys Therapeutics. He is a cofounder and is on the Board of Directors of Ternalys Therapeutics. In the past (over 12 months ago), he has also advised Alcyone, Amasa, Bexion, Biogen, GSK Oncorus, Merck, Tocagen, Ziopharm, Stemgen, NanoTx., Ziopharm Oncology, Cerebral Therapeutics, Ceregene, GSK, Merck, Janssen, Karcinolysis, Shanghai Biotech, Sangamo Therapeutics, and Voyager Therapeutics. He has received research support from NIH, U.S. Department of Defense, American Brain Tumor Association, National Brain Tumor Society, Alliance for Cancer Gene Therapy, Neurosurgical Research Education Foundation, Advantagene, NewLink Genetics, and Amgen. He also is a named inventor on patents related to oncolytic HSV1 (U.S. patent nos. US10,806,761B2, US10,232,002B1, and US2017/0015757A1) and noncoding RNAs (PCT/US2019/029988). K.L.L. is a founder of Travera. He is a consultant for Travera, Bristol Meyers Squibb, Blaze Bioscience, and Integragen and has received research funding from Bristol Meyers Squibb and Eli Lily. P.Y.W. received research support from Astra Zeneca, Black Diamond, Bristol Meyers Squibb, Celgene, Chimerix, Eli Lily, Erasca, Genentech/Roche, Kazia, MediciNova, Merck, Novartis, Nuvation Bio, Servier, Vascular Biogenics, and VBI Vaccines, and he is an advisory board/consultant for Astra Zeneca, Black Diamond, Celularity, Chimerix, Day One Bio, Genenta, Glaxo Smith Kline, Merck, Mundipharma, Novartis, Novocure, Nuvation Bio, Prelude Therapeutics, Sapience, Servier, Sagimet, Vascular Biogenics, and VBI Vaccines. O.J. is a coinventor of the IMD technology (U.S. patent no. US10390702B2) and is a consultant with compensation (non-equity) to Kibur Medical. The other authors declare that they have no competing interests.

Data and materials availability:

All data associated with this study are present in the paper or the Supplementary Materials. Additional data files related to transcriptomics and mass spectrometry that are not presented in the manuscript or in the supplementary materials section are available at 10.5281/zenodo.8226867. Microdevices were fabricated in the Jonas laboratory at Brigham and Women’s Hospital. Access to the IMD can be shared through a standard licensing agreement with MassGeneralBrigham.

REFERENCES AND NOTES

1. Cancer Genome Atlas Research Network, Comprehensive genomic characterization defines human glioblastoma genes and core pathways. Nature 455, 1061–1068 (2008).[PMC free article] [PubMed] [Google Scholar]

2. Bhaskaran V, Nowicki MO, Idriss M, Jimenez MA, Lugli G, Hayes JL, Mahmoud AB, Zane RE, Passaro C, Ligon KL, The functional synergism of microRNA clustering provides therapeutically relevant epigenetic interference in glioblastoma. Nat. Commun 10, 442 (2019).[PMC free article] [PubMed] [Google Scholar]

3. Louis DN, Perry A, Wesseling P, Brat DJ, Cree IA, Figarella-Branger D, Hawkins C, Ng HK, Pfister SM, Reifenberger G, Soffietti R, von Deimling A, Ellison DW, The 2021 WHO classification of tumors of the central nervous system: A summary. Neuro Oncol. 23, 1231–1251 (2021).[PMC free article] [PubMed] [Google Scholar]

4. Katsigiannis S, Grau S, Krischek B, Er K, Pintea B, Goldbrunner R, Stavrinou P, MGMT-positive vs MGMT-negative patients with glioblastoma: Identification of prognostic factors and resection threshold. Neurosurgery 88, E323–E329 (2021). [PubMed] [Google Scholar]

5. Stupp R, Mason WP, van den Bent MJ, Weller M, Fisher B, Taphoorn MJB, Belanger K, Brandes AA, Marosi C, Bogdahn U, Curschmann J, Janzer RC, Ludwin SK, Gorlia T, Allgeier A, Lacombe D, Cairncross JG, Eisenhauer E, Mirimanoff RO, Radiotherapy plus concomitant and adjuvant temozolomide for glioblastoma. New England J. Med 352, 987–996 (2005). [PubMed] [Google Scholar]

6. Lim-Fat MJ, Youssef GC, Touat M, Iorgulescu JB, Whorral S, Allen M, Rahman R, Chukwueke U, McFaline-Figueroa JR, Nayak L, Clinical utility of targeted next-generation sequencing assay in IDH-wildtype glioblastoma for therapy decision-making. Neuro Oncol. 24, 1140–1149 (2022).[PMC free article] [PubMed] [Google Scholar]

7. Wen PY, Weller M, Lee EQ, Alexander BM, Barnholtz-Sloan JS, Barthel FP, Batchelor TT, Bindra RS, Chang SM, Chiocca EA, Cloughesy TF, DeGroot JF, Galanis E, Gilbert MR, Hegi ME, Horbinski C, Huang RY, Lassman AB, Le Rhun E, Lim M, Mehta MP, Mellinghoff IK, Minniti G, Nathanson D, Platten M, Preusser M, Roth P, Sanson M, Schiff D, Short SC, Taphoorn MJB, Tonn J-C, Tsang J, Verhaak RGW, von Deimling A, Wick W, Zadeh G, Reardon DA, Aldape KD, van den Bent MJ, Glioblastoma in adults: A Society for Neuro-Oncology (SNO) and European Society of Neuro-Oncology (EANO) consensus review on current management and future directions. Neuro Oncol. 22, 1073–1113 (2020).[PMC free article] [PubMed] [Google Scholar]

8. Weller M, Butowski N, Tran DD, Recht LD, Lim M, Hirte H, Ashby L, Mechtler L, Goldlust SA, Iwamoto F, Rindopepimut with temozolomide for patients with newly diagnosed, EGFRvIII-expressing glioblastoma (ACT IV): A randomised, double-blind, international phase 3 trial. Lancet Oncol. 18, 1373–1385 (2017). [PubMed] [Google Scholar]

9. Miller TW, Traphagen NA, Li J, Lewis LD, Lopes B, Asthagiri A, Loomba J, De Jong J, Schiff D, Patel SH, Tumor pharmacokinetics and pharmacodynamics of the CDK4/6 inhibitor ribociclib in patients with recurrent glioblastoma. J. Neurooncol 144, 563–572 (2019).[PMC free article] [PubMed] [Google Scholar]

10. Mellinghoff IK, Ellingson BM, Touat M, Maher E, Macarena I, Holdhoff M, Cote GM, Burris H, Janku F, Young RJ, Ivosidenib in isocitrate dehydrogenase 1–mutated advanced glioma. J. Clin. Oncol 38, 3398–3406 (2020).[PMC free article] [PubMed] [Google Scholar]

11. Brandner S, McAleenan A, Kelly C, Spiga F, Cheng H-Y, Dawson S, Schmidt L, Faulkner CL, Wragg C, Jefferies S, MGMT promoter methylation testing to predict overall survival in people with glioblastoma treated with temozolomide: A comprehensive metaanalysis based on a Cochrane Systematic Review. Neuro Oncol. 23, 1457–1469 (2021).[PMC free article] [PubMed] [Google Scholar]

12. Herrlinger U, Tzaridis T, Mack F, Steinbach JP, Schlegel U, Sabel M, Hau P, Kortmann R-D, Krex D, Grauer O, Lomustine-temozolomide combination therapy versus standard temozolomide therapy in patients with newly diagnosed glioblastoma with methylated MGMT promoter (CeTeG/NOA–09): A randomised, open-label, phase 3 trial. Lancet 393, 678–688 (2019). [PubMed] [Google Scholar]

13. Mansouri A, Hachem LD, Mansouri S, Nassiri F, Laperriere NJ, Xia D, Lindeman NI, Wen PY, Chakravarti A, Mehta MP, MGMT promoter methylation status testing to guide therapy for glioblastoma: Refining the approach based on emerging evidence and current challenges. Neuro Oncol. 21, 167–178 (2019).[PMC free article] [PubMed] [Google Scholar]

14. Ben-David U, Siranosian B, Ha G, Tang H, Oren Y, Hinohara K, Strathdee CA, Dempster J, Lyons NJ, Burns R, Genetic and transcriptional evolution alters cancer cell line drug response. Nature 560, 325–330 (2018).[PMC free article] [PubMed] [Google Scholar]

15. Ben-David U, Ha G, Tseng Y-Y, Greenwald NF, Oh C, Shih J, McFarland JM, Wong B, Boehm JS, Beroukhim R, Patient-derived xenografts undergo mouse-specific tumor evolution. Nat. Genet 49, 1567–1575 (2017).[PMC free article] [PubMed] [Google Scholar]

16. Verhaak RG, Hoadley KA, Purdom E, Wang V, Qi Y, Wilkerson MD, Miller CR, Ding L, Golub T, Mesirov JP, Integrated genomic analysis identifies clinically relevant subtypes of glioblastoma characterized by abnormalities in PDGFRA, IDH1, EGFR, and NF1. Cancer Cell 17, 98–110 (2010).[PMC free article] [PubMed] [Google Scholar]

17. Gularyan SK, Gulin AA, Anufrieva KS, Shender VO, Shakhparonov MI, Bastola S, Antipova NV, Kovalenko TF, Rubtsov YP, Latyshev YA, Investigation of Inter- and intratumoral heterogeneity of glioblastoma using TOF-SIMS. Mol. Cell. Proteomics 19, 960–970 (2020).[PMC free article] [PubMed] [Google Scholar]

18. Liu Y, Li Z, Zhang M, Zhou H, Wu X, Zhong J, Xiao F, Huang N, Yang X, Zeng R, Rolling-translated EGFR variants sustain EGFR signaling and promote glioblastoma tumorigenicity. Neuro Oncol. 23, 743–756 (2021).[PMC free article] [PubMed] [Google Scholar]

19. Neftel C, Laffy J, Filbin MG, Hara T, Shore ME, Rahme GJ, Richman AR, Silverbush D, Shaw ML, Hebert CM, An integrative model of cellular states, plasticity, and genetics for glioblastoma. Cell 178, 835–849 (2019).[PMC free article] [PubMed] [Google Scholar]

20. Vigneswaran K, Boyd NH, Oh S-Y, Lallani S, Boucher A, Neill SG, Olson JJ, Read RD, YAP/TAZ transcriptional coactivators create therapeutic vulnerability to verteporfin in EGFR-mutant glioblastoma. Clin. Cancer Res 27, 1553–1569 (2021).[PMC free article] [PubMed] [Google Scholar]

21. Vogelbaum MA, Krivosheya D, Borghei-Razavi H, Sanai N, Weller M, Wick W, Soffietti R, Reardon DA, Aghi MK, Galanis E, Phase 0 and window of opportunity clinical trial design in neuro-oncology: A RANO review. Neuro Oncol. 22, 1568–1579 (2020).[PMC free article] [PubMed] [Google Scholar]

22. Sanai N, Phase 0 clinical trial strategies for the neurosurgical oncologist. Neurosurgery 85, E967–E974 (2019).[PMC free article] [PubMed] [Google Scholar]

23. Jonas O, Landry HM, Fuller JE, Santini JT Jr., Baselga J, Tepper RI, Cima MJ, Langer R, An implantable microdevice to perform high-throughput in vivo drug sensitivity testing in tumors. Sci. Transl. Med 7, 284ra257 (2015). [PMC free article] [PubMed] [Google Scholar]

24. Bhagavatula S, Thompson D, Ahn SW, Upadhyaya K, Lammers A, Deans K, Dominas C, Ferland B, Valvo V, Liu G, A miniaturized platform for multiplexed drug response imaging in live tumors. Cancer 13, 653 (2021). [PMC free article] [PubMed] [Google Scholar]

25. Sanai N, Polley M-Y, McDermott MW, Parsa AT, Berger MS, An extent of resection threshold for newly diagnosed glioblastomas. J. Neurosurg 115, 3–8 (2011). [PubMed] [Google Scholar]

26. Stupp R, Gander M, Leyvraz S, Newlands E, Current and future developments in the use of temozolomide for the treatment of brain tumours. Lancet Oncol. 2, 552–560 (2001). [PubMed] [Google Scholar]

27. Dietlein F, Thelen L, Reinhardt HC, Cancer-specific defects in DNA repair pathways as targets for personalized therapeutic approaches. Trends Genet. 30, 326–339 (2014). [PubMed] [Google Scholar]

28. Nguyen Q-D, Lavdas I, Gubbins J, Smith G, Fortt R, Carroll LS, Graham MA, Aboagye EO, Temporal and spatial evolution of therapy-induced tumor apoptosis detected by caspase-3–selective molecular imaging. Clin. Cancer Res 19, 3914–3924 (2013). [PubMed] [Google Scholar]

29. Ozaki M, Haga S, Ozawa T, In vivo monitoring of liver damage using caspase-3 probe. Theranostics 2, 207–214 (2012).[PMC free article] [PubMed] [Google Scholar]

30. Molinaro AM, Hervey-Jumper S, Morshed RA, Young J, Han SJ, Chunduru P, Zhang Y, Phillips JJ, Shai A, Lafontaine M, Association of maximal extent of resection of contrastenhanced and non–contrast-enhanced tumor with survival within molecular subgroups of patients with newly diagnosed glioblastoma. JAMA Oncol. 6, 495–503 (2020).[PMC free article] [PubMed] [Google Scholar]

31. Yamashita S, Yokogami K, Matsumoto F, Saito K, Mizuguchi A, Ohta H, Takeshima H, MGMT promoter methylation in patients with glioblastoma: Is methylation-sensitive high-resolution melting superior to methylation-sensitive polymerase chain reaction assay? J. Neurosurg 130, 780–788 (2018). [PubMed] [Google Scholar]

32. Naz S, Sowers A, Choudhuri R, Wissler M, Gamson J, Mathias A, Cook JA, Mitchell JB, Abemaciclib, a selective CDK4/6 inhibitor, enhances the radiosensitivity of non–small cell lung cancer in vitro and in vivo. Clin. Cancer Res 24, 3994–4005 (2018).[PMC free article] [PubMed] [Google Scholar]

33. El Botty R, Coussy F, Hatem R, Assayag F, Chateau-Joubert S, Servely J-L, Leboucher S, Fouillade C, Vacher S, Ouine B, Inhibition of mTOR downregulates expression of DNA repair proteins and is highly efficient against BRCA2-mutated breast cancer in combination to PARP inhibition. Oncotarget 9, 29587–29600 (2018).[PMC free article] [PubMed] [Google Scholar]

34. Gallego O, Nonsurgical treatment of recurrent glioblastoma. Current oncology 22, 273–281 (2015). [PMC free article] [PubMed] [Google Scholar]

35. Li H, Stokes W, Chater E, Roy R, De Bruin E, Hu Y, Liu Z, Smit EF, Heynen GJ, Downward J, Decreased glutathione biosynthesis contributes to EGFR T790M-driven erlotinib resistance in non-small cell lung cancer. Cell Discov. 2, 1–14 (2016). [PMC free article] [PubMed] [Google Scholar]

36. Abdelmohsen K, Gerber PA, von Montfort C, Sies H, Klotz L-O, Epidermal growth factor receptor is a common mediator of quinone-induced signaling leading to phosphorylation of connexin-43: Role of glutathione and tyrosine phosphatases. J. Biol. Chem 278, 38360–38367 (2003). [PubMed] [Google Scholar]

37. Roos NJ, Aliu D, Bouitbir J, Krähenbühl S, Lapatinib activates the Kelch-like ECH-associated protein 1-nuclear factor erythroid 2-related factor 2 pathway in HepG2 cells. Front. Pharmacol 11, 944 (2020).[PMC free article] [PubMed] [Google Scholar]

38. Dowless M, Lowery CD, Shackleford T, Renschler M, Stephens J, Flack R, Blosser W, Gupta S, Stewart J, Webster Y, Abemaciclib is active in preclinical models of ewing sarcoma via multipronged regulation of cell cycle, DNA methylation, and interferon pathway signaling. Clin. Cancer Res 24, 6028–6039 (2018).[PMC free article] [PubMed] [Google Scholar]

39. Schaer DA, Beckmann RP, Dempsey JA, Huber L, Forest A, Amaladas N, Li Y, Wang YC, Rasmussen ER, Chin D, The CDK4/6 inhibitor abemaciclib induces a T cell inflamed tumor microenvironment and enhances the efficacy of PD-L1 checkpoint blockade. Cell Rep. 22, 2978–2994 (2018). [PubMed] [Google Scholar]

40. Tatarova Z, Blumberg DC, Korkola JE, Heiser LM, Muschler JL, Schedin PJ, Ahn SW, Mills GB, Coussens LM, Jonas O, Gray JW, Multiple spatial systems analysis of local nanodoses drug responses predicts effective treatment combinations of immunotherapies and targeted agents in mammary carcinoma. biorxiv 2021.09.01.458631 (2021). 10.1101/2021.09.01.458631. [CrossRef] [Google Scholar]

41. Gilad Y, Gellerman G, Lonard DM, malley BWO, Drug combination in cancer treatment—From cocktails to conjugated combinations. Cancer 13, 669 (2021). [PMC free article] [PubMed] [Google Scholar]

42. Jin X, Kim LJ, Wu Q, Wallace LC, Prager BC, Sanvoranart T, Gimple RC, Wang X, Mack SC, Miller TE, Targeting glioma stem cells through combined BMI1 and EZH2 inhibition. Nat. Med 23, 1352–1361 (2017).[PMC free article] [PubMed] [Google Scholar]

43. Narayan RS, Molenaar P, Teng J, Cornelissen FM, Roelofs I, Menezes R, Dik R, Lagerweij T, Broersma Y, Petersen N, A cancer drug atlas enables synergistic targeting of independent drug vulnerabilities. Nat. Commun 11, 2935 (2020).[PMC free article] [PubMed] [Google Scholar]

44. Grossman SA, Ye X, Lesser G, Sloan A, Carraway H, Desideri S, Piantadosi S, Immunosuppression in patients with high-grade gliomas treated with radiation and temozolomide. Clin. Cancer Res 17, 5473–5480 (2011).[PMC free article] [PubMed] [Google Scholar]

45. Litterman AJ, Zellmer DM, Grinnen KL, Hunt MA, Dudek AZ, Salazar AM, Ohlfest JR, Profound impairment of adaptive immune responses by alkylating chemotherapy. The J Imm 190, 6259–6268 (2013). [PMC free article] [PubMed] [Google Scholar]

46. Karachi A, Dastmalchi F, Mitchell DA, Rahman M, Temozolomide for immunomodulation in the treatment of glioblastoma. Neuro Oncol. 20, 1566–1572 (2018).[PMC free article] [PubMed] [Google Scholar]

47. Jordan JT, Sun W, Hussain SF, DeAngulo G, Prabhu SS, Heimberger AB, Preferential migration of regulatory T cells mediated by glioma-secreted chemokines can be blocked with chemotherapy. Cancer Immunol. Immunother 57, 123–131 (2008).[PMC free article] [PubMed] [Google Scholar]

48. Jonas O, Calligaris D, Methuku KR, Poe MM, Francois JP, Tranghese F, Changelian A, Sieghart W, Ernst M, Pomeranz Krummel DA, First in vivo testing of compounds targeting group 3 medulloblastomas using an implantable microdevice as a new paradigm for drug development. J. Biomed. Nanotechnol 12, 1297–1302 (2016).[PMC free article] [PubMed] [Google Scholar]

49. Dominas C, Bhagavatula S, Stover EH, Deans K, Larocca C, Colson YL, Peruzzi P, Kibel AS, Hata N, Tsai LL, The translational and regulatory development of an implantable microdevice for multiple drug sensitivity measurements in cancer patients. IEEE Trans Biomed Eng 69, 412–421 (2022).[PMC free article] [PubMed] [Google Scholar]

50. Dominas C, Deans K, Packard R, Jonas O, Preparation and sterilization of an implantable drug-delivery microdevice for clinical use. MethodsX 8, 101382 (2021).[PMC free article] [PubMed] [Google Scholar]

51. Davidson SM, Jonas O, Keibler MA, Hou HW, Luengo A, Mayers JR, Wyckoff J, Del Rosario AM, Whitman M, Chin CR, Direct evidence for cancer-cell-autonomous extracellular protein catabolism in pancreatic tumors. Nat. Med 23, 235–241 (2017).[Google Scholar]

52. Ahn SW, Ferland B, Jonas OH, An interactive pipeline for quantitative histopathological analysis of spatially defined drug effects in tumors. J Pathol Inform 12, 34 (2021).[PMC free article] [PubMed] [Google Scholar]