Systemic chemokine-modulatory regimen combined with neoadjuvant chemotherapy in patients with triple-negative breast cancer

Abstract

Background

Higher cytotoxic T lymphocyte (CTL) numbers in the tumor microenvironment (TME) predict pathologic complete response (pCR) to neoadjuvant chemotherapy (NAC) and positive long-term outcomes in triple-negative breast cancer (TNBC). pCR to NAC is achieved only in 30–40% of patients. The combination of NAC with pembrolizumab increases the pCR rate but at the cost of immune-related adverse events (irAEs). Based on these considerations, we tested if systemic infusion of the chemokine modulatory regimen (CKM; selective toll-like receptor 3 (TLR3) agonist rintatolimod, interferon (IFN)-α2b, and cyclooxygenase-2 (COX-2) inhibitor celecoxib) regimen can be safely combined with NAC to enhance intratumoral CTL numbers and NAC effectiveness.

Methods

Phase I study NCT04081389 evaluated nine patients with early-stage TNBC who received 3 weeks of paclitaxel with CKM (dose-escalation of IFN-α2b), followed by 9 weeks of paclitaxel alone, dose-dense doxorubicin and cyclophosphamide, and surgery. Primary and secondary endpoints were safety and clinical efficacy, respectively.

Results

The combination treatment was well-tolerated with no dose-limiting toxicities or irAEs. 5/9 patients achieved pCR and one patient had microinvasive disease (ypTmic). We observed elevated IFN signature and uniform decreases in CTL numbers (average 8.3-fold) in the blood of all treated patients. This was accompanied by reciprocal uniform increases in CD8β (overall 5.9-fold), CD8α/FoxP3 (2.11-fold), and CCL5 (4.73-fold) transcripts in TME, particularly pronounced in patients with pCR. Multiplex immunohistochemistry revealed selectively increased numbers of CTL (but not regulatory T cells) in both the epithelial and stromal tumor compartments and early decreases in the numbers of αSMA⁺ vascular/stromal cells in the tumors of all pCR patients.

Conclusions

Combined paclitaxel/CKM regimen was safe, with desirable TME changes and preliminary indications of promising pCR+ypTmic of 66%, comparable to the combination of NAC with pembrolizumab.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• The presence of CD8⁺ cytotoxic T cells (CTL) in the tumor microenvironment (TME) predicts response to both chemotherapy and immune checkpoint inhibition in triple-negative breast cancer (TNBC) and other cancers. The current standard of care in early-stage TNBC, aggressive four-drug chemotherapy (doxorubicin, cyclophosphamide, paclitaxel and carboplatin) with pembrolizumab is associated with significant immune-related adverse events (irAEs). Since paclitaxel activates intratumoral nuclear factor-κB (NF-κB), which can promote tumor survival, but also acts as a co-factor in the induction of CTL-attractants by the chemokine modulatory regimen (CKM) composed of rintatolimod (selective toll-like receptor 3 (TLR3)-ligand), interferon-α2b and celecoxib, we evaluated the safety and preliminary efficacy of systemic CKM combined with paclitaxel-based chemotherapy.

WHAT THIS STUDY ADDS

• Our phase I study in early-stage TNBC assessed the feasibility of including CKM with paclitaxel chemotherapy followed by doxorubicin and cyclophosphamide, before the resection of the residual tumor. We report the safety of this combination (no irAEs) and enhanced CTL infiltration into the TME, with a promising residual cancer burden-0+I rate of 66%.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Our data provide a rationale for the prospective evaluation of systemic CKM as an addition to the current standard of care or as a venue to de-escalate the toxic chemotherapy component.

Background

Triple-negative breast cancer (TNBC) represents 15–18% of all breast cancer (BC) diagnoses. It is associated with a poor prognosis and early recurrences, representing an area of particularly high unmet therapeutic need. About 20% of patients with TNBC are under 40 years of age at the time of diagnosis.¹ Neoadjuvant chemotherapy (NAC) regimen involving either docetaxel/cyclophosphamide for early stage 1 disease or doxorubicin/cyclophosphamide/taxane for higher risk disease has been a long-term standard of care. Data from ECOG 1199, CALGB 9741 and SWOG 0221 indicate that the dose-dense schedule of doxorubicin/cyclophosphamide followed by either dose-dense or weekly paclitaxel is preferred for high-risk disease.2,4

Pathological complete response (pCR) refers to the absence of invasive cancer in the breast and/or axillary lymph nodes. Achieving pCR following NAC is a desirable outcome and a surrogate marker for survival benefit but is achieved only in about 35% of patients.⁵ Recently, KEYNOTE-522 demonstrated that the addition of a checkpoint inhibitor pembrolizumab to taxane-based chemotherapy can improve pCR to 65%.^{6 7} Similarly, improved overall survival (OS) in the metastatic setting observed in KEYNOTE-355 which added pembrolizumab to chemotherapy,⁸ indicates a critical role played by the immune system in improving clinical outcomes.

Retrospective evaluation of two cohorts of patients with neoadjuvant BC (GeparDuo and GeparTrio) showed that the baseline tumor-infiltrating lymphocytes (TILs) in breast tumor microenvironment (TME) predict the success of taxane-anthracycline containing treatment and a higher probability of pCR.⁹ Lymphocyte-rich BC had a pCR of 40%, compared with only 3–7% pCR in the TIL-poor tumors.⁹ A recent pooled analysis showed that higher TILs in early-stage TNBC are associated with improved survival.¹⁰ Detailed analyses indicate a particular role of intratumoral CD8⁺ cytotoxic T lymphocytes (CTLs) in improved disease-free and OS,^{11 12} while local prevalence of FoxP3⁺ regulatory T cells (Tregs) and myeloid-derived suppressor cells (MDSC) were associated with poor prognosis and impaired survival.^{13 14} Therefore, higher ratios of CD8⁺/FoxP3⁺ TILs are associated with an increased probability of pCR.^{15 16} CTL migration depends on two main chemokine receptors, CXCR3 and CCR5.17,20 Previous studies have shown the critical roles of the intratumoral production of ligands for CCR5 (CCL5/RANTES ligand) and CXCR3 (CXCL9/MIG, CXCL10/IP10 and CXCL11/ITAC ligands) in CTL infiltration.18,23 CCL22 is responsible for the attraction of CCR4-expressing FoxP3⁺ Tregs.1924,27

To selectively enhance intratumoral CTL accumulation, without increasing Treg influx, our group has developed the chemokine modulatory regimen (CKM) composed of rintatolimod (toll-like receptor 3 (TLR3) agonist), interferon-α2b (IFN-α2b), and celecoxib (cyclooxygenase-2 (COX-2) inhibitor).^{24 25} This combination selectively induces CTL-attracting chemokines with a concomitant decrease in MDSC and Treg-favoring chemokines, thus increasing the ratio of CD8⁺T/Tregs in the colon, bladder, and prostate TMEs ex vivo,2527,29 with minimal impact on healthy tissues.²⁵ These data suggested the uniqueness of CKM as it can be used as a systemic treatment to induce desirable changes in the TMEs. Our recent clinical trial NCT03599453 demonstrated that in accordance with the preclinical data, systemic CKM resulted in over 10× selective increases in CTL markers in advanced TNBC.³⁰

The components of CKM synergistically induce CTL attractants but suppress Treg attractants.^{24 25} In our recent phase 1 dose-escalation study of local (intraperitoneal) CKM combined with cisplatin in patients with ovarian cancer, we observed local upregulation of IFN-signature, including CTL attracting chemokines in the peritoneal washes, although we could not detect enhanced influx of CTLs themselves, potentially due to their transition to the solid fraction of the tumors, which were not available.³¹

Paclitaxel is known to activate intratumoral nuclear factor-κB (NF-κB),³² the factor which synergizes with the COX2 pathway in the induction of Treg attractants in untreated tumors, but can act as a co-factor in the intratumoral induction of CTL attractants by CKM.^{24 25} These considerations and our ex vivo observations that CKM preferentially induces CTL attractants in NF-κB-rich TME but not healthy tissues,²⁵ suggested that selective intratumoral accumulation of CTLs in patients with paclitaxel-treated TNBC can be achieved by systemic CKM. We hypothesized that a combination of systemic CKM with NAC can selectively enhance intratumoral CTL accumulation and the antitumor efficacy of paclitaxel-based NAC. To test this hypothesis and evaluate the feasibility and safety of adding systemic CKM to chemotherapy, we designed a phase I dose-escalation clinical trial (NCT04081389; see online supplemental file 1) in a neoadjuvant setting in TNBC.

Methods

Study design and participants

We conducted a single-center (Roswell Park Comprehensive Cancer Center) open-label phase I accelerated titration design trial of rintatolimod and celecoxib±IFN-α2b in combination with standard paclitaxel followed by dose-dense doxorubicin/cyclophosphamide chemotherapy in early-stage TNBC at four dose levels (NCT04081389; online supplemental file 1). The first dose level did not have IFN, the second dose level was 5 million units/m² IFN, the third dose level was 10 million units/m², and the fourth dose level was 20 million units/m².

To be eligible for the study, patients had to be female, ≥18 years of age, have a pathologically confirmed diagnosis of resectable TNBC by the College of American Pathologists criteria, have measurable disease, and be eligible for NAC. Patients must also have had an Eastern Cooperative Group performance status of 0–2, adequate organ function/hematologic counts, and an echocardiogram with left ventricular ejection fraction ≥55%. Patients were excluded if they were treated with systemic immunosuppressive agents, including steroids within the last 3 weeks, or if they presented with active autoimmune disease, or if they had a diagnosis of invasive cancer within the last 3 years, or if they were diagnosed with inflammatory breast cancer, or if they had metallic implants not compatible with an MRI, or if they had serious mood disorders (as IFN-α use is associated with depression) or prior cardiac events, such as, acute coronary syndrome, myocardial infarction, or ischemia, or New York Heart Association (NYHA) class III or IV heart failure (due to cardiac toxicity associated with anthracyclines), or if they had a history of upper gastrointestinal ulceration or bleeding, or if they had allergic reaction or hypersensitivity to nonsteroidal anti-inflammatory drugs (NSAIDs), or if they presented with grade 1 or higher neuropathy. Biopsies were successfully obtained from patients at dose level 3 (two patients) and 4 (all three patients).

Procedures

All study procedures were implemented after written informed consent was obtained as mandated by the Institutional Review Board of Roswell Park Comprehensive Cancer Center. Patients underwent screening with pretreatment staging CT scans, breast ultrasonography, mammography, transthoracic echocardiography, and breast MRI following a diagnostic core biopsy confirming invasive TNBC. Clinically suspicious lymph nodes were biopsied prior to initiation of therapy.

Administration of the CKM components

The study treatment consisted of nine doses of rintatolimod, celecoxib±IFN-α2b administered on days 0, 1, 2, 7, 8, 9, 14, 15, 16 concurrently with paclitaxel (80 mg/m²), which was administered intravenously once weekly for 12 weeks. On day 1 of weeks 1–3, patients received paclitaxel and CKM therapy on the same day. First, celecoxib 200 mg was given orally, followed by paclitaxel 80 mg/m² intravenously over 1 hour, with premedications including dexamethasone 12 mg intravenously (subsequently reduced to 4 mg intravenously if patient tolerated first cycle well), diphenhydramine 50 mg intravenously and famotidine 20 mg intravenously. One-hour later, IFN-α2b (dose levels 2–4; over 30 min) followed by rintatolimod 200 mg intravenously (over 2.5 hours) were administered. A second dose of celecoxib was given 12 hours after the initial dose. On days 2 and 3 of weeks 1–3, the participants only received celecoxib, IFN-α2b (dose levels 2–4) and rintatolimod, without paclitaxel. After completion of 3 weeks of combination treatment, weekly paclitaxel was given for an additional nine doses (weeks 4–12). Once the initial 12 weeks of rintatolimod, celecoxib±IFN-α2b plus paclitaxel therapy were completed, the patients received four doses every 2 weeks of intravenously doxorubicin/cyclophosphamide (60/600 mg/m²) with pegfilgrastim growth factor support and usual premedication/antiemetics. Patients with residual disease after undergoing neoadjuvant therapy were allowed to receive adjuvant capecitabine or olaparib therapy at the discretion of their treating oncologist.

Outcomes

The primary endpoints were the safety and toxicity of the study therapy using the Common Terminology Criteria for Adverse Events (CTCAE) V.5.0 and to find the recommended phase II dose (RP2D). Dose-limiting toxicities (DLT) were defined as persistent grade 2 or grade 3 and higher adverse events (AEs) occurring during the first 3 weeks of study treatment leading to treatment delays. Immune-related adverse events (irAEs) were collected until 3 months after completing the combination treatment. The RP2D was the highest dose level where less than one of three patients experienced a DLT. Participants who did not have a DLT and who did not complete six out of nine doses (CKM) of combination treatment (CKM and paclitaxel) were replaced. Secondary endpoints were efficacy (pCR rate; breast MRI response), recurrence-free survival (RFS) and OS. pCR was defined as the absence of invasive cancer in the breast and/or axillary lymph nodes (ypT0/is N0). OS was defined as the time from treatment initiation until death or the last follow-up. RFS was defined as the time from treatment initiation until disease recurrence, death or the last follow-up. Exploratory endpoints included analysis of immune infiltrates, in blood and tumors and comparison of response assessment using Response Evaluation Criteria in Solid Tumors (RECIST) V.1.1 and immune-related RECIST (irRECIST).

Peripheral blood was obtained on day 0 (prior to the start of CKM and paclitaxel) and then 2 hours post completion of CKM infusion on day 0. A fresh biopsy was performed at baseline and after completion of concomitant CKM and paclitaxel on week 3 for patients at higher dose levels. Four tumor cores and two cores of non-tumor surrounding tissue were obtained. Three cores of tumor and one core of non-tumor surrounding tissue were placed in Dulbecco’s phosphate-buffered saline. One core of tumor was processed to formalin-fixed paraffin-embedded (FFPE) block per institutional standards.

Multicolor flow cytometry

Peripheral blood mononuclear cells (PBMCs) were isolated from heparinized blood obtained from patients before and after treatment by centrifugation over the Ficoll-Hypaque density gradient. The cells were then subjected to surface staining with antibodies, APC-H7 anti-CD3, PE-Cy7 anti-CD4, BV510 anti-CD8, BUV395 anti-CXCR3, BV-421 anti-PD-1, PerCP-Cy5.5 anti-CCR4 and BV605 anti-CCR5. Intracellular staining for Alexa-488 Foxp3 and PE GrB was carried out after fixation and permeabilization of cells. The samples were acquired on an LSR II Flow Cytometer and data were analyzed using FlowJo V.10.8 1 software.

Quantitative gene expression

Biopsy strings of tumor or non-tumor tissue were cut into three pieces and placed into Lysing Matrix E Tubes (MP Biologicals) containing RLT buffer (RNeasy Kit; Qiagen) and agitated using an FP120 homogenizer (MP Biologicals). The total RNA was extracted using the RNeasy Kit (Qiagen), 250 ng of RNA were used for complementary DNA (cDNA) synthesis, subsequent cDNA was used to quantify messenger RNA (mRNA) expression. All analysis was performed on the CFX 96 system (Bio-Rad). Commercially available TaqMan primers (Thermo Fisher Scientific Life Technologies) were used to evaluate the local expression of immune cell markers (CD8α, GrB, FoxP3) and key chemokines involved in the attraction of the effector cells (CCL5 and CXCL10), Treg (CCL22) and MDSCs (CXCL12). The expression of each gene was normalized to the HPRT1 housekeeping gene.

Multiplex immunohistochemistry and image analysis

Multiplexed chromogenic immunohistochemical (IHC) assay for high-dimensional tissue analysis called Multiplexed Immunohistochemical Consecutive Staining on Single Slide (MICSSS) was performed. An automated immunostainer (Leica Bond RX, Leica Biosystems) was used to bake slides and perform iterative staining, consisting of chromogenic revelation via 3-amino-9-ethylcarbazole (Vector Laboratories) and counterstaining with hematoxylin like previously described.³³ After each round of staining, slides were removed, mounted with a glycerol-based mounting medium, and scanned to obtain digital images using an Aperio AT2 scanner and Aperio ImageScope DX visualizer software V.12.3.3 (Leica). After scanning, slide coverslips were removed in hot water (~50°C) and tissue sections were bleached. This process is repeated for the length of the panel. Primary antibodies are presented in online supplemental table 1.

For multiplexed imaging, we internally processed our samples to quantitatively analyze localization and coexpression patterns from MICSSS images. The tissue annotation was done by pathologists using QuPath software³⁴ to identify the following four regions: whole tissue (including necrosis, fibrosis, normal tissue); tumor area only; epithelial component of the tumor area; and the stromal component of the tumor area. For the analysis, we used the svs multiresolution, pyramidal images obtained per marker after staining. Each image underwent a brief quality control step to ensure tissue masking appropriately captured the tissue area. Both the AEC chromogen stain and the hematoxylin nuclear counterstain were extracted from each image via a dynamically determined deconvolution matrix. Then, each image was split into smaller tiles to permit computational analysis. Each tile from the first stained image was matched to the respective tile from the sequential stained images and then elastically registered using the extracted hematoxylin nuclear stain and SimpleElastix open-source software. Then, by using an iterative nuclear masking via StarDist, we produced a composite semantic segmentation for nuclei residing in the series of tiles. Each nucleus was artificially expanded by several pixels to simulate a cytoplasm per cell and that was coherent with membrane marker staining. Lastly, cellular resolution metadata was acquired for all cells in the final cell mask, including AEC and hematoxylin intensity properties (percentiles, dynamic ranges, etc) and morphological characteristics (circularity, area, etc) per cell. A final data frame was appended for each tile processed, producing one data frame representing all the cellular metadata per sample. To unbiasedly determine positive and negative cells per marker, we used an unsupervised classification technique to cluster cell populations, followed by a supervised approach where we would evaluate each cluster as positive or negative per marker. First, the metadata aggregated per sample was collected, transformed to z-scores, randomized, and split into subsamples per batch. Each batch was processed in parallel: data for each marker was transformed, clustered, and collapsed into multiple groups by principal components analysis and Uniform Manifold Approximation and Projection. Then, we performed the final quality control to manually attribute which clusters were positive or negative. This produced a final cellular-resolution data frame containing binary marker classification that was used for downstream localization, marker coexpression, tissue annotation and reconciliation and statistical analyses.

For the downstream analysis, samples were quantified within the four aforementioned regions for the following cell compartments: PanCK (PANCK⁺ CD3⁻ CD8⁻ FOXP3⁻ CD20⁻ NKP46⁻ CD68⁻), all T cells (CD3⁺ CD20⁻ NKP46⁻ CD68⁻), CD8 T cells (CD3⁺ CD8⁺ FOXP3⁻ NKP46⁻ CD68⁻), non-CD8 T cells (CD3⁺ CD8⁻ FOXP3⁻ NKP46⁻ CD68⁻), Tregs (CD3⁺ CD8⁻ FOXP3⁺ CD20⁻ NKP46⁻ CD68⁻), NK cells (CD3⁻ CD8⁻ FOXP3⁻ CD20⁻ NKP46⁺ CD68⁻), macrophages (CD3⁻ CD8⁻ FOXP3⁻ CD20⁻ NKP46⁻ CD68⁺), PD-L1 macrophages (CD3⁻ CD8⁻ FOXP3⁻ CD20⁻ NKP46⁻ CD68⁺ PDL1⁺), B cells (CD3⁻ CD8⁻ FOXP3⁻ CD20⁺ NKP46⁻ CD68⁻), stroma and vasculature αSMA (ASMA⁺ CD3⁻ CD8⁻ FOXP3⁻ CD20⁻ NKP46⁻ CD68⁻), COX2 (COX2⁺ CD3⁻ CD8⁻ FOXP3⁻ CD20⁻ NKP46⁻ CD68⁻), and PD-L1 (PDL1⁺ CD3⁻ CD8⁻ FOXP3⁻ CD20⁻ NKP46⁻). Due to low sample numbers, only nominal unadjusted paired t-test values between pre and post were calculated and shown (with significance defined as p<0.05 and trend defined as p<0.1).

Statistical analysis

Sample size justification: The sample size was based on the primary analysis, which evaluates the safety of the treatment combination using an accelerated titration design with four dose levels. Therefore, the observed sample size is dependent on the dose-safety profile, which is unknown a priori. As a post hoc power analysis for the correlative endpoints (immune milieu), assuming within-subject correlations of 0.05–0.65, a sample size of n=9 provides 80% power (at a significance level of 0.05) to detect changes of 1.45–0.9 SDs post-treatment versus pretreatment.

Analytical plan: The patient demographic and clinical characteristics were summarized in the overall sample and by dose level using appropriate descriptive statistics. OS and RFS were summarized using standard Kaplan-Meier methods, where the median and 2-year rates were estimated using 95% CIs. AEs were summarized by attribution and grade using frequencies and relative frequencies. Differences in immune milieu between pretreatment and post-treatment samples were analyzed using a linear mixed model, accounting for within-subject clustering. All analyses were conducted in SAS V.9.4 (Cary, North Carolina, USA) at a significance level of 0.05, where no adjustments were made for multiple testing of correlative endpoints.

RNA sequencing library preparation and sequencing

RNA was extracted using a Qiagen miRNA easy kit following the manufacturer’s recommended protocol. Quantitative assessment of the purified total RNA was then accomplished by using a Qubit Broad Range RNA Kit (Thermo Fisher). The RNA was further evaluated qualitatively by a TapeStation 4200 (Agilent Technologies). The RNA sequencing (RNA-seq) libraries were prepared using the RNA HyperPrep Kit with RiboErase (HMR) kit (Roche Sequencing Solutions), from 100 ng total RNA following the manufacturer’s instructions. Final libraries were validated for appropriate size on a 4,200 TapeStation D1000 ScreenTape (Agilent Technologies), quantitated using a Kapa Biosystems quantitative PCR kit, and pooled together in an equimolar fashion, following experimental design criteria. Each pool was denatured and diluted to 350 pM with 1% PhiX control library added. The resulting pool was then loaded into the appropriate NovaSeq Reagent cartridge for 100 paired-end sequencing and sequenced on a NovaSeq 6000 following the manufacturer’s recommended protocol (Illumina).

Bioinformatic analysis

Paired-end sequencing reads were first pre-processed by FastQC (V.0.11.8)³⁵ for sequencing base quality control. The reads were mapped to the GRCh38 human reference genome and GENCODE (V.38)³⁶ annotation database using STAR (V.2.7.9a).³⁷ Alignment files were indexed using SAMtools (V.1.14).³⁸ A second pass quality control (QC) was done using alignment output with RSeQC (V.4.0.0)³⁹ in order to examine abundances of genomic features, splicing junction saturation, gene-body coverage and rRNA content to check for potential library preparation issues. Gene expression was quantified using FeatureCounts (V.2.0.0)⁴⁰ against corresponding gene annotation obtaining a raw counts matrix.

Differential expression (DE) analyses were performed using DESeq2 (V.1.18.1)⁴¹; DE genes were reported using fold change (FC)>1.2 and adjusted-p value<0.1 (Benjamini-Hochberg’s false discovery rate (FDR) method) cut-off. Raw counts were transformed using a variance stabilization function implemented in DESeq2 package⁴¹ to visualize DE results. Cell-type enrichment analysis was performed on all samples using xCell⁴² inputting transcription per million (TPM) values quantified by RSEM (V.1.3.3).⁴³ Selected immune-cell population enrichment values obtained from xCell output were used to compare between relevant clinical groups. Gene Set Enrichment Analysis (GSEA) (V.7.4)⁴⁴ was used to examine statistically significant enriched pathways (FDR<0.05) against Hallmark and C2cp mSigDB⁴⁵ databases. DESeq2 normalized expression values were used as input to run GSEA.

Data availability

RNA-seq data have been deposited to Gene Expression Omnibus under accession number GSE260989 and can be accessed using the reviewer’s token ulyheocoppmffkt. The data generated in this study are available on request from the corresponding author.

Results

Patient accrual details and demographics

From December 2019 to March 2022, 12 patients were consented. Three patients did not meet eligibility criteria and nine patients were accrued to the study (figure 1A: study schema; figure 1B: Consolidated Standards of Reporting Trials diagram). The study was designed as an accelerated titration study. One patient was accrued at dose level 1 (#1 DL1), one patient at dose level 2 (#2 DL2), four patients at dose level 3 (#3 DL3, #4 DL3, #5 DL3, #6 DL3) and three patients at dose level 4 (#7 DL4, #8 DL4, #9 DL4). One patient (#4 DL3) was non-compliant with the required number of CKM doses (minimum of six doses) and thus, only eight patients were considered evaluable for the primary endpoint of DLTs. Patient demographics are shown in table 1.

Figure 1. Study design. (A) Study schema. Systemic chemokine modulation (CKM; 30 min-long IFN-α2b infusion followed by 2.5-hour-long infusion of rintatolimod) was given as nine daily infusions (3 days per week) over 3 weeks (days 0, 1, 2, days 7, 8, 9, and days 14, 15, 16) along with weekly paclitaxel 80 mg/m². CKM consisted of intravenous rintatolimod (200 mg) and oral celecoxib (two doses of 200 mg)±IFN-⍺2b at dose-escalation from 0 (no IFN-⍺2b; dose level 1) through 5 MU/m² (dose level 2) and 10 MU/m² (dose level 3) to 20 MU/m² (dose level 4). Stars represent the timing of tumor biopsies performed before and after the CKM at the higher dose levels. (B) Consolidated Standards of Reporting Trials flow diagram. This depicts screening, enrollment and follow-up of participants in the trial. The trial enrolled nine patients, eight of whom were evaluable for the primary endpoint of safety. IFN, interferon.

Table 1. Baseline characteristics of the study population.

	N (%)
Gender
Females	9 (100)
Age
Median	47 (range 37–55)
Race
White	8 (89)
Hispanic	1 (11)
Stage
I	2 (22)
IIA	6 (67)
IIB	1 (11)
Lymph node positive	2 (22)
Tumor
T1	3 (33)
T2	6 (67)
Grade
III	9 (100)
Histology
Invasive ductal carcinoma	9 (100)
Dose level
1	1 (11)
2	1 (11)
3	4 (44)
4	3 (33)

Clinical and demographic characteristics of the nine participants on the study.

Safety of CKM combination with weekly paclitaxel

The data cut-off for the safety and efficacy analyses was February 16, 2024. There were no AEs that qualified as DLT or irAE at any of the dose levels tested. Apart from patient #4 DL3 (who forgot to take celecoxib), all other patients completed at least 6 doses of IFN-α2b and rintatolimod and 12 doses of celecoxib taken twice a day (overall 6 doses of CKM) and were considered evaluable for the primary endpoint (online supplemental figure S1). A summary of AEs is presented in online supplemental table 2. The most common worst-grade AEs, which were possibly, probably, or definitely attributed to study therapy (CKM, paclitaxel, doxorubicin and cyclophosphamide) are listed in table 2 (CKM), (online supplemental figure 2A) (paclitaxel), and online supplemental figure 2B) (doxorubicin or cyclophosphamide), respectively.

Table 2. Treatment-related adverse events (TRAEs) attributed to components of chemokine modulaton regimen (CKM).

System organ class	Interferon-⍺2b-related TRAE	Grade 1	Grade 2	Grade 3
Gastrointestinal disorders	Nausea	1	0	0
General disorders	Chills	1	0	0
	Fatigue	4	0	0
	Fever	1	1	0
Investigations	ALT increased	0	1	0
	AST increased	1	0	0
	Neutropenia	0	0	1
Musculoskeletal and connective tissue	Generalized muscle weakness	1	0	0
	Myalgia	1	0	0
Nervous system disorders	Headache	1	0	0
System organ class	Rintatolimod-related TRAE	Grade 1	Grade 2	Grade 3
Gastrointestinal disorders	Constipation	1	0	0
	Nausea	2	0	0
General disorders	Chills	2	0	0
	Fatigue	4	0	0
Musculoskeletal and connective tissue	Myalgia	1	0	0
Nervous system disorders	Dizziness	1	0	0
Vascular disorders	Hypotension	1	0	0
System organ class	Celecoxib	Grade 1	Grade 2	Grade 3
None

Any grade adverse events attributable to CKM are reported.

ALTalanine aminotransferaseASTaspartate aminotransferase

The most frequent AEs attributed to CKM were nausea, chills, fatigue, fever, alanine transaminase/aspartate transaminase increase, myalgia, dizziness, headache, constipation, and hypotension, which were grade 1 or 2. Grade 3 treatment-related adverse events included neutropenia (4/9 patients), attributed to CKM (1/9 patients) or paclitaxel (3/9 patients) or doxorubicin/cyclophosphamide (1/9 patients), pneumonia (1/9 patients) and anemia (1/9 patients) attributed to doxorubicin/cyclophosphamide. The only serious AE in the study was pneumonia (2/9 patients), one attributed to doxorubicin/cyclophosphamide (1/9 patients), and the other one was a grade 3 event, unrelated to study treatment. Both the events of pneumonia were managed with antibiotics.

Clinical outcomes in patients receiving CKM/NAC regimen: Rates of pCR, microinvasive disease (ypTmic) and interim breast MRI responses

The pCR rate (ypT0/is N0) was 55% (5/9 patients), which included wo of the two node-positive patients. Among the remaining four patients who did not achieve a pCR, two (patients #3 DL3, #8 DL4) were classified as residual cancer burden-I (RCB-I), 1 of which was ypTmic. The other two patients (#1 DL1, #4 DL3) were classified as RCB-II. Three patients received adjuvant capecitabine and one patient with germline Breast Cancer gene (BRCA) mutation received adjuvant olaparib. Figure 2 shows the waterfall plot (left) with the per cent change in tumor size by breast MRI from baseline to pre-surgery and the swimmer’s plot (right) with dose levels, clinical stage of the tumors, RCB status, RFS and OS.

Figure 2. Radiologic and pathologic responses post neoadjuvant treatment. Breast MRI responses (left) after completion of neoadjuvant treatment were measured by RECIST V.1.1 and irRECIST for the nine patients on dose levels (DL) 1–4. Waterfall plot shows the % changes in tumor size from baseline to pre-surgery. The clinical stage of the tumors is depicted using AJCC staging. The pathological stage (right) is reported as RCB-0, RCB-I and RCB-II. As of the data cut-off of February 16, 2024, two patients on follow-up progressed and developed metastatic disease and died. The table lists the doses of celecoxib, IFN-α2b, rintatolimod at each DL and occurrences of grade 3 or higher treatment-related adverse events (TRAEs), attributed to paclitaxel, CKM or doxorubicin and cyclophosphamide (AC). Abbreviations: AC, doxorubicin (Adriamycin) and cyclophosphamide; AJCC, American Joint Committee on Cancer; CKM, chemokine modulatory regimen (rintatolimod, IFN-α2b, celecoxib); IFN, interferon; irRECIST, immune related Response Evaluation Criteria in Solid Tumours; RCB, residual cancer burden; RECIST, Response Evaluation Criteria in Solid Tumors.

With a median follow-up time of 29.9 months (range: 21.6–43.9 months), there were two distant BC recurrences, which both resulted in BC-related mortality. There were no recurrence events in RCB 0–1 patients. Median RFS was not reached (NR; 95% CI, 15.0 months to NR) with a 2-year rate of 75%. Median OS was also NR (95% CI, 21.7 months to NR) with a 2-year rate of 75%.

The breast MRI response was evaluated after completion of neoadjuvant CKM and paclitaxel and pre-surgery using RECIST V.1.1 and irRECIST. There was concordance in measurements by RECIST and irRECIST (although only one MRI scan is available for irRECIST measurement, and a follow-up repeat MRI was not performed). Among the 7/9 patients who underwent breast MRI after CKM and paclitaxel, one attained radiographic complete response (CR), one had partial response, four had stable disease, while one experienced progressive disease. Thus, the interim breast MRI response was 28.6% (online supplemental figure 3).

Systemic CKM and paclitaxel induced transient decreases in circulating CTLs. Blood samples were obtained pretreatment (day 0) and on the same day post-treatment after completion of CKM and paclitaxel. Similar to the observations from our prior TNBC clinical trial (NCT03599453), which included no chemotherapy component,³⁰ longitudinal flow cytometry analysis of the circulating immune cells demonstrated rapid decreases in the numbers of circulating effector-type CD3⁺CD4⁻CD8⁺ CTLs and of CD3⁺CD4⁻CD8⁺GrB⁺ CTLs. These decreases were statistically significant both in patients with pCR (p<0.001 and p=0.002 for each cell type) and in patients with residual disease (respectively, p=0.004 and p=0.011) (figure 3A). These results were mirrored by real-time quantitative polymerase chain reaction (RT-qPCR), which showed consistent decreases in CD8α (to an average of 0.12 and median of 0.16 of baseline; p=0.004) and CD8β (average of 0.25 and median of 0.19 of baseline; p=0.005). Interestingly, there was no difference in GrB expression (p=0.108) (figure 3B), which may suggest a compensatory (GrB-activating) effect of CKM on the remaining CTLs.

Figure 3. Decreased CD8⁺T cell counts in the blood of patients directly after CKM with paclitaxel treatment. Acute changes in the immune cell subset composition of peripheral blood were measured at the end (same day) of treatment with CKM and paclitaxel. (A) Multiparameter flow cytometry analysis of circulating lymphocytes (N=8) for patients (PTS) 2–9. The blue lines represent changes in blood in patients who attained a pCR and the red lines represent blood changes in patients with non-pCR. A statistically significant decrease in CD8⁺ T cells and GrB⁺ CD8⁺ T cells was observed. (B) Expression of cytotoxic T lymphocyte markers (CD8α and CD8β) and their effector status (GrB) was measured using RT-PCR for PTS 3–9 (N=7). Data is expressed as ratios of the individual markers to the housekeeping gene, HPRT1. Decreases in both CD8α and CDβ transcripts are observed with no change in GrB transcript measured by RT-PCR (N=7). (C) Short-term changes in the immune signature in peripheral blood mononuclear cells induced by CKM and paclitaxel. Note the changes in chemokines, chemokine receptors, IFN-inducible and IFN-regulatory genes which are upregulated post-treatment (N=7) for PTS 3–9 (>1.2-fold change, FDR<0.1). (D) Combination of CKM and paclitaxel downregulates expression of the Wnt family member 7A, granzyme M, granzyme K, CXCL8, CXCL2, CXCR3, CD8A and CD8B (N=7) for PTS 3–9 (>1.2-fold change, FDR<0.1). (E) xCell immune cell average enrichment scores for the various immune cell subtypes across the pretreatment and same day post-treatment points for PTS 3–9 in the pCR and non-pCR groups. The left panel legend shows the cell subtypes in the order they appear in each stacked bar from left to right. The right panel shows post-treatment decreases in the signatures of the CD4⁺ and CD8⁺ T cells, and a concomitant increase in B cells and macrophage signatures, consistent with their preferential retention within the circulation. CKM, chemokine modulatory regimen; FDR; false discovery rate; IFN, interferon; pCR, pathological complete response RT-qPCR, real-time quantitative polymerase chain reaction.

We also observed statistically significant decreases in the number of circulating total CD4⁺ T cells. Unexpectedly, this same trend was also observed in total FoxP3⁺ cells and CXCR4⁺ FoxP3⁺ cells in the circulation, among both patients with pCR and those without (online supplemental figure 4A), which may suggest a direct Treg-reprogramming effect of CKM. These results were mirrored by RT-PCR (online supplemental figure 4B) which revealed decreases in total circulating CD4⁺ T cells (to an average of 0.54/median of 0.32 of baseline; p=0.026), which was statistically significant only among the patients with pCR (average 0.36/median 0.36 of baseline, p=0.038). These changes resulted in a significant decrease in the ratios of CD8α/FoxP3 (average 0.11/median 0.05 of baseline; p=0.014), and CD8β/FoxP3 (average 0.18/median 0.16; p=0.002). We also observed an average 15.9-fold/median=13.8-fold increase in programmed death-ligand 1 (PD-L1) (p<0.001) and an average 7.04-fold/median=7.25-fold increase in PD-L2 (p<0.001) in the circulation among all patients, suggestive of upregulation of PD-L1 on myeloid cells. Unexpectedly, there was also a decrease in programmed cell death protein-1 (PD-1) in the PBMCs of patients with non-pCR (average 0.35/median 0.39 of baseline, p=0.012).

CKM and paclitaxel induce differential changes in circulating immune markers. The post-treatment blood samples showed higher expression of immune genes, such as, IFN-stimulated genes; chemokines CXCL10, CXCL11, CXCL9, and CCL5; chemokine receptor XCR1; IFN inducible genes, IFN regulatory factors, signal transducer and activator of transcription (STAT), and the transmembrane genes, as determined by bulk RNA-seq (figure 3C).

Consistent with the results of the RT-PCR and flow cytometry, we observed downregulation of Wnt family member 7A, granzyme M, granzyme K, CXCL8, CXCL2, CXCR3, CD8A and CD8B (figure 3D), consistent with the efflux of CXCR3⁺CTLs from the circulation to the tumor. Similarly, GSEA showed upregulation of IFN-α, IFN-γ and inflammatory response pathways (online supplemental figure 5A). Immune deconvolution analysis using xCell showed a decrease in CD4⁺ T cells, CD8⁺ T cells, but a concomitant expansion of B cells and M1 and M2 macrophages in post-treatment versus pretreatment samples (figure 3E).

Gene expression only elevated in post-treatment blood samples of pCR patients was analyzed. This included elevated expression of XCR1, CCL5, CCR7, CCR2, TNFRSF17, IL-6, IL-6-AS1, IL-2RA, LTF, and immunoglobulin heavy and light chains (online supplemental figure 5B), raising the possibility that CKM-driven immune patterns can act as early predictors of pCR. In contrast, we could not identify any pretreatment predictors of pCR.

Online supplemental figure 5C shows the differential post-treatment expression of genes in patients who subsequently attained pCR versus non-pCR. We observed higher expression of immunoglobulin genes, IGLV8-61, IGHG3 and IGLV6-57, IL-12RB2 and XCR1 (expressed on cDC1 subset of dendritic cells involved in antigen presentation⁴⁶) in patients who subsequently attained pCR, providing additional indication that CKM/paclitaxel-induced immune signature may be used as a pCR predictor. Similarly, GSEA showed upregulation in cell proliferation (E2F targets, G2M checkpoint, mitotic spindle, MYC targets) and immune signaling (IFN-γ response, interleukin (IL-6)-Janus kinase-STAT 3) pathways, combined with the downregulation of angiogenesis-signaling, hypoxia-signaling and tumor necrosis factor (TNF)-signaling pathways, in the post-treatment blood samples of pCR patients, compared with non-pCR patients.

Improved chemokine production patterns and CTL markers in the TME of patients receiving systemic CKM and paclitaxel

Since CD8⁺ T-cell infiltration is a prognostic marker that predicts pCR and improved long-term outcomes in patients with TNBC, our exploratory analyses involved treatment-associated changes in intratumoral CTL markers in the total biopsy volume. Paired pre-CKM and post-CKM/paclitaxel biopsies (weeks 0 vs 3) were successfully obtained in four patients with pCR: patients (#5 DL3, #6 DL3, #7 DL4, #9 DL4), but only in one patient with non-pCR (#8 DL4), limiting our ability to interpret the non-pCR data (figure 4A). Comparison of the pre/post CKM/paclitaxel mRNAs showed a trend towards an overall increase in CD8α (average 6.88-fold/median 2.86-fold; p=0.312) and an average 5.9-fold/median 5.39-fold increase in CD8β (p=0.094). The increase in CD8β was statistically significant in the patients with pCR (average 7.07-fold/median 5.67-fold increase; p=0.016); with a trend towards CD8α increase in the same subset of patients (8.3-fold average/3.03-fold median increase, p=0.096). We further noted significant increases in the ratios of CD8α/FoxP3 (average 2.11-fold/median 2.15-fold; p=0.018), which was particularly significant in patients with pCR (average 2.18-fold increase and median 2.21-fold increase, p=0.009).

Figure 4. Intratumoral increases in cytotoxic T lymphocyte markers after systemic CKM and paclitaxel. (A) Intratumoral levels of CD8⍺, CD8β and chemokine expression (normalized for HPRT1) were measured using RT-PCR in tumor biopsies obtained at baseline and at 3 weeks of CKM/paclitaxel regimen (see figure 1 for study schema) for patients on the higher CKM dose levels (PTS 5–9). The blue lines represent patients who attained a pCR (N=4) and the red line represents the non-pCR patient (N=1; dose level 4). Statistically significant increases in CD8β, CD8α/FoxP3, CCL5 and CXCL12 mRNA were observed. (B) Gene Set Enrichment Analysis (GSEA) expression level changes (normalized enrichment scores (NES)) for PTS 5–9 identified multiple immune signaling groups which increased significantly post CKM/paclitaxel, while proliferation gene families were significantly decreased (FDR<0.05). CKM, chemokine modulatory regimen; FDR; false discovery rate; pCR, pathological complete response; PTS, patients; RECIST, Response Evaluation Criteria in Solid Tumors.

To gain insight into the mechanism of the CKM-driven enhancement of the CTL signature, we analyzed changes in the expression of CD8⁺T cell-attracting and Treg-attracting chemokines (figure 4A). The average intratumoral expression of CCL5 mRNA (chemokine binding to CTL-expressed CCR5) showed a trend to post-treatment increase overall (average 4.73-fold/median=2.88-fold; p=0.099) but was statistically significant in the patients with pCR (average 5.49-fold/median 4.06-fold; p=0.030).

Unexpectedly, we also observed a 5.23-fold average/median 4.45-fold increase in the undesirable chemokine CXCL12 (p=0.021), particularly significant in patients with pCR (average 6.06-fold/median 4.93-fold; p=0.005; figure 4A). We also observed trends towards increased ratios of CD8α/CD4 (average 2.67-fold/median 2.31-fold; p=0.054) and CCL5/CCL22 (average 5.09-fold/median 2.85-fold; p=0.058) selectively among tumors with pCR (online supplemental figure 6). GSEA (figure 4B) revealed elevated immune signaling pathways associated with allograft rejection, IFN-α-response, IFN-γ-response, complement, IL-6-Janus kinase-STAT 3, IL-2-STAT 5, and general inflammatory signature in the post-treatment TMEs. Reduced proliferation-related genes (E2F targets, G2M checkpoints, MYC targets, mammalian target of rapamycin complex1, glycolysis and cholesterol hemostasis) were also noted, interestingly with reduced early and late estrogen response signature (figure 4B).

We used multiplex IHC (figure 5) for spatial profiling, to compare the intratumoral distribution and frequencies of immune cells before and after the CKM/paclitaxel regimen. Figure 5A–B show representative pre/post-treatment tumor samples and trends towards the increased proximity of CD8⁺ T cells to PanCK⁺ cancer cells post-treatment, preferentially seen in the patients who subsequently attained pCR. Cell enumeration (figure 5C) demonstrated a post CKM/paclitaxel overall increase in intratumoral T cells (p=0.028) being particularly pronounced in tumors from patients who achieved pCR (p=0.009). This increase included both CD8⁺ T cells and CD4⁺ T cells, but not Tregs. The patients with pCR also demonstrated strong decreases in stromal and vascular marker, α-smooth muscle actin (αSMA) expression in whole tumor tissue and in their stromal areas (respectively, p=0.006 and p=0.02; figure 5C). αSMA is a marker of epithelial to mesenchymal transition and an increase in αSMA expression in the tumor with non-pCR while decrease post-treatment in tumors with pCR is consistent with the role of αSMA positive cells in promoting tumor growth.⁴⁷ No significant changes in other markers (PD-L1, PD-1, FOXP3, COX2, CD68, CD20) were observed (data not shown).

Figure 5. CKM and paclitaxel treatment decreases stromal vascular markers while increasing cytotoxic T lymphocytes in the tumors. MICSSS was performed for high-dimensional tissue analysis on tumors at baseline and at 3 weeks at the completion of CKM and paclitaxel. The whole tumor area (excluding normal, necrosis or fibrosis), tumor nests (epithelial) and stromal elements were analyzed. (A) Pseudo-immunofluorescence images from MICSSS used to visualize each marker, are shown for a representative patient with pCR (PT 5) and non-pCR (PT 8), before and after the CKM/paclitaxel. Note that the patient with pCR, but not the non-pCR patient, showed post-treatment CD3 and CD8 infiltration, and decreases in the PanCK-positive and α-smooth muscle actin (α-SMA)-positive cells. Each ROI) is 500 μm × 400 μm. (B) The blue lines represent patients who attained a pCR (N=4) and the red line represents the non-pCR patient (N=1; dose level 4). Treatment with CKM and paclitaxel induces a trend towards closer apposition of CD8⁺T cells to the nearest PanCK⁺ cancer cells in patients with pCR (patients 5, 6, 7, 9) but not in the single patient with non-pCR (PT 8). (C) The blue lines represent patients who attained a pCR (N=4) and the red line represents the non-pCR patient (N=1). There were consistent post-treatment increases in total T cells, CD8⁺ T cells, and CD8⁻ T cells (identified as CD4⁺ T cells) but no changes in regulatory T cells (Tregs) in the epithelial and stromal/vascular areas of the tumors and associated decreases in the prevalence of the cells expressing αSMA, selectively affecting patients who subsequently attained pCR. CKM, chemokine modulation; Multiplexed Immunohistochemical Consecutive Staining on Single Slide, MICSSS; NK, natural killer; pCR, pathological complete response; PT, patient; ROI, region of interest.

Discussion

Our study demonstrates the safety of combining paclitaxel-based NAC of TNBC with systemic CKM and provides a preliminary indication of the immunologic and clinical effectiveness of this combination. The safety data on the combination therapy suggest that CKM can be administered at the highest dose, concurrently with neoadjuvant weekly paclitaxel chemotherapy. The main AEs that were significantly different from those associated with chemotherapy alone were related to immune activation during treatment with CKM (fevers, chills, nausea, fatigue, myalgias, headache, dizziness). However, these AEs were transient and easily managed with antipyretics. Due to the limitation of the accelerated titration design and the small sample size, in subsequent phase II trials, we may include safety lead-ins or Bayesian safety monitoring to further explore the safety profile of this regimen.

Our preliminary efficacy results suggest that adding CKM to NAC may increase the probability of achieving a pCR (55%), compared with the expected pCR with standard anthracycline/taxane chemotherapy (30–40%), and a comparable pCR rate to the KEYNOTE-522 regimen. Interestingly, both patients with lymph-node-positive disease achieved pCR, showing an early promise of this combination. The probability of achieving pCR was higher at the higher IFN-α dose levels, although the small sample size of the study along with its accelerated titration design gives this notion a very preliminary character.

The increases in CTLs in the TME of all patients who subsequently attained pCR and a concomitant decrease in the circulation suggests that the CKM/paclitaxel combination triggers an inflammatory reaction in the tumor and CTL transition from blood to the TME. Among patients with pCR, there was an upregulation of immunoglobulin genes in the circulation compared with patients with residual disease, which is consistent with an increase in B cells in the circulation post-treatment with CKM/paclitaxel combination, indicative of B-cell retention in the blood. Similarly, treatment with CKM/paclitaxel also increased XCR1, the chemokine receptor expressed on cDC1s,⁴⁸ selectively in patients who attained pCR. Since cDC1s are key to cross-presenting tumor-derived antigens to T cells,⁴⁸ the current data suggest that activation of cDC1s, in addition to the attraction of immune effector cells, may contribute to the CKM-driven activation of CTLs and antitumor effects. In addition, the treatment also enhanced IFN-α and IFN-γ signature in the TME, along with downregulation of MYC targets, G2M checkpoint and E2F targets; these changes were associated with a higher likelihood of achieving pCR. Availability of only one pair of matched tumor samples from non-pCR patient limits our ability to interpret these findings, but all patients with pCR, demonstrated uniform increases in T-cell recruitment to the TME.

Unexpectedly, in addition to all the desirable changes, we also observed the CXCL12 increases in the TMEs of the paclitaxel/CKM-treated patients, which is in contrast to our observations in a prior study where patients with advanced TNBC received CKM alone.³⁰ Both paclitaxel itself⁴⁹ and PGE2, inducible in the TME by activated CTLs and natural killer cells⁵⁰ or paclitaxel-base chemotherapy,⁵¹ have been shown to increase circulating CXCL12 in preclinical models and patients with cancer.49,51 Since CXCL12 is an undesirable Treg-and MDSC-attracting chemokine, inducible in an NF-kB-dependent and PGE₂-dependent manner,^{24 52} this unexpected effect suggests the ability of paclitaxel to activate this pathway as well as opportunities to enhance the clinical efficacy of the CKM/NAC regimen using additional blockers of PGE₂ pathway or using inhibitors of CXCL12 and CXCR4 interaction, such as AMD 3100.

Our study also showed that several immune-related gene signatures were upregulated in the blood in patients whose tumors achieved pCR, while not being upregulated in non-pCR, suggesting the feasibility of developing on-treatment biomarkers of effective anticancer immunity as a first step towards designing patient-response driven therapy, in order to reduce overall morbidity/mortality.

At the time when the trial was designed, there was limited data regarding the impact of adding carboplatin or immunotherapeutic agents, such as pembrolizumab. The addition of carboplatin increases the pCR rate to a level similar to that observed in our current cohort, and recent reports have shown that it also improves RFS and OS, although this has been limited to the group of patients less than 50 years old.⁵³ Recent data from I-SPY2 and KEYNOTE-522 show an encouraging increase in the pCR rate up to 65%, with early survival data demonstrating a favorable response with the addition of pembrolizumab.^{6 54} Our study population was similar to that of KEYNOTE-522 except that there were 2/9 stage I patients in our study. Adding CKM to immune checkpoint inhibitors in the neoadjuvant setting may improve the pCR rate further, by altering the TME.

Another interesting implication of the current study which deserves to be evaluated further in a larger follow-up trial is that radiographic response seen in 2/7 patients after neoadjuvant therapy with CKM and paclitaxel may be used as a basis to allow such patients to undergo an earlier surgery, without receiving subsequent doxorubicin and cyclophosphamide, which constitute the most toxic component of NAC. Our planned phase II trial will evaluate the combination of CKM, NAC (docetaxel and carboplatin) and pembrolizumab followed by an interim MRI and possible omission of doxorubicin and cyclophosphamide if radiographic CR on interim MRI. The chemotherapy backbone of carboplatin and docetaxel in the planned trial would be similar to the ongoing phase 3 SCARLET trial investigating this chemotherapy combination with pembrolizumab (NCT05929768) and comparing it head-to-head with the KEYNOTE 522 regimen. The design of such studies will involve the evaluation of the CKM/paclitaxel/pembrolizumab-induced molecular signatures in the blood and TME of patients who achieve pCR, to validate their ability to predict outcomes and help guide treatment decisions.

Major limitations of our study include its small sample size and multiple comparisons within its exploratory endpoints. Other limitations include the availability of biopsy material from only one patient with non-pCR, the presence of stage I disease in two of nine studied patients (the pCR rate in this group is already high) and the single-arm character of the study, highlighting the need for larger confirmatory studies. Within these limitations, our study provides early indications of the safety of CKM/NAC regimen and its ability to drive immunologic and clinical effects. We conclude that these initial results warrant phase II trials to test if CKM is clinically effective in high-risk early-stage TNBC and if it can be used to reduce the intensity and toxicity of the current treatments. Our additional studies will also test the applicability of the current findings to other groups of patients receiving taxanes.

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information.