Dual PD-1/LAG-3 Blockade in Breast Cancer: Navigating Benefit and Risk

Abstract

The addition of immune checkpoint inhibitors to neoadjuvant chemotherapy has reshaped the treatment landscape for early-stage triple-negative breast cancer (TNBC). However, for hormone receptor-positive, HER2-negative (HR⁺/HER2⁻) breast cancer, the most common subtype, the benefits of immunotherapy have remained limited. The I-SPY2 trial, recently published in JAMA Oncology, evaluated the combination of the anti-PD-1 antibody cemiplimab and the anti-LAG-3 antibody fianlimab (the PCF regimen) in early-stage high-risk HER2-negative breast cancer [1]. The results showed remarkable efficacy alongside safety concerns that warrant careful consideration.

In this adaptively randomized phase 2 trial, 76 evaluable patients were stratified by standard biomarker subtypes, including 40 with HR⁺/HER2⁻ disease and 36 with TNBC. 350 patients served as historical controls. The primary efficacy endpoint was the pathologic complete response (pCR) rate after completion of neoadjuvant therapy. Overall, the estimated pCR rate of the PCF regimen reached 44% (95% confidence interval [CI], 34%–53%), more than doubling the 21% (95% CI, 17%–25%) observed in the historical control arm, meeting I-SPY2′s prespecified graduation criteria across the evaluated biomarker signatures. Notably, in HR⁺ /HER2⁻ disease, a subtype traditionally considered immunologically “cold,” the pCR rate increased from 14% (95% CI, 9%–19%) to 36% (95% CI, 23%–49%). These foundings are encouraging but should be interpreted cautiously. The subtype analyses exploratory without multiplicity adjustment, and I-SPY2 was not designed for formal interaction testing. Given the broad confidence intervals, confirmation of these benefits in adequately powered trials is required.

The biomarker findings were notable. Among HR⁺ /HER2⁻ patients with ImPrint-positive tumors, the pCR rate reached 91%, compared with only 28% among those with ImPrint-negative tumors. In TNBC, the corresponding rates were 83% and 28%, respectively [1, 2]. Collectively, these data suggest that dual PD-1/LAG-3 blockade has potent biological activity in breast cancer, although prospective validation is still required. The biological rationale for this approach is illustrated in Figure 1. PD-1 and LAG-3 are non-redundant inhibitory checkpoints frequently co-expressed on exhausted CD8⁺ tumor-infiltrating lymphocytes (TILs). Co-blockade relieves two independent brakes, promoting stem-like CD8⁺ T cell expansion and restoring antitumor cytotoxicity.

Beyond breast cancer, the PD-1/LAG-3 combination has shown promise across multiple tumor types. In melanoma, the fixed-dose combination of relatlimab and nivolumab has been approved, establishing the clinical value of this dual blockade strategy [3]. In recurrent glioblastoma, a phase 1 trial demonstrated a 12-month overall survival rate of 52.2% with combination therapy, compared with 34.8% with relatlimab monotherapy [4]. In alveolar soft part sarcoma, dual blockade with sintilimab and IBI110 has also shown durable antitumor activity [5]. These cross-tumor data support the biological rationale for PD-1/LAG-3 co-blockade and highlight its broad therapeutic potential.

However, these efficacy gains were accompanied by substantial safety concerns. Adrenal insufficiency occurred in 21% of patients receiving the PCF regimen, with 11% experiencing grade 3 or 4 hypophysitis. Notably, most cases developed after immunotherapy completion, reflecting the delayed onset of immune-related toxicity and necessitating prolonged endocrine surveillance. In the I-SPY2 platform, adrenal insufficiency was defined as morning serum cortisol level < 3 mcg/dl in the absence of recent glucocorticoid exposure and occurred more frequently with dual immune modulation than with single-agent immune checkpoint inhibitor therapy (18.1% vs 8.5%). More than 20% of cases were diagnosed after treatment completion, and all affected patients remained on glucocorticoid replacement at last follow-up [6]. These finding underscore the potential for delayed and persistent endocrine toxicity, although long-term reversibility remains uncertain. In the curative-intent setting of early-stage breast cancer, where patients anticipate decades of disease-free survival, the risk of persistent adrenal insufficiency represents a severe, lifelong medical burden [1]. The toxicity profile limits the clinical feasibility of the regimen, and this specific combination will not be advanced in its current form. Baseline endocrine evaluation, serial monitoring during and after immunotherapy, and clear pathways for long-term glucocorticoid management should therefore be integral to future risk–benefit assessment.

These findings should also be interpreted in the context of the historical control arm design and current standards of care. The I-SPY2 platform used historical chemotherapy control arms that lacked carboplatin and pembrolizumab, agents that now constitute standard neoadjuvant therapy for TNBC. In KEYNOTE-522, pembrolizumab plus chemotherapy achieved a pCR rate of 64.8% (95% CI, 59.9%–69.5%), compared with 51.2% (95% CI, 44.1%–58.3%) for chemotherapy alone [7]. Thus, although the 53% pCR rate with PCF in TNBC (95% CI, 39%–67%) represented a substantial improvement over historical chemotherapy (29%; 95% CI, 22%–36%), its pCR rae was numberically lower than that reported with pembrolizumab plus chemotherapy in KEYNOTE-522, although cross-trial differences preclude direct dfficacy comparisons. In contrast, the principal therapeutic relevance of PCF may lies in HR⁺/HER2⁻ disease, where effective immunotherapeutic strategies remain an unmet need [2].

The PCF trial is not the first dual immunotherapy regimen to encounter this dilemma in breast cancer. The BELLINI trial evaluated neoadjuvant nivolumab plus ipilimumab (anti-PD-1 + anti-CTLA-4) in early-stage TNBC, similarly observing immune activation alongside substantial toxicity [8]. Together, these findings highlight a recurringefficacy-toxicity challenge with intensified dual immune checkpoint blockade (dual-IO) in breast cancer. In the curative-intent setting, the potential for enhanced antitumor activity must therefore be weighed carefully against immune-related toxicity. Whether particular endocrine organs are especially susceptible to dual-IO, and whether this susceptibility is influenced by the hormonal-immune microenvironment of breast cancer,warrants dedicated translational investigation.

ImPrint, a 53-gene expression signature, achieved an odds ratio (OR) of 14.5 for predicting pCR across five immunotherapy arms of I-SPY2, substantially higher than that of conventional biomarkers such as tumor grade III, MammaPrint ultra-high risk, and ER-low status (all OR < 7.5). Approximately 51% of TNBC and 26% of HR⁺/HER2⁻ tumors were ImPrint-positive, with sensitivity of 83% and specificity of 89% in HR⁺/HER2⁻ patients. However, because these findings are exploratory and derive from relatively small subgroups across multiple immunotherapy arms, they should not yet be regarded as clinically actionable. Importantly, approximately 73% of HR⁺/HER2⁻ tumors in the PCF group were ImPrint-negative, yet still had a 28% pCR rate, indicating that ImPrint negativity does not equate to an absense of benefit.

Unlike the PD-L1 combined positive score (CPS), which is derived from a single-protein immunohistochemical assay, ImPrint captures the coordinated expression of a 53-gene immune program, more than 90% of which are immune-related. This broader functional readout may provide complementary information on the tumor immune microenvironment that is not captured by PD-L1 CPS, and may explain why the latter has shown inconsistent predictive performance, particularly in HR⁺/HER2⁻ disease, in which PD-L1 expression is frequently low. If prospectively valiadated, ImPrint-based patient selection could substantially improve the risk-benefit balance of dual-IO strategies by enriching for patients most likely to benefit while potentially sparing others from unnecessary toxicity. Complementing ImPrint with BluePrint molecular subtyping may further refine patient selection.

Moving forward, a more productive objective may be to refine the therapeutic index of the PD-1/LAG-3 axis rather than to abandon it. Sustained peripheral T cell activation resulting from dual blockade of two non-redundant inhibitory pathways may drive off-target toxicities. Because PD-1 and LAG-3 cooperatively maintain peripheral self-tolerance, simultaneous inhibition could trigger autoreactive T cell responses and cytokine release involving the pituitary and adrenal glands. Bispecific antibody formats, designed to preferentially engage T cells co-expressing both checkpoints within the tumor microenvironment, may offer a strategy to concentrate blockade at the tumor site while limiting systemic exposure. Preclinical evidence suggests that these molecules may preferentially target CD8⁺ TILs over regulatory T cells, promoting the expansion of stem-like CD8⁺ T cells for sustained tumor control (Figure 2). Additional mitigation strategies include Fc engineering to reduce effector function and biomarker-based risk stratification using baseline autoantibodies or HLA genotyping.

However, these translational claims remain hypothetical. Systemically administered bispecific antibodies may still bind peripheral PD-1/LAG-3, and key questions remain regarding receptor occupancy, pharmacokinetics/pharmacodynamics, cytokine release syndrome, Fc-mediated effector function, and tissue-specific targeting. Clinical data for bispecific antibodies in early breast cancer remain very limited. Tobemstomig (RO7247669) is in phase 2 evaluation for PD-L1-positive metastatic TNBC (NCT05852691) [9], while tebotelimab (MGD013) has demonstrated manageable safety across solid tumors [10]. The predictive value of ImPrint also requires prospective validation in independent cohorts. Future research should prospectively validate ImPrint as a companion diagnostic, incorporate toxicity-prediction biomarkers into adaptive platform trials, and investigate the interaction between checkpoint blockade and the hormonal microenvironment of breast cancer. The bispecific approach, therefore, remains investigational. Overall, while the PCF regimen may not enter clinical practice in its current form, the trial points toward a potential direction for biomarker-guided, rationally designed dual immunotherapy in early-stage breast cancer.

Kaixiang Feng: writing – original draft (lead), investigation (equal). Fanxiang Yin: investigation (equal), writing – original draft (equal). Mengting Chen: investigation (equal), formal analysis (equal), visualization (equal). Zongbi Yi: conceptualization (lead), supervision(lead), writing – review and editing (lead).

The authors have nothing to report.

The authors have nothing to report.

The authors declare no conflicts of interest.

The authors have nothing to report.