Pathologic features of response to neoadjuvant anti-PD-1 in resected non-small-cell lung carcinoma: a proposal for quantitative immune-related pathologic response criteria (irPRC)

Abstract

Introduction

Lung cancer is the leading cause of cancer death worldwide. Recent improvements in survival of patients with unresectable non-small-cell lung cancer (NSCLC) treated with PD-1/PD-L1 immune checkpoint blocking agents have generated interest in extending their use into the neoadjuvant setting. We recently published the results of a single arm, phase II clinical trial of neoadjuvant nivolumab in resectable stage I–IIIA NSCLC. In that study, while only 10% (2/20) of patients had radiographic objective responses by the time of surgical resection, 45% (9/20) of resected tumors showed a major pathologic response (MPR) [defined as ≤10% residual viable tumor (RVT) at the time of resection] using criteria for neoadjuvant chemotherapy (cMPR) [1]. (See Table ​Table11 for terms and definitions related to pathologic response assessment.)

Due to the distinct mechanisms of action between conventional chemotherapy and immunotherapy, we hypothesized that the histopathologic features of pathologic response to immune checkpoint blockade may diverge from those reported for chemotherapy [2, 3], as suggested by anecdotal descriptions of tumor histology following anti-PD-1 [4, 5]. A standardized approach to the assessment of pathologic response is necessary for reliable interpretation of the post-treatment resection specimens in patients receiving neoadjuvant anti-PD-1 therapy and is particularly critical given the multitude of neoadjuvant immunotherapy clinical trials currently underway.

In the current study, we systematically assessed the features of histopathologic response to neoadjuvant anti-PD-1 in a previously described cohort of patients with NSCLC [6], and characterized features of immune-mediated tumor clearance. We then used the observed histologic features of immune-mediated regression to create a provisional scoring system for pathologic response to neoadjuvant anti-PD-1.

Materials and methods

Results

Tumor bed identification by radiographic and gross pathologic correlation

A strong correlation was identified between the largest lesional diameter on the presurgical resection CT scan (taken post-immunotherapy), and the largest diameter measured during gross examination of the resection specimen by pathology (Figure ​(Figure1A1A and B). Microscopic examination of the ‘tumor mass’ revealed that the composition ranged from complete replacement of tumor by fibroinflammatory areas of immune-mediated regression (referred to here as ‘regression beds’) in patients with pCR, to masses composed nearly entirely of viable tumor, consistent with no treatment response (Figure ​(Figure1C).1C).

Open in a separate window

Figure 1.

Radiographic and gross pathologic measurements of ‘tumor mass’ include areas of immune-mediated tumor regression. (A) There is a significant correlation between the radiographic measurement of the post-treatment lung mass on CT scan and pathologic measurement of the mass in the resected lung specimen. The grossly measured mass includes the microscopically evident regression bed. (B) Representative example of post-neoadjuvant anti-PD-1 treatment lung mass measured on CT scan and the corresponding resected lung specimen. The mass measures ∼3.8 cm in greatest dimension by both modalities, though microscopic examination shows the mass is composed of only 50% RVT. (C) Microscopic examination reveals post-treatment lung masses are composed of varying proportions of fibroinflammatory regression stroma and RVT. Representative photomicrographs from a complete pathologic responder (top), partial pathologic responder (middle; residual tumor with black asterisk, regression bed with white asterisk), and tumors showing no treatment effect (bottom). Original magnification: ×400, all panels.

Histopathologic features of immune-mediated regression in NSCLC resection specimens

The regression bed is characterized by dense immune infiltrates with features of activation [tertiary lymphoid structure (TLS), dense tumor infiltrating lymphocytes (TIL) infiltrates and plasma cells, granuloma formation], along with features of cell death (cholesterol clefts, interstitial foamy macrophages), and tissue repair/wound healing, such as neovascularization and new, proliferative fibrosis (Figure ​(Figure2A).2A). These features were absent or rarely present in specimens from nonresponders. In particular, the percent of the tumor bed surface area occupied by new fibrosis and the presence of neovascularization were two of the key distinguishing features between cMPR specimens versus specimens showing cNR to therapy (P ≤ 0.001 for both; Figure ​Figure2B).2B). In contrast to TLS, unstructured lymphoid aggregates were present in specimens from both responders and nonresponders (NR). The features of cell death such as necrosis, cholesterol clefts, giant cells, and foamy macrophages, as well as acellular fibrosis and mixed inflammatory infiltrates, have been previously reported in specimens from patients treated with neoadjuvant chemotherapy [2, 8]. However, the specific immunologic features of TLS and dense plasma cells accompanied by the proliferative fibrosis and neovascularization of wound healing/tissue repair have not been detailed in chemotherapy patients and may reflect the different mechanism of action of immunotherapeutic agents.

Open in a separate window

Figure 2.

Histologic features of pathologic response to neoadjuvant anti-PD-1 in NSCLC. (A) Post-treatment NSCLC specimens with MPR to therapy showed a distinct pattern of immune-mediated tumor regression, characterized by histologic features of immune activation, tumor cell death, and tissue repair. (B) Association of individual histologic features with MPR/complete pathologic response.

Following the nomination of immune-mediated features of regression identified when comparing the two extremes of the cohort (pCR/cMPR versus cNR specimens), all specimens were then individually assessed for each feature. Most cases showing any degree of response exhibited a moderate to severe grade of TIL in association with a number of the other histologic findings, most commonly neovascularization, proliferative fibrosis, dense plasma cells, cholesterol clefts, and TLS (Figure ​(Figure3A).3A). An ‘outside-in’ pattern of regression was typically present, meaning that regression beds were observed surrounding residual tumor and abutting normal background lung tissue (Figure ​(Figure3B3B and C). The peripheral location of the regression bed has important implications for how specimens are sampled for histologic analysis and how the total tumor bed area is defined microscopically.

Open in a separate window

Figure 3.

Individual features of pathologic response co-localize in the regression bed surrounding residual tumor. (A) Features of pathologic response identified in MPR specimens are also present in the regression bed of ‘partially responding tumors’. However, the tumors from patients 11, 14, and 16 (black box outlines on heat map) had minimal features suggestive of immune-mediated regression. These cases were notable for immune exclusion (patient 11, see Figure 5) and disease relapse (patients 14 and 16). (B) Photomicrograph of a partial responder with 30% cRVT (blue outline). The regression bed (green outline) typically surrounds the residual tumor foci, suggesting an ‘outside-in’ pattern of tumor regression. (C) A patient with a partially responding keratinizing, squamous cell NSCLC (bottom panel) had a regression bed containing residual keratin pearls (top panel, asterisk), further supporting the interpretation of this distinctive peritumoral region as ‘regression bed’. (D) Features of immune-mediated pathologic response may also be identified in lymph nodes, as shown here for patient 4. The double-headed arrow denotes the extent of the regression bed, which is positioned between normal lymph node tissue (asterisk) and RVT (black outline). These findings are notable, as they suggest that immune-mediated responses can be identified and potentially quantified using irPRC, even in tissues normally rich in immune cells. Additionally, the histologic features of regression bed are distinctive enough that responses in previously undiagnosed microscopic disease may also potentially be estimated. Original magnifications: (B) ×20, (C) ×400, (D) ×20 (top panel), and ×200 (bottom panels).

The distinct features of immune-mediated regression observed in the lung were also observed in the resected draining lymph nodes from two patients (patients 4 and 13) (Figure ​(Figure3D).3D). Based on pretreatment imaging, only one of those patients (patient 13) was suspected to have nodal involvement, with neither having pathologically confirmed pretreatment nodal disease.

irPRC: a novel scoring system

Following identification of pathologic features of response to anti-PD-1 therapy, irPRC were developed (supplementary Table S1, available at Annals of Oncology online). Using this approach, %irRVT = viable tumor area/total tumor bed area, whereby the total tumor bed = regression bed + RVT + necrosis (Figure ​(Figure4A).4A). If multiple primary tumor foci are present, the areas from each are summed such that the %irRVT is a representation of the total primary tumor burden. Due to the distinct histologic features of immune-mediated regression, it may also be possible to score %irRVT in lymph nodes (supplementary Figure S3, available at Annals of Oncology online).

Open in a separate window

Figure 4.

Proposal for reproducible, quantitative irPRC. (A) %irRVT is assessed by dividing the total surface area of RVT (circled in blue) by the total tumor bed area ×100. The total tumor bed area (circled in green) is composed of the regression bed area+RVT area+areas of necrosis (supplementary Table S1, available at Annals of Oncology online). The lymph node RVT (LN-irRVT) is calculated separately, but using the same approach, i.e. by measuring the cross-sectional area of RVT and dividing by total tumor bed area. If there are multiple lymph node foci, the areas of each component are summed together to give a %LN-irRVT that represents the total tumor burden in all the lymph nodes assessed. (B) irPRC improves interpathologist reproducibility for percent RVT (%irRVT). Four pathologists blinded to clinical outcome were trained on the tumor from patient 1, and then scored tumors from patients 2 to 20 by both irPRC and chemotherapy criteria. The mean±SD %irRVT and %cRVT scores for each case are shown in black and blue, respectively. Strong concordance was observed among pathologists for the specimens with very little residual tumor when using either the chemotherapy criteria or irPRC. In contrast, large discrepancies in scores for %cRVT were seen in specimens from partial responders. (C) The interpathologist variability in scoring of each case is presented as the range of %RVT scores from four pathologists using the chemotherapy (cRVT) and immune-related (irRVT) criteria (left), shown by blue and black lines, respectively. For each method, the median overall range is presented (right). The proposed immune-related response criteria significantly reduce interpathologist variability in scoring relative to the chemotherapy criteria in these post-immunotherapy specimens.

Discussion

Immune checkpoint blockade in the neoadjuvant setting represents one of the next frontiers in cancer immunotherapy. Clinical trials are underway for patients with NSCLC and multiple other solid tumor types; however, it will be years before survival outcomes are available. MPR has been put forth as a surrogate for survival in lung cancer patients receiving neoadjuvant chemotherapy [9], and scoring systems for calculating the %cRVT in that setting have been recommended [8, 9]. Assessment of pathologic response has the advantage of providing an early indication of therapeutic efficacy within weeks or months instead of waiting years to accrue survival data [10, 11]. It may also allow physicians to determine whether or not additional adjuvant therapy is necessary postresection.

Immunotherapies have a mechanism of action distinct from chemotherapy. Here, we systematically identify the histopathologic features of treatment response to anti-PD-1 and use them to form the basis of a novel scoring system. Recognition of the specific histologic features of immune-mediated regression, a.k.a ‘the regression bed’, promoted higher reproducibility amongst the pathologists, which will be key in accurately predicting outcomes for individual patients and facilitating comparisons between different clinical trials [12]. It is possible that a %irRVT different than the 10% threshold associated with cMPR will be associated with improved progression-free and/or overall survival in this setting, especially if immunotherapy has the anticipated effect of driving improved outcomes by suppressing or eradicating micrometastatic disease.

We also explicitly demonstrated that the regression bed accounts for the previously reported discrepancy between radiographic response and pathologic response [6]. Specifically, 18 out of the 20 patients assessable for pathologic response had stable disease and the remaining 2 patients had a partial response by RECIST 1.1 assessment, despite varying amounts of RVT identified microscopically. These findings support using a ‘disease control’ metric, which includes patients with ‘stable disease’ along with objective responders (complete and partial radiographic responders by RECIST) when assigning patients to categories of response versus nonresponse. It remains to be determined whether nuanced signals on PET-CT or other sophisticated imaging approaches can be used to distinguish between regression bed and residual tumor.

To date, most studies that have analyzed post-treatment tumor specimens from patients with advanced cancer treated with anti-PD-1 have focused on the effector role of cytotoxic T-lymphocytes in tumor eradication. Notably, the neoadjuvant setting provides a unique window for studying drug mechanism of action. In the post-treatment resection specimens from responders, not only were dense TILs observed, but plasma cells and TLS were also enriched. TLS are thought to be key for supporting both local and systemic T- and B-cell antitumor responses [13]. Additionally, B cells are thought to be involved in accelerating wound healing, and have been associated with neoangiogensis and new fibrosis [14], both of which are key features observed in the regression beds formed during immune-mediated tumor regression. Other immune cell subsets, including T-regs, Th2, and M2 cells, as well as cells dying from the so-called ‘immunogenic cell death’ have also been implicated in stimulating tissue repair [15–17], highlighting the complex interplay between specific immune cell subsets, tumorigenesis, tumor cell death, and tissue repair and regeneration.

The extent of immune-mediated tumor clearance in this cohort may largely be a function of the interval between treatment initiation and definitive resection. Histologic patterns suggestive of antigenic but incompletely eliminated tumors were observed. Either with more time, or with additional immunotherapy, there may have been more extensive or complete tumor clearance in these thoroughly immune-infiltrated tumors. Interestingly, we identified a post-treatment tumor with a striking pattern of immune exclusion. Activation of the canonical β-catenin pathway has been associated with T-cell exclusion in melanoma, bladder cancer, and ovarian carcinoma [18–20] and may therefore be implicated in the pattern of immune exclusion identified in our study. In the neoadjuvant setting, when the resistant clone is confined to the primary tumor bed, clearance is likely achieved by the definitive resection surgery. However, if micrometastases of the resistant clone have seeded, these patients may require strategic adjuvant approaches, e.g. chemotherapy, stimulator of interferon genes complex (STING) agonists, or even β-catenin inhibitors [21, 22].

Standardized assessments of %RVT have the potential to facilitate trial design and accelerate drug development. However, it is important to note that %cRVT as a surrogate of survival was validated in the setting of neoadjuvant chemotherapy. While we have outlined a proposed approach to assessment of %irRVT in patients receiving immunotherapy, these findings will need to be confirmed when parameters such as long-term disease-free and overall survival become available on this cohort, and through future clinical trials with larger patient cohorts in the phase II and III settings. Based on our findings in this study, we suggest sampling a full cross-section of tumor through the largest dimension to facilitate the most-accurate assessment of immune-mediated tumor regression, supplementary Figure S6, available at Annals of Oncology online. We also suggest the minimum pathologic parameters gathered should include those in the Case Report Form (supplementary Table S3, available at Annals of Oncology online), enabling longer-term assessments to be made regarding which pathologic features are most closely linked to long-term outcomes. Similar assessments will be required when neoadjuvant immunotherapy is administered to patients with other tumor types, and when immunotherapy is combined with chemotherapy and other antitumor agents.

Supplementary Material

Acknowledgements

References