Pan-immune-inflammation value predicts immunotherapy response and reflects local antitumor immune response in rectal cancer

2.1 Patients and materials

This study included 915 treatment-naïve patients with RC who had undergone curative surgery at two different institutions. Between January 2018 and January 2021, 679 RC patients from Peking Union Medical College Hospital (PUMCH) were included in the discovery cohort (PUMCH cohort). Between January 2015 and January 2017, 236 RC patients from the 7th Medical Center of the PLA General Hospital (PLAGH) were included in the validation cohort (PLAGH cohort). Patients with distant metastasis, non-adenocarcinoma, or those receiving neoadjuvant therapy such as chemotherapy or chemoradiotherapy were excluded. Only the patients with complete data were included, and the follow-up period was calculated from the date of surgery to the last censor.

Furthermore, this study included 16 RC patients from PLAGH who received neoadjuvant chemoradiotherapy combined with an anti-PD-1 antibody (tislelizumab) between October 2023 and May 2024.

2.2 Clinicopathological characteristics

Clinicopathological characteristics included sex, age, T stage, N stage, lymph node ratio (LNR), tumor diameter, tumor grade (G1-2: well/moderate G3: poor), tumor location, tumor deposits, vascular invasion, and perineural invasion. The blood-based PIV was calculated based on preoperative peripheral blood data using the formula: PIV = (neutrophil count [10⁹/L] × platelet count [10⁹/L] × monocyte count [10⁹/L])/lymphocyte count (10⁹/L). Tumor markers, including carbohydrate antigen 19–9 (CA19-9) and carcinoembryonic antigen (CEA), were measured within 1 week before surgery.

The assessed follow-up outcomes were disease-free survival (DFS) and cancer-specific survival (CSS). DFS was defined as the time between radical surgery and recurrence during follow-up. CSS was defined as the time between radical surgery and death due to RC during follow-up.

2.3 Immunohistochemistry

Immunohistochemistry analysis was performed on 4-μm-thick sections following established protocols.^{20, 21} The procedure involved deparaffinization, rehydration, incubation with 3% H₂O₂, heat-induced antigen retrieval, incubation with primary and secondary antibodies (CD8 [ab101500, Abcam], CD20 [ab9475, Abcam], CD21 [TA327627, ZSGB-BIO], CD23 [TA801554, ZSGB-BIO], and human lymphocyte antigen (HLA) class I [88274T, CST]), and diaminobenzidine staining. Sections were then stained with hematoxylin, dehydrated, and mounted on neutral resin. Scanning was performed using a Leica CS2 and analyzed using an Aperio ImageScope.

2.4 Assessment of tumor regression score (TRS)

Tumor regression score in the post-neoadjuvant therapy resected specimens was determined according to the AJCC guidelines.²¹ TRS was classified as follows: 0, no viable cancer cells; 1, single cells or rare small groups of cancer cells; 2, residual cancer outgrown by fibrosis; and 3, extensive residual cancer with no evident tumor regression.^{21, 22} Patients with TRS 0 and 1 were classified as responders, whereas those with TRS 2 and 3 were considered nonresponders. Specifically, TRS 0 and TRS 1–3 indicate complete pathological response (pCR) and incomplete pathological remission (non-pCR), respectively.

2.5 Assessment of TILs and tertiary lymphoid structures (TLS)

Hematoxylin and eosin-stained tissue sections were examined to assess TIL intensity and TLS maturity. TILs are lymphocytes infiltrating the tumor stroma within the borders of invasive tumors. TIL intensity was evaluated using International TILs Working Group criteria,²³ categorizing tumors into three groups: low (TILs <10%), intermediate (10% ≦ TILs <50%), and high (TILs ≧50%) TIL. The classification of TLS maturity is as follows: aggregate (Agg) containing primarily clusters of B cells (CD20) without FDC network (CD21) and GC (CD23); primary follicles (FL-1) containing clusters with an FDC network (CD21) without GC (CD23); and secondary follicles (FL-2) containing clusters with an FDC network (CD21) and GC (CD23).^{20, 21} Patients were classified into the Agg group when only Agg was present, without FL1 or FL2; FL1 group when FL1 was present without FL2; and FL2 group when FL2 was present. The HLA class I status, determined by EMR8-5, is categorized as either high or low.²⁴ HLA I-High was defined as complete and heterogeneous membrane staining in more than 80% of the tumor cells. HLA I-Low was defined as membrane staining in less than 80% of the tumor cells.

2.6 RNA-seq and analysis

We conducted RNA-seq analysis on fresh-frozen tissues from 59 patients, including 19 PIV-high and 39 PIV-low samples. Total RNA was extracted using TRIzol® Reagent, according to the manufacturer's instructions (Invitrogen), and genomic DNA was removed using DNase I (TaKara). RNA purification, reverse transcription, library construction, and sequencing were performed by Shanghai Majorbio Bio-pharm Biotechnology Co., Ltd. following the manufacturer's instructions (Illumina). The transcriptome library was prepared using the TruSeqTM RNA sample preparation kit (Illumina) with 1 μg of total RNA. mRNA was isolated using oligo(dT) beads, fragmented, and reverse transcribed into double-stranded cDNA using the SuperScript kit (Invitrogen) with random hexamer primers (Illumina). The cDNA underwent end-repair, phosphorylation, and “A” base addition as per the manufacturer's protocol. Libraries were size selected for 300 bp fragments on 2% low-range ultra agarose, followed by PCR amplification with Phusion DNA polymerase (NEB) for 15 cycles. The libraries were quantified using TBS380 and sequenced on the Illumina NovaSeq 6000 (2 × 150 bp read length).

2.7 Bioinformatics analyses

The KEGG Legacy, HALLMARK, and ImmuneSigDB gene sets were downloaded from the Molecular Signatures Database.²⁵ The immune regulatory molecule²⁶ and angiogenesis-related genes²⁷ (poor-prognosis angiogenesis genes [PPAG] and good-prognosis angiogenesis genes [GPAG]) were obtained from previous studies. The ESTIMATEScore, ImmuneScore, StromalScore, and TumorPurity were evaluated using the ESTIMATE method.²⁸ Immunological features, including presence of immune cells and activity of immune-related pathways, were determined using the GSVA package.²⁹ Detailed information on gene enrichment of immune cells, immune-related pathways, TLS, and interferon (IFN) is provided in Table S1. The R package clusterProfiler was used for functional gene annotation.

2.8 Evaluation of prognostic value

Kaplan–Meier survival analysis was performed to evaluate the risk factors for DFS and CSS in RC patients. Univariate and multivariate Cox regression analyses were used to identify significant independent prognostic factors. A nomogram integrating four independent prognostic factors was created to estimate DFS probability in the PUMCH cohort (n = 679). Calibration curves were drawn to compare the predictive and actual survival probabilities at the 3 and 5 years to assess the nomogram's predictive performance. The predictive value of the nomogram was validated in the PLAGH cohort (n = 236).

2.9 Statistical analysis

All statistical analyses were performed using R software (version 4.2.1) and Prism (version 8.0). Categorical variables were assessed using the chi-square test and Fisher's exact test, and continuous variables were analyzed using the Mann–Whitney U test. The threshold for DFS was evaluated,³⁰ and patients were divided into two groups based on whether their PIV, LNR, CEA, and CA19-9 values were above or below the optimal cutoff. All statistical analyses were two sided, with p < 0.05 indicating statistical significance. Figures were generated using R (version 4.2.1), Adobe Illustrator (version 26.0.1), and Prism (version 8.0) software.

3.1 Patients' clinicopathological characteristics and PIV

This study included 915 treatment-naïve RC patients who had undergone radical surgery (679 and 236 patients from the PUMCH and PLAGH cohorts, respectively). Overall, 558 (61%) were men, with a median age of 62 years (interquartile range [IQR], 55–69). The median follow-up time was 33.1 months (IQR, 25.6–49.68) and 60 months (IQR, 60–60) for the PUMCH and PLAGH cohorts, respectively. Demographic characteristics such as sex, tumor location, LNR, tumor grade, vascular invasion, perineural invasion, CEA levels, and tumor diameter were comparable between the cohorts (Table 1). However, the proportion of patients with older age (p < 0.001), pT4 (p = 0.018), pN+ (p = 0.009), tumor deposits (p = 0.025), and lower CA19-9 levels (p = 0.045) were higher in the PUMCH cohort (Table 1).

3.2 Prognostic value of PIV in RC

To investigate the prognostic value of PIV, Kaplan–Meier survival analysis was performed to explore its association with DFS and CSS. The distribution of PIV is shown in Figure 1A. The optimal cutoff value of PIV for DFS was 363.906 in the PUMCH cohort (Figure 1B). Patients were then categorized into PIV-high and PIV-low groups based on this cutoff value. Additionally, we evaluated the optimal cutoff values for other indicators (CEA, CA19-9, and LNR) for DFS in the PUMCH cohort, and patients were categorized into high and low groups based on these values in both the PUMCH and PLAGH cohorts (Table S2). In the PUMCH cohort, the PIV-high group exhibited poorer DFS (p = 0.0021; Figure 1C) and CSS (p = 0.019; Figure 1D). Similarly, in the PLAGH cohort, the PIV-high group demonstrated an inverse association with both DFS (p < 0.001; Figure 1E) and CSS (p = 0.014; Figure 1F).

Multivariate Cox regression models were constructed to determine whether PIV was an independent prognostic factor. Specifically, univariate Cox regression analysis identified that PIV, tumor grade, vascular invasion, perineural invasion, pT, LNR, tumor deposits, CEA levels, and CA19-9 levels were associated with both DFS and CSS in the PUMCH cohort (Tables 2 and S3). Furthermore, univariate Cox regression analysis identified that PIV, tumor location, vascular invasion, perineural invasion, pT, LNR, tumor deposits, tumor diameter, CEA level, and CA19-9 levels were associated with DFS, and PIV, pT, LNR, tumor deposits, tumor diameter, CEA level, and CA19-9 level were associated with CSS in the PLAGH cohort (Tables 2 and S3). We constructed DFS- and CSS-based multivariate Cox models for both the PUMCH and PLAGH cohorts. In the PUMCH cohort, significant factors for DFS included PIV, LNR, tumor deposits, and vascular invasion (Figure 2A), and PIV and LNR were significant factors for CSS (Figure 2B). Furthermore, in the PLAGH cohort, PIV, LNR, and tumor deposits were significant predictors in both DFS- (Figure 2C) and CSS-based Cox models (Figure 2D). These results indicate that PIV levels are independent prognostic factors in RC.

3.3 Pan-immune-inflammation value nomogram predicts RC prognosis

To enhance the prognosis prediction in RC patients, we developed a nomogram to predict DFS at 3 and 5 years based on multivariable Cox regression analysis in the PUMCH cohort. The nomogram included four independent prognostic factors: vascular invasion, tumor deposits, LNR, and PIV (Figure 3A). Calibration curves indicated good consistency between the actual and predicted probabilities of 3- and 5-year DFS in the PUMCH cohort (Figure 3B). To verify the predictive accuracy of the constructed nomogram, 236 RC patients from the PLAGH cohort were included as the external validation cohort. The calibration curve showed that the nomogram maintained good predictive value in the PLAGH cohort (Figure 3C). Collectively, these findings underscore the independent prognostic significance of PIV in RC patients and highlight the utility of the nomogram in identifying patients at high risk for recurrence and poor outcomes.

3.4 Pan-immune-inflammation value levels are associated with distinct clinical characteristics in RC

We further investigated the association between PIV and clinical and pathological factors in RC. In the PUMCH cohort, high PIV was significantly associated with male sex (male: 71 vs. 59%, p = 0.026), upper tumor location (upper: 50 vs. 35%, p = 0.021), elevated CEA levels (median [IQR]: 4.3 [2.2–8.17] vs. 3.12 [1.9–6.54], p = 0.01), and larger tumor diameter (median [IQR]: 4.6 [3–6] vs. 3.5 [2.5–4.5], p < 0.001) (Table 3). In the PLAGH cohort, high PIV was significantly associated with elevated CEA levels (median [IQR]: 10.1 [3.55–20.4] vs. 2.55 [1.5–5.73], p < 0.001), elevated CA19-9 levels (median [IQR]: 20.8 [9.88–61.55] vs. 11.4 [6.5–21.5], p = 0.006), and larger tumor diameter (median [IQR]: 4.75 [3.38–6] vs. 3.3 [2.4–4.5], p < 0.001) (Table S4).

Logistic multivariable analysis provided additional insights. In the PUMCH cohort, high PIV was associated with larger tumor diameter (≥5 cm: adjusted odds ratio [aOR], 3.43; 95% CI, 2.11–5.56; p < 0.001, Table 4). In the PLAGH cohort, logistic multivariable analysis indicated that high PIV was associated with larger tumor diameter (≥5 cm: aOR, 4.05; 95% CI, 1.48–11.07; p = 0.006), upper tumor location (upper: aOR, 3.62; 95% CI, 1.16–11.31; p = 0.027), and elevated CEA levels (high: aOR, 7.06; 95% CI, 2.13–23.42; p = 0.001) (Table 4).

3.5 Potential interplay between the antitumor immune responses and PIV in RC

Systemic inflammatory response may influence local immune responses via the peripheral blood.^{18, 19} We further evaluated the immunogenic signature according to PIV levels using RNA sequencing. KEGG pathway analysis identified several enriched pathways in the PIV-low group, including allograft rejection, antigen processing, and presentation (Figure 4A). Conversely, pathways enriched in the PIV-high group included ECM–receptor interactions, cancer pathways, axon guidance, and focal adhesion (Figure 4A). In the HALLMARK gene set analysis, the PIV-low group was associated with IFN pathways-related gene sets, whereas the PIV-high group was associated with gene sets related to angiogenesis, epithelial–mesenchymal transition (EMT), and hypoxia (Figure 4B). Notably, oncogenic gene sets such as KRAS and TGF-ß were strongly activated in the PIV-high group, while MYC-related gene sets were strongly suppressed in the PIV-low group (Figure 4B, Table S5). Furthermore, immunologic signature gene set analysis indicated that CD8⁺ T cell activation and regulatory T cell function-related gene sets were enriched in the PIV-low group (Figure 4C). In contrast, the inflammatory response-related gene sets were enriched in the PIV-high group (Figure 4C).

Moreover, we evaluated immune functions and immune cell infiltration according to the PIV status (Figure 4D). Compared with PIV-high patients, PIV-low patients exhibited higher levels of immune cell markers (B cells, dendritic cells [DCs], macrophages, mast cells, neutrophils, pDCs, T helper cells, Th1 cells, Th2 cells, TIL, and regulatory T cells [Tregs]) and increased activation of immune-related pathways (cytokine and cytokine receptor [CCR], checkpoint, inflammation promoting, T cell coinhibition, T cell costimulation, and IFN) (Figure 4D). Furthermore, we evaluated the expression of IFN pathway-related genes between the two groups and found that the PIV-low group had higher expression of IFN pathway-related genes (Figure 4E).

Next, we explored the differences in expression of immunomodulatory genes between the PIV-low and PIV-high groups. The expression of immune regulatory molecule genes was higher in the PIV-low group than in the PIV-high group (Figure 4F). In particular, genes related to antigen presentation (HLA-DQA1, HLA-DQB2, HLA-DRB1, and HLA-DRB5), cell adhesion (SELP), inhibitory immune checkpoint (PDCD1 [PD-1], CTLA4, HAVCR2 [TIM-3], LAG3, EDNRB, and BTLA), and agonistic immune checkpoint (TNFRSF18 [GITR]) were upregulated in the PIV-low group. Conversely, genes related to the inhibitory immune checkpoint (VTCN1) and immunomodulatory ligands (IL-1α and VEGF-ß) were upregulated in the PIV-high group.

To further characterize the TIME between the PIV-low and PIV-high groups, we evaluated the ESTIMATEScore (p < 0.05), ImmuneScore (p = 0.0520), StromalScore (p = 0.0656), and TumorPurity (p < 0.05) using the ESTIMATE method (Figure 4G). Additionally, we assessed the angiogenesis-related genes between the PIV-low and PIV-high groups by evaluating GPAG and PPAG. Compared with the PIV-high group, the expression of GPAG-related genes was upregulated in the PIV-low group, whereas the expression of PPAG-related genes was not significantly different between the two groups (Figure 4H).

3.6 Degree of immune cell infiltration patterns correlated with PIV levels in RC

RNA sequencing analysis revealed that low PIV was associated with an increased antitumor immune response; therefore, we evaluated the intensity of TILs, maturity of TLS, density of CD8+ cells, and density of HLA-I using H&E and IHC staining in the PLAGH cohort (n = 236), and the representative microscopic images are shown in Figure 5A–D.

Among them, 31 (13%), 106 (45%), and 99 (42%) patients had high, intermediate, and low TILs, respectively. Notably, the TIL-low group exhibited a statistically significant increase in PIV compared with the TIL-intermediate (p = 0.031, Figure 5E) and TILs-high groups (p = 0.012, Figure 5E). However, no significant difference in PIV was observed between the TIL-high and TIL-intermediate groups (Figure 5E).

Regarding TLS maturity, 111 (47%), 57 (24%), and 68 (29%) patients had Agg, FL1, and FL2 TLS, respectively. Our findings revealed that PIV was significantly lower in the FL-2 TLS group than in the Agg TLS (p = 0.01) and FL-1 TLS (p = 0.019) groups (Figure 5F). However, no statistically significant differences in PIV were observed between the FL-1 and Agg TLS groups (Figure 5F). Additionally, the PIV was significantly lower in the GC TLS than in the non-GC TLS (p = 0.0018) groups (Figure 5G).

Subsequently, we quantified the density of CD8⁺ T cells in the patients, with a median CD8⁺ T cell density of 135.54 (IQR, 84.4–204.19). Patients were categorized into CD8-high and -low groups based on the median CD8+ T cell density. PIV was significantly lower in the CD8-high than in the CD8-low group (p = 0.015) (Figure 5H). Similarly, PIV was significantly lower in the HLA I-high group than in the HLA I-low group (p = 0.024) (Figure 5I).

Subsequently, we assessed the prognostic impact of tumor microenvironment indicators, TLS, TILs, CD8, and HLA-I, and found that all were associated with prognosis (Table S6).

Transcriptome sequencing was performed to evaluate the relationship between PIV and TLS hallmark gene sets, which revealed that compared with the PIV-high group, the expression levels of CCL4, CCL21, CD79B, CD1D, CETP, RBP5, and PTGDS were significantly increased in the PIV-low group (Figure 5J). Enrichment analysis further indicated a significantly increased TLS score expression level in the PIV-low compared with the PIV-high group (Figure 5K).

3.7 Association between PIV levels and sensitivity to immunotherapy in RC patients

We evaluated the predictive value of PIV in 16 RC patients treated with anti-PD-1 antibody at PLAGH. The baseline characteristics of the patients are summarized in Table S7. The pCR rate was 31.25% (5/16), with 6 (37.5%), 5 (31.25%), and 0 (0.0%) patients achieving a TRS of 1, 2, and 3, respectively. Notably, pre-ICI PIV levels were significantly associated with response to immunotherapy, such as responders versus nonresponders (141.0 vs. 302.5, p = 0.0192) and pCR versus non-pCR (115.6 vs. 171.3, p = 0.0380) (Figure 6A). Furthermore, post-ICI PIV levels were significantly associated with response to immunotherapy (responders: 134.3 vs. 372.5, p = 0.0133), although there was no significant difference between pCR and non-pCR (130.8 vs. 180.3, p = 0.0897) (Figure 6B).

We further evaluated the patterns of ICI-induced changes in PIV. No significant difference was observed between pre- and post-ICI PIV in the ICI-treated group (p = 0.1754, Figure 6C), responders (p = 0.5771, Figure 6D), nonresponders (p = 0.3125, Figure 6E), pCR (p = 0.6250, Figure 6F), and non-pCR groups (p = 0.2402, Figure 6G).