Tumor Immune Response as a Biomarker for Metastasis-Free Survival of Breast Cancer and Immune Checkpoint Inhibition Therapy: A Retrospective Cohort Study

Abstract

Background:

Breast cancers (BRCs) can be classified into 6 molecular subtypes based on gene expression profiles. Previous research suggests that tumor-infiltrating lymphocytes are associated with metastasis-free survival (MFS) in triple-negative and HER2-overexpressing BRC.

Objectives:

Our study aims to investigate further how the immune response (IR) may impact MFS in different molecular subtypes of BRC.

Design:

A single hospital-based retrospective cohort study.

Methods:

A training series of 327 BRCs was used to identify 297 IR transcripts that were correlated with the T cell–associated CD3D transcript or the B cell-associated CD19 transcript. Using these IR transcripts, each of the 6 molecular subtypes was hierarchically clustered into high and low immune responders. An IR score based on the average of the 297 IR transcripts was determined for each BRC. Correlations between the IR score and 3 signatures for IR or response to immune checkpoint inhibition therapy (ICIT) were investigated. A series of 884 BRCs from public datasets was used for confirmation, and the other independent series of 988 BRCs was used for validation.

Results:

For subtype I, high immune responders had a statistically significantly better MFS than low immune responders in all the training, confirmation, and validation series by Kaplan-Meier survival analysis (P = .0039, .049, .039, log-rank test). The same trend was observed for subtype II (P = .16, .052, .015) and subtype IV (P = .0078, .0002, .12). Our IR scores were linearly correlated with the Teschendorff, the T-effector and IFNg, and the T–cell inflamed signatures for IR or ICIT. The IR scores were also linearly correlated with the expression of 6 different immune checkpoint genes.

Conclusions:

Tumor IR is a biomarker for MFS for BRCs of I, II, and IV subtypes. Our study supports the potential use of the IR score for identifying patients responsive to ICIT.

Plain Language Summary

Tumor-infiltrating lymphocytes (TILs) have been used as prognostic factors in several solid tumors, including breast cancer (BRC). The traditional approach is based on pathological evaluation of the extents of TILs, and the results are known to have significant inter-observer variations. Previously, we have classified BRC into 6 molecular subtypes. In the present study, we used genome-wide RNA expressing profiles to identify immune response (IR) transcripts and establish an IR score. We showed that IR scores were correlated with TILs, and a high IR score predicted a better metastasis-free survival (MFS) in subtype I (basal-cell triple negative), subtype II (ER-negative HER2-overexpressing), and subtype IV (luminal B-like) BRC. These results significantly confirm and extend previously published data on the significance of TILs in BRC. In addition, the IR score also correlated with known immune checkpoint inhibition therapy (ICIT) predictive signatures, implying potential use for ICIT.

Keywords

Tumor immune response breast cancer molecular subtypes immune checkpoint inhibition therapy chemotherapy metastasis-free survival

Introduction

The clinical importance of tumor-infiltrating lymphocytes (TILs) on survival has been reported in many studies for breast cancer (BRC).¹ Nevertheless, the beneficial effect of high TILs in patients with BRC was confined mainly to triple-negative or HER2+ subtypes.¹ In contrast, a high degree of TILs was reported as a negative prognostic factor in luminal/HER2-negative tumors.^2
-3 Most of these studies have been limited to BRC subtyping according to estrogen receptor (ER), progesterone receptor (PR), and HER2 status, as determined by immunohistochemistry (IHC) and fluorescence in situ hybridization.

Earlier, we established 6 BRC molecular subtypes based on the expression of 783 selected genes in 327 patients with BRC.⁴ All 6 molecular subtypes exhibit unique clinical and molecular features, and these features can be used to guide treatments.⁵ Subtypes I, II, and III are ER negative, whereas subtypes IV, V, and VI are ER positive. Subtype I BRC showed a high degree of concordance with the basal-like intrinsic subtype of Perou⁵ and is highly sensitive to adjuvant chemotherapy. Subtype II (ER negative, HER2 overexpressing) BRC has amplification of the ERBB2 gene but is negative for ER and PR gene expression. This subtype has high cell cycling and proliferative activity. Subtype III (ER negative, mostly no HER2 overexpression) is a non-basal-like triple-negative breast cancer (TNBC). Subtype IV (ER positive, mostly no HER2 overexpression) is equivalent to the luminal B intrinsic subtype. This subtype has high cell proliferation activity and poor survival outcome. Notably, subtype IV exhibits high expression of TOP2A and is highly sensitive to anthracyclines, resulting in significantly improved long-term survival.⁴ Subtype V (ER and PR positive, and no HER2 overexpression) is a low-risk subset of luminal A. It is an indolent BRC with a very low cell proliferation activity. This subtype is insensitive to cytotoxic chemotherapy agents but has excellent long-term survival. Subtype VI (ER positive, and mostly no HER2 over-expression) is a subset of luminal A and normal breast-like. It has a higher risk of distant metastasis than subtype V, but lower than subtypes I, II, III, and IV. In this gene expression classification scheme, the ER, the PR, and the HER2 statuses can also be reliably and quantitatively determined based on the expression level of transcripts and are highly correlated with the results obtained by IHC.

Immune responses in tumors other than BRC, as well as the use of immune response (IR) scores as biomarkers, particularly as predictors of responses to immune checkpoint inhibition therapy (ICIT), are also under intense investigation.^6
-8

In this study, our objective is to investigate the association between the expression of IR genes and clinical outcomes in molecular subtypes identified through gene expression profiling. We conducted a study on our published cohort (GSE20685)⁴ to investigate the expression of tumor IR genes and how IR gene expression might impact metastasis-free survival (MFS) in different molecular subtypes of BRC. Our result showed the differential impact of IR on MFS in different molecular subtypes of BRC. The study was followed up with a combined cohort of 5 published datasets: GSE7390,⁹ GSE2603,¹⁰ GSE2990,¹¹ GSE2034,¹² and GSE1112,¹³ as well as an expanded, larger independent cohort of 988 patients (GSE272291). We also established a scoring method to quantify IR. The IR score was correlated to the expression of key immune checkpoint genes and 3 published gene expression signatures for IR and ICIT prediction.^14
-16 The results of our study are reported herein.

Methods

Patients and microarray data

A gene expression profiling dataset previously reported by us on 327 patients with BRC using Affymetrix HG-U133 Plus 2 microarrays was used as the training cohort (GSE20685).⁴ The patients in the training cohort were diagnosed and treated between 1991 and 2004. Five previously published public datasets, including GSE7390,⁹ GSE2603,¹⁰ GSE2990,¹¹ GSE2034,¹² and GSE1112,¹³ were pooled and used as the confirmation cohort. A gene expression dataset for a cohort of 988 patients with BRC from our study was used as the independent validation cohort (GSE272291). All patients diagnosed with BRC and treated at KFSYSCC between 1997 and 2014 were included in the study. A total of 988 patients were identified. Clinical information including age, gender, tumor staging, and so on, was retrieved from medical records between March 13, 2014 and February 28, 2015. Survival data were collected from medical records between March 13, 2014 and August 2, 2017. Authors had access to information that could identify individual participants during or after data collection. The clinical characteristics of the training cohort and the validation cohort are summarized in Table 1. Gene expression intensities of each microarray were pre-processed with the MAS5.0 software,¹⁷ log₂ transformed, and then quantile-normalized to the GSE20685 dataset. Details for sample processing, gene expression profiling, subtyping, determination of ER, PR, and HER2 status, clinical correlations, and survival analysis were the same as described in our earlier study.⁴ A STROBE checklist from the EQUATOR Network can be found in Supplemental Table S1 for data reporting transparency.¹⁸

Table 1.

Comparisons between the training and the validation series.

		Training		Validation
		327		988
		Count	%	Count	%	P (Fisher exact)
Age at diagnosis
	<50	203	62%	507^a	51%	.0009
	⩾50	124	38%	478^a	49%
Treatment year
Before or in 1996		124	38%	3	0.30%	<.0001
In or after 1997		203	62%	985	98.70%
TNM stage
	I	69	21%	225	23%	.0009
	II	147	45%	497	50%	(I/II vs III/IV)
	III	103	31%	226	23%
	IV^b	8	2%	6	1%
	Unknown^c	0	0%	34	3%
Positive LN
	0	137	42%	452	46%	.0146
	1-3	88	27%	301	30%	(0-3 vs ⩾4)
	4-9	61	19%	119	12%
	⩾10	37	11%	111	11%
	Unknown^c	4	1%	5	1%
Nuclear grade
	1	33	10%	102	10%	0.0686
	2	90	28%	315	32%	(1 & 2 vs 3)
	3	202	62%	536	54%
	Unknown^c	2	1%	35	4%
ER ^d
	ER+	222	68%	622	63%	.1107
	ER−	105	32%	366	37%
Her2 ^d
	Her2+	71	22%	240	24%	.3680
	Her2−	256	78%	748	76%
PR ^d
	PR+	154	47%	435	44%	.3367
	PR−	173	53%	553	56%
Neoadjuvant chemotherapy
	-	296	90.52%	918	92.91%	.01052
	+	30	9.17%	49	4.96%
	Unknown	1	0.31%	21	2.13%
Adjuvant chemotherapy
	-	54	16.51%	98	9.92%	.00265
	+	268	81.96%	864	87.45%
	Unknown	5	1.53%	26	2.63%
Radiation therapy
	-	175	53.52%	276	27.94%	2.65E-16
	+	145	44.34%	677	68.52%
	Unknown	7	2.14%	35	3.54%
Hormonal therapy
	-	96	29.36%	263	26.62%	.8316
	+	227	69.42%	601	60.83%
	Unknown	4	1.22%	124	12.55%
Herceptin
	-	0	0.00%	2	0.20%	NA
	+	2	0.61%	140	14.17%
	Unknown	325	99.39%	846	85.63%
Median follow-up, year
	Median	7.7		4.7		<.0001
95% CI lower bound		0.23		0.15
95% CI upper bound		12.77		9.11
Metastasis
	Events	83	25.38%	186	18.83%	.014

For 3 cases, the age at diagnosis is unknown.

MFS in stage IV patients is assigned as zero.

Not included in the calculation of survival curves.

Cutoff for ER (205225_at): 10.9309; PR (228443_at): 10.053; Her2 (216836_s_at): 13.4173.

Selection of IR genes

Genome-wide mRNA expression profiles of 327 training cases (GSE20685) were used to select probesets of genes associated with IR. Probesets were selected according to the following 2 criteria. The first selection criterion was probesets with expression intensities greater than 512 in at least 10 out of 327 training cases. The second selection criterion was probesets showing a greater than 2-fold change in expression levels between the 10th and the 90th percentiles of expression levels of the 327 training cases. Out of 54 675 probesets in the Affymetrix HG-U133 Plus2 microarray, 18 682 probesets met these 2 criteria. Then, the expressions of the selected probesets were correlated with the expression intensity of the CD3D gene (Affymetrix probeset ID 213539_at) or the CD19 gene (Affymetrix probeset ID 206398_s_at) using a quadratic model on all 327 samples in the training series. Probesets with P values < 10⁻⁵⁰ were selected. If multiple probesets for the same gene were selected, only the probe set showing the highest range of the expression levels was chosen. A total of 297 probesets/genes met the criteria and were used as the IR genes for the study. These probesets are listed in Supplemental Table S2. DAVID Gene Ontology Term Enrichment Analysis (https://david.ncifcrf.gov/) was used to determine how these 297 genes were associated with IR. The association with IR was also investigated using Ingenuity Pathway Analysis (QIAGEN, Redwood City, California).

Scoring IR based on gene expression

The IR score was calculated after gene expression of the confirmation and the validation series was quantile-normalized against the training series. For each of the 297 IR genes from a BRC case, a z-score of the gene was calculated. The IR score of the BRC case was the average of the 297 z-scores. To calculate the z-score of a particular gene, the formula was used: (gene expression intensity − M)/STD. In the formula, the mean (M) and standard deviation (STD) of expression were based on the distribution of the gene’s expression among 327 cases of the training set. Supplemental Tables S3 and S4 provide details for processing and calculating the IR score from the gene expression of the 297 IR transcripts.

Quantification of intratumoral CD3+ T-cell infiltration

For quantitation of tumor-infiltrating CD3+ T cells in a BRC sample, the most representative tumor section was selected and immunohistochemically stained for CD3 using the mouse anti-human CD3 monoclonal antibody (clone: UCHT1) (Dako, Carpinteria, California), iVIEW DAB Detection kit, and a BenchMark ULTRA automated staining system (Ventana Medical Systems, Inc., Tucson, Arizona). To enumerate tumor-infiltrating CD3+ T cells, a line was drawn across the largest dimension of the tumor section, and the line was then divided perpendicularly into 4 equal parts. An area of each part was randomly marked with a circle using a Nikon microscope object marker. CD3+ T cells were enumerated using an eyepiece with a 10 mm² micrometer grid at a magnification of 200X. All CD3+ T cells in the squares over the circled areas were determined. The sum of CD3+ T cells in the squares of all 4 circled areas was determined as intratumoral CD3+ T cells and correlated with the IR score.

Statistical analyses

The probability of MFS was estimated with the Kaplan-Meier (KM) survival analysis. Multivariate COX regression was used to obtain the hazard ratio of IR score, adjusted for confounders such as age, tumor stage, node stage, and nuclear grade. Missing data were excluded from statistical analysis. Gene set enrichment analysis was carried out using DAVID gene ontology term enrichment analysis (https://david.ncifcrf.gov/tools.jsp). Statistical analyses were performed using the R software package (version 4.3) from Bioconductor (http://www.bioconductor.org). Two-way unsupervised hierarchical clustering analysis in the R software was used to divide BRCs of each or all molecular subtypes into high and low immune responders based on 297 IR genes.

Result

Pathway enrichment analysis of the selected 297 genes

The selected probesets for 297 genes were first studied using the DAVID gene ontology online software. Two hundred eighty-seven out of a total of 297 genes were found to be annotated in one or more of the 35 biological process terms associated with different immunological functions (Table 2). Two hundred ten of the 297 selected genes are included in more than one of the listed biologic process terms. All P values for the 35 immunologically related processes were <10⁻⁹ (Table 2). We also studied these 297 genes using Ingenuity Pathway Analysis. The most significant pathways identified were cell-mediated IR (P = 10⁻¹⁷), immune cell trafficking (P = 10⁻¹⁵), tissue development (P = 10⁻¹⁰), and humoral IR (P = 10⁻⁸). These results showed a high degree of association between the selected genes and immune processes. These 297 genes were used as IR signature genes in our studies.

Table 2.

Pathway enrichment analysis.^a

Immune system process	115	2.56E-71
Immune response	90	3.95E-58
Leukocyte activation	45	3.26E-34
Lymphocyte activation	42	3.43E-34
Regulation of immune system process	50	1.45E-30
T-cell activation	32	1.67E-28
Regulation of lymphocyte activation	33	1.77E-27
Regulation of leukocyte activation	33	8.26E-26
Regulation of T-cell activation	28	4.04E-24
Regulation of immune response	33	1.69E-21
T-cell differentiation	20	2.04E-19
Lymphocyte differentiation	23	4.68E-19
Positive regulation of lymphocyte activation	22	2.28E-18
Leukocyte differentiation	24	7.46E-18
Positive regulation of immune response	24	7.75E-17
Regulation of lymphocyte differentiation	17	1.95E-15
Positive regulation of T-cell activation	18	2.50E-15
Defense response	42	4.07E-15
Hematopoiesis	27	6.86E-15
Hematopoietic or lymphoid organ development	28	8.72E-15
Regulation of T-cell differentiation	15	3.50E-14
Immune system development	28	3.84E-14
Immune response–regulating cell surface receptor signaling pathway	13	1.46E-12
Antigen processing and presentation of peptide or polysaccharide antigen via MHC class II	12	1.88E-12
Antigen processing and presentation	16	3.18E-12
Positive regulation of lymphocyte differentiation	12	3.94E-12
Immune effector process	19	4.00E-12
T-cell selection	10	4.55E-12
Immune response–activating cell surface receptor signaling pathway	12	1.50E-11
Antigen receptor–mediated signaling pathway	11	5.82E-11
Immune response–regulating signal transduction	13	6.04E-11
Positive regulation of biological process	71	9.56E-11
Cellular defense response	13	1.76E-10
Immune response–activating signal transduction	12	4.50E-10
Leukocyte-mediated immunity	14	9.45E-10

Of the 297 genes, 287 were annotated in one or more of the listed immunological functions.

Correlation of IR score with intratumoral T lymphocyte counts

To quantify the IR based on the expression of IR genes, an IR score was devised using a mean of z-scores as described in the “Materials and Methods” section. To investigate whether the IR score derived from the selected 297 genes was biologically relevant at a cellular level, we correlated our IR scores with the number of intratumoral CD3+ T lymphocytes. This study was performed in 30 randomly selected BRC samples from our training series. Intratumoral CD3+ T lymphocytes were enumerated after immunohistochemical (IHC) staining. The results of our study as shown in Figure 1 reveal a linear positive correlation between the 2 parameters (r = .63, P = .0002). This result supported the immunological relevance of the 297 IR genes and our IR score.

Figure 1.

Correlation of immune response (IR) score with intratumoral CD3+ T cells. Thirty BRC cases were randomly selected from the training series for the study. X-axis: CD3+ T cells enumerated from the immunohistochemically stained tumor sections; Y-axis: IR score. The Pearson correlation analysis was conducted. The Pearson coefficient and the P value were shown in the figure. The result showed a high degree of linear correlation.

Correlation of IR signature gene expression with MFS: High immune responders had a better MFS than low immune responders in the training and the confirmation sets, but not the validation set

Next, we used the 297 IR signature genes to unsupervised hierarchical clustering analysis on the training dataset (GSE20685) comprising 327 BRC samples. These 327 patients were classified into a high IR group and a low IR group without considering their molecular subtypes. The first node of the dendrogram for 327 BRC cases was used to divide patients into high (n = 152) and low (n = 175) immune responders (Figure 2A). The MFS between these 2 groups of patients was compared. The result indicated that patients with higher expression of the IR genes had better MFS as determined by KM survival analysis (P = .0074) (Figure 2B).

Figure 2.

Metastasis-free survival (MFS) differences between low and high immune responders in the training, the confirmation, and the validation series. Low and high immune responders were identified according to the hierarchical clustering analysis described in the Materials and Methods section. (A) and (B): training series; (C) and (D): confirmation series; (E) and (F): validation series. A, C, & E; Hierarchical clustering of the training, the confirmation, and the validation cases to classify patients into high (red) and low immune responders (green); B, D, & F: Kaplan-Meier metastasis-free survival curves for the high (red) and the low (green) immune responders in 3 series. The p values of the log-rank test are shown in each figure of the MFS analysis. The sample number and the number of observed metastatic events are also shown in the figures. X-axis: years; Y-axis: metastasis-free survival (MFS).

To confirm this finding, we conducted a similar comparative study on a pool of 884 patients with BRC collected from 5 published BRC gene expression datasets (see Materials and Methods). Of 884 cases, survival data were available in 860 cases. The result of hierarchical clustering analysis showed that 363 patients were high immune responders and 497 were low immune responders (Figure 2C). A comparison of MFS between the 2 groups of patients again revealed that patients with BRC with higher expression of IR signature genes had a better MFS by KM survival analysis (P = .0027) (Figure 2D).

A similar comparison of MFS was performed on the independent validation series of 988 BRCs. Of 988 cases, survival data were available in 987 cases. There were 363 high and 497 low immune responders (Figure 2E). Patients with BRC with higher expression of IR signature genes did not show a significant difference between high and low immune responders by KM survival analysis (P = .41, Figure 2F). The lack of a significant difference is likely attributable to improved treatment and survival outcomes for more recently diagnosed and treated patients with BRC. The validation dataset consisted of patients from a more recent period compared to the training cohort and the cohorts in the confirmation series (see Discussion). In addition, the slightly better MFS survival observed in high immune responders in the validation dataset (Figure 2F) may further explain the absence of differences between high and low immune responders.

Prognostic impact of IRs on MFS according to BRC molecular subtypes

It is well known that MFS is different between different BRC molecular subtypes. We therefore investigated the MFS difference between high and low immune responders according to each BRC molecular subtype. Hierarchical clustering analyses using the IR signature genes were conducted to classify high and low immune responders for each molecular subtype in all 3 series of BRCs (Supplemental Figure S1). MFS was compared between high and low immune responders for each molecular subtype (Figure 3). Kaplan-Meier survival analysis showed that high immune responders had a better MFS than low immune responders (P < 0.05) for subtype I in all 3 series (P = .0039, .0489, and .0391, respectively), subtype II in the confirmation (P = .052) and the validation (P = .015) series, and subtype IV in the training (P = .0078) and confirmation (P = .0002) series.

Figure 3.

Metastasis-free survival (MFS) differences between the high and the low immune responders in each molecular subtype. (A): training series; (B): confirmation series; (C): validation series. X-axis: follow-up time in years; Y-axis: MFS. Green: low immune responders; Red: high immune responders. Kaplan-Meier survival curves, P-values of the log-rank test, and sample numbers with the observed metastatic events in parentheses were shown in the MFS figures for each subtype. The heat maps of hierarchical clustering analyses for each molecular subtype of all 3 series are shown in Fig. S1. The results showed that a higher immune response was associated with better MFS in subtype I of all 3 series, in subtype II of the confirmation and the validation series, and in subtype IV of the training and the confirmation series.

To confirm the findings obtained from the hierarchical clustering analyses, we adopted a more quantitative approach using the IR score to investigate differences in MFS between high and low immune responders in each molecular subtype. In this approach, high immune responders were defined as having an IR score equal to or greater than the median IR score and low immune responders as having an IR score smaller than the median IR score. We then compared rates of 5-year MFS between high and low immune responders in all 3 series. Table 3 shows that high immune responders had a better 5-year MFS rate than low immune responders in subtype I of all 3 series (P = .0079, .0006, and .0136 for the training, the confirmation and the validation, respectively), subtype II of the confirmation (P = .0132) and the validation (P = .04), and subtype IV of all 3 series (P = .027, <.0001, and .079 for the training, the confirmation and the validation, respectively). In summary, the results from KM MFS analyses (Figure 3) and the 5-year MFS rate (Table 3) comparisons consistently show that high immune responders have better MFS outcomes in subtypes I, II, and IV of BRC.

Table 3.

Comparisons of metastasis-free survival rate at 5 years between high and low immune responders.

	Training			Confirmation			Validation
	High	Low	P	High	Low	P	High	Low	P
I	100%	67%	.0079	75%	50%	6.00E-04	87%	69%	.0136
II	65%	62%	.4159	76%	52%	0.0132	77%	60%	.04
III	90%	74%	.0571	75%	33%	.2219	77%	73%	.3378
IV	78%	59%	.0274	76%	53%	<.0001	77%	68%	.0791
V	90%	100%	1.00	88%	88%	.45	92%	96%	.7996
VI	87%	80%	.2811	93%	88%	.0987	85%	89%	.8342

Survival data in 20 out of 880 cases in the confirmation series were not available. One case from the validation series before 1997 was excluded. The high and the low immune responders were defined by an IR score (>the median as high and <median as low).

Bold values indicates p<0.1.

In multivariate COX regression showed that the hazard ratio associated with a high IR score adjusted for age, tumor stage, lymph node stage, or nuclear grade, was 0.41 (95% CI: 0.25-0.66, at P = .0002) for subtype I; 0.55 (95% CI: 0.37-0.81, at P = .0026 for subtype II; 0.60 (95% CI: 0.37-0.99, at P = .044 for subtype III; 0.57 (95% CI: 0.42-0.79, at P = .0005 for subtype IV. The hazard ratio was not significant: 1.32 (95% CI: 0.30-5.80, at P = .71 for subtype V; 0.89 (95% CI: 0.60-1.34, at P = .60 for subtype VI. Additional statistical analysis to estimate confounders and bias in the survival analysis was performed and summarized in Supplemental Tables S5 to S7, according to the STROBE checklist from the EQUATOR Network (Supplemental Table S1).¹⁸

Differential distribution of IR scores in BRC molecular subtypes

To further investigate possible causes for the differential impact of IR on MFS in different molecular subtypes of BRC, we examined the distribution of IR scores in each molecular subtype for all 3 series.

As shown in Figure 4, the median IR scores of all 6 BRC subtypes were 0.17, 0.28, 1.15, −0.16, −0.79, and −0.20 in the training series (Figure 4A); −0.04, −0.04, 0.42, −0.49, −0.79, and −0.38 in the confirmation public series (Figure 4B); and 0.35, 0.20, 0.45, −0.21, −0.92, and −0.28 in the validation series (Figure 4C). Among the 6 subtypes, subtype V had the lowest IR score, and subtype III had the highest IR score. This is true in all 3 series. Interestingly, as already shown in the preceding paragraphs, the 5-year MFS of subtypes III and V is both independent of IR (Table 3). The data for subtypes III, V, and VI also showed a significantly tighter distribution of IR scores in all 3 series. The tighter distribution and lower IR difference within each of the subtypes III, V, and VI could have contributed to the lack of discernible IR effect on MFS in molecular subtypes III, V, and VI of BRC.

Figure 4.

Box-Whisker and violin plots of IR scores of each BRC molecular subtype in 3 series. X-axis: molecular subtypes; Y-axis: the IR score. (A): training series; (B): confirmation series; (C): validation series. The median IR scores, for subtypes I to IV, respectively, were 0.17, 0.28, 1.15, −0.16, −0.79, and −0.20 in the training series (A); −0.04, −0.04, 0.42, −0.49, −0.79, and −0.38 in the confirmation series (B); and 0.35, 0.20, 0.45, −0.21, −0.92, and −0.28 in the validation series (C). Statistical comparisons revealed significant differences in IR among subtypes (P < 10⁻¹⁵ by Kruskal-Wallis test for all 3 series), with the highest IR score for subtype III and the lowest IR score for subtype V.

Correlation of IR score with reported IR signatures for ICIT

Immune response signatures based on 8-gene T-effector and IFNg signature¹⁹ and 18-gene T-cell-inflamed GEP signature²⁰ (Supplemental Table S8) were reported to associate with an increased response rate to ICIT in non–small cell lung cancer. Thus, it is interesting to learn how our IR score is correlated with the 8-gene T-effector and IFNg signature and 18-gene T-cell-inflamed GEP signature

We used genes in each ICIT-prediction gene signature to determine the IR scores for all 3 BRC series. We then correlated our IR scores with the reported IR scores for each ICIT-prediction gene signature. The results shown in Figure 5 revealed a high degree of positive linear correlations between our IR scores and the scores of 2 predictive signatures for ICIT in all 3 series (Figure 5A to F). This finding suggests that our IR score may serve as a biomarker to identify patients with BRC who are more likely to respond to ICIT.

Figure 5.

Correlations between immune response (IR) score and scores of 3 different IR signatures in 3 BRC series. Three different IR signatures were the 8-gene T-effector and IFNg signature,²⁰ the 18-gene T-cell-inflamed GEP signature,¹⁹ and the 7-gene IR module for breast cancer.²¹ Figure 5A to C shows the IR correlations with the 8-gene T-effector and IFNg signature. Figure 5D to F shows the IR correlations with the 18-gene T-cell-inflamed GEP signature. Figure 5G to I shows the IR correlations with the 7-gene IR module for breast cancer. From left to right: the training, the confirmation, and the validation series. X-axis: the IR score; Y-axis: the 8, 18, or 7-gene signature score. The Spearman correlation coefficients are shown in the figure. All P values are <10⁻¹⁵.

To further demonstrate the immunological relevance of our IR score, we correlated our IR score with the 7-gene Teschendorff IR signature, which was reported as prognostic for ER-negative BRC²¹ (Supplemental Table S8). Again, a significant proportional correlation was observed in all 3 BRC series (Figure 5G to I).

Correlation of IR score with immune checkpoint genes

Recent studies indicate that the higher expression of PD-L1 gene on tumor cells and immune cells can be used to identify responders to ICIT.^22,23 In addition, the combination of LAG-3 and PD-1 inhibition has been reported to enhance the antitumor IR and improve progression-free survival compared with PD-1 inhibition treatment alone in advanced melanoma patients.²⁴ To learn how the IR score might correlate with the expression of different immune checkpoint genes, we conducted a correlation study between our IR score and the expression of immune checkpoint genes, including CTLA4, CD274 (PD-L1), PDCD1 (PD-1), LAG-3, HAVCR2 (TIM3), and IDO1 on the validation series. The results showed that our IR score was proportionally and significantly correlated with the expression of all 6 immuno-checkpoint inhibition genes (Figure 6).

Figure 6.

Correlation of immune response (IR) score with the expression of 6 immune checkpoint genes. The Pearson correlation was performed. The Pearson coefficients and the P values are shown in the figure. Y-axis: IR score. X-axis: gene expression intensity in log₂ scale, including CTLA4 (A), CD274 (B), PDCD1 (C), LAG3 (D), HAVCR2 (E), and IDO1 (F). The training and validation sets were combined for the correlation study.

Discussion

Earlier, we established 6 BRC molecular subtypes using gene expression profiles of 327 BRC samples (GSE20685). We demonstrated the unique clinical and molecular characteristics of these subtypes and showed how they can be used to guide treatment optimization.⁴ As previously reported, TILs were reported to associate with a better prognosis in the basal-like triple-negative BRC²⁵ and the HER2-overexpressing BRC,²⁶ but not the luminal subtype BRC.³ Also, the intratumoral T cells and B cells were associated with the IR and the immunotherapy outcomes of BRC.^27
-29

We selected the 297 IR genes by correlating the expression of genes with those of CD3D and CD19 (Table 2, Supplemental Table S2). Our IR score is, therefore, biased toward T-cell and B-cell-mediated immunity. We did not specifically look for natural killer (NK) cells or macrophage-specific genes because the success of ICIT indicates that antitumor immunity is primarily mediated by T-cell IR. Nevertheless, many of the 297 genes selected for our IR scoring are known to be expressed in NK cells (eg, ACAP1, AIM2, AOAH, GNLY, GZMA, GZMB, GZMH, GZMK, KLRK1, KLRD1) or macrophages (eg, ADAMDEC1, CCL5, CCR2, CCR5, CD52, CD53, CD74, CD97, CTSS, CXCL11, CXCL9, CXCR4, FGL2, GBP1, GBP4, GBP5, GMFG, IL10RA, IL18BP, INPP5D, IRF8, ITGAL, LILRB1, LYZ). Therefore, these genes are not absolutely specific for NK cells or macrophages. Thus, our IR signature does not encompass the full spectrum of the IR. Our IR score is mainly focused on the T-cell and B-cell-mediated immunity.

To evaluate whether our IR score could be functionally relevant and serve as the molecular equivalent of TILs, we performed a correlation study between the IR scores and the tumor-infiltrating CD3+ T cells in 30 BRC specimens (Figure 1). The tumor-infiltrating CD3+ T cells were assessed using IHC. This study aims to demonstrate the functional relevance of the IR scores by gene expression, rather than gaining insight into the mechanisms responsible for antitumor immunity. A study to correlate the IR scores with the tumor-infiltrating CD19+ B cells was not performed, as the high abundance of B cells in BRC sections made accurate quantification technically challenging.

We then studied how IR affected MFS in high and low immune responders. The results of our study showed significantly better MFS in high immune responders of the training series (Figure 2A and B). This finding was further confirmed using the confirmation series of 880 BRCs (Figure 2C and D). Nevertheless, the finding was not as evident in the independent validation cohort of 988 BRCs (Figures 2E to F). The lack of difference is likely related to the improved survival outcome in the validation cohort, as this cohort of patients was from a more recent time period and received different treatments (Supplemental Figure S2).

Considering the possibility of differential impact of IR on MFS in different molecular subtypes of BRC, we then conducted our comparative study between high and low immune responders within each molecular subtype. We first classified high and low immune responders using hierarchical clustering analyses in each molecular subtype and performed KM survival comparison. The results of this study showed differential impacts of the IR on MFS in BRC molecular subtypes (Figure 3). We then used the IR score to investigate the 5-year MFS rate difference between high and low immune responders within each molecular subtype (Table 3). The results of both approaches showed that high immune responders had a better MFS than low immune responders in subtype I BRC of all 3 series, subtype II of the confirmation and validation series, and subtype IV of all 3 series with a trend in the validation series. The lack of a significant impact of the IR on the MFS in the training series for subtype II BRC may be due to a smaller sample size in the training series and a less prominent effect of the IR on the MFS in subtype II BRC. For subtype IV BRC, a weaker IR effect was noted in the validation series. The weakening IR effect could be related to the fact that subtype IV patients in the validation series were from a later time period compared with patients in the training series (Supplemental Figure S2) and had anthracycline in the adjuvant chemotherapy regimens. Patients of the training series were diagnosed and treated between 1991 and 2004 (Table 1, Supplemental Figure S2). Many of them received chemotherapy regimens without anthracyclines (eg, CMF). Anthracycline containing chemotherapy regimens is known to significantly survival of subtype IV patients due to their highly expressed TOP2A gene.⁴ The improved survival outcome from anthracycline-containing chemotherapy regimens could have blunted the beneficial impact of the IR in subtype IV patients with BRC of the independent validation cohort.

In contrast to subtypes II and IV, our data showed a consistent IR impact on subtype I BRC (basal-like TNBC) across all 3 series but not on subtype III (non-basal-like triple-negative BRC). This finding supports the importance to subclassify triple-negative BRC regarding the IR impact as previously reported by Lee et al³⁰ In addition, our study showed that, within the luminal subtypes, the IR is prognostic for the survival of subtype IV (a major subset of luminal B) BRC, but not subtypes V (a major subset of luminal A) and VI (a mixture of luminal B and normal breast-like BRC). The observed differential impact of the IR in different BRC molecular subtypes in part could be related to differential degrees of IRs among different subtypes of BRC (Figure 4). Most subtype III BRCs had the highest IR among all 6 subtypes, and subtype V BRCs had the lowest IR. In addition, subtype V and VI BRCs exhibited the smallest variation in IR scores. These unique features combined with more effective treatments in recent periods could have contributed to the absence of the IR impact on MFS in subtypes III, V, and VI in our study. Although it is not entirely clear why the IR does not impact MFS in subtypes III, V, and VI, the high IR scores in subtype III and the relatively low IR scores with reduced variations in subtypes V and VI (Figure 4) as mentioned could have contributed to the absence of the IR impact on MFS in these 3 BRC subtypes. It is not known whether the high immunogenicity of subtype III BRC and the low immunogenicity of subtypes V and VI could be the underlying cause, and deserves further investigation.^31
-33

Earlier, an 8-gene T-effector and IFNg signature²⁰ and an 18-gene T-cell-inflamed gene expression profile (GEP) signature¹⁹ were reported as predictive for identifying non–small cell lung cancer patients responsive to ICIT. It is therefore of interest to correlate our IR scores with the scores of the 8-gene T-effector and IFNg signature and the 18-gene T-cell-inflamed GEP signature (Supplemental Table S8). The results of this study showed high degrees of proportional correlation (Figures 5A-F). To further support the functional relevance of our IR scoring method, we also correlated our IR scores with the 7-gene IR module for BRC²¹ (Supplemental Table S8). Again, we found a high degree of significant proportional correlation between the Teschendorff IR module and our IR scores (Figures 5G to I). The results of all these correlation studies indicate that our IR score based on gene expression may be useful in identifying patients with BRC who are responsive to ICIT.

For immune checkpoint inhibition treatment with anti-PD1 or anti-PD-L1 monoclonal antibodies in different solid tumors, PD-L1 expression as determined by IHC staining is often used as the predictive biomarker. For BRC, PD-L1 expression is also used to guide PD1 or PD-L1 targeted ICIT for metastatic or high-risk early triple-negative BRC.^22,34
-37 To further investigate whether our gene expression-based IR score can be used as a potential biomarker to guide ICIT, we studied the correlation of IR scores with the expression of PD-L1 and other immune checkpoint inhibitory genes including CTLA4, PD1, LAG3, TIM3, and IDO1. As shown in Figure 6, our IR scores are significantly correlated with all immune checkpoint inhibitory genes including PD-L1. Considering the inherent observer-dependent variation of IHC interpretation, the results of our study suggest that the more quantitative and stable IR score may be a more reliable biomarker for identifying patients responsive to ICIT.

The clinical relevance of TILs in different subtypes of BRC is well documented.^3,38 Besides TILs and tumor microenvironment,³⁹ intratumoral microbiota also can contribute to the effect of immunotherapy.⁴⁰ Other immune-related biomarkers for predicting response to immunotherapy are also studied in an ongoing clinical trial.⁴¹ Recent studies highlighted the potential roles of IR in redefining BRC subtypes, and for transforming chemotherapy and ICIT.^42
-44 How our IR score reported herein may contribute to the optimization of ICIT warrants further investigation.

Limitations

The results of our study suggest that the IR score reported herein may serve as a biomarker to identify patients with BRC who may benefit from ICIT in relation to specific molecular subtypes. Nevertheless, it is important to note that our IR score is restricted to the expression of genes primarily associated with T-cell and B-cell-mediated immunity. Our study did not encompass many genes specifically related to NK and myeloid cells. Further study is necessary for a more thorough characterization of the IRs involving NK and myeloid cells, alongside T and B cells, and to understand how a more comprehensive IR signature may influence the prediction of ICIT outcomes.

Conclusions

T-cell and B-cell-mediated IR has a differential impact on MFS in different BRC molecular subtypes. The IR signature of 297 genes reported herein predicts a better MFS in subtype I (basal-cell triple negative), subtype II (ER-negative HER2-overexpressing), and subtype IV (luminal B-like) patients with BRC. The IR score of BRCs shows high degrees of correlation with TILs, PD-L1 gene expression, and 2 distinct predictive signatures for ICIT. Our IR score may serve as a valuable biomarker for ICIT and warrants further investigation.

Supplemental Material

sj-docx-1-bcb-10.1177_11782234251363665 – Supplemental material for Tumor Immune Response as a Biomarker for Metastasis-Free Survival of Breast Cancer and Immune Checkpoint Inhibition Therapy: A Retrospective Cohort Study

Supplemental material, sj-docx-1-bcb-10.1177_11782234251363665 for Tumor Immune Response as a Biomarker for Metastasis-Free Survival of Breast Cancer and Immune Checkpoint Inhibition Therapy: A Retrospective Cohort Study by Chung-Wu Lin, Kai-Ming Chang, Wen-Hui Ku and Kuo-Jang Kao in Breast Cancer: Basic and Clinical Research

Supplemental Material

sj-docx-3-bcb-10.1177_11782234251363665 – Supplemental material for Tumor Immune Response as a Biomarker for Metastasis-Free Survival of Breast Cancer and Immune Checkpoint Inhibition Therapy: A Retrospective Cohort Study

Supplemental material, sj-docx-3-bcb-10.1177_11782234251363665 for Tumor Immune Response as a Biomarker for Metastasis-Free Survival of Breast Cancer and Immune Checkpoint Inhibition Therapy: A Retrospective Cohort Study by Chung-Wu Lin, Kai-Ming Chang, Wen-Hui Ku and Kuo-Jang Kao in Breast Cancer: Basic and Clinical Research

Supplemental Material

sj-JPG-10-bcb-10.1177_11782234251363665 – Supplemental material for Tumor Immune Response as a Biomarker for Metastasis-Free Survival of Breast Cancer and Immune Checkpoint Inhibition Therapy: A Retrospective Cohort Study

Supplemental material, sj-JPG-10-bcb-10.1177_11782234251363665 for Tumor Immune Response as a Biomarker for Metastasis-Free Survival of Breast Cancer and Immune Checkpoint Inhibition Therapy: A Retrospective Cohort Study by Chung-Wu Lin, Kai-Ming Chang, Wen-Hui Ku and Kuo-Jang Kao in Breast Cancer: Basic and Clinical Research

Supplemental Material

sj-JPG-9-bcb-10.1177_11782234251363665 – Supplemental material for Tumor Immune Response as a Biomarker for Metastasis-Free Survival of Breast Cancer and Immune Checkpoint Inhibition Therapy: A Retrospective Cohort Study

Supplemental material, sj-JPG-9-bcb-10.1177_11782234251363665 for Tumor Immune Response as a Biomarker for Metastasis-Free Survival of Breast Cancer and Immune Checkpoint Inhibition Therapy: A Retrospective Cohort Study by Chung-Wu Lin, Kai-Ming Chang, Wen-Hui Ku and Kuo-Jang Kao in Breast Cancer: Basic and Clinical Research

Supplemental Material

sj-xlsx-2-bcb-10.1177_11782234251363665 – Supplemental material for Tumor Immune Response as a Biomarker for Metastasis-Free Survival of Breast Cancer and Immune Checkpoint Inhibition Therapy: A Retrospective Cohort Study

Supplemental material, sj-xlsx-2-bcb-10.1177_11782234251363665 for Tumor Immune Response as a Biomarker for Metastasis-Free Survival of Breast Cancer and Immune Checkpoint Inhibition Therapy: A Retrospective Cohort Study by Chung-Wu Lin, Kai-Ming Chang, Wen-Hui Ku and Kuo-Jang Kao in Breast Cancer: Basic and Clinical Research

Supplemental Material

sj-xlsx-4-bcb-10.1177_11782234251363665 – Supplemental material for Tumor Immune Response as a Biomarker for Metastasis-Free Survival of Breast Cancer and Immune Checkpoint Inhibition Therapy: A Retrospective Cohort Study

Supplemental material, sj-xlsx-4-bcb-10.1177_11782234251363665 for Tumor Immune Response as a Biomarker for Metastasis-Free Survival of Breast Cancer and Immune Checkpoint Inhibition Therapy: A Retrospective Cohort Study by Chung-Wu Lin, Kai-Ming Chang, Wen-Hui Ku and Kuo-Jang Kao in Breast Cancer: Basic and Clinical Research

Supplemental Material

sj-xlsx-5-bcb-10.1177_11782234251363665 – Supplemental material for Tumor Immune Response as a Biomarker for Metastasis-Free Survival of Breast Cancer and Immune Checkpoint Inhibition Therapy: A Retrospective Cohort Study

Supplemental material, sj-xlsx-5-bcb-10.1177_11782234251363665 for Tumor Immune Response as a Biomarker for Metastasis-Free Survival of Breast Cancer and Immune Checkpoint Inhibition Therapy: A Retrospective Cohort Study by Chung-Wu Lin, Kai-Ming Chang, Wen-Hui Ku and Kuo-Jang Kao in Breast Cancer: Basic and Clinical Research

Supplemental Material

sj-xlsx-6-bcb-10.1177_11782234251363665 – Supplemental material for Tumor Immune Response as a Biomarker for Metastasis-Free Survival of Breast Cancer and Immune Checkpoint Inhibition Therapy: A Retrospective Cohort Study

Supplemental material, sj-xlsx-6-bcb-10.1177_11782234251363665 for Tumor Immune Response as a Biomarker for Metastasis-Free Survival of Breast Cancer and Immune Checkpoint Inhibition Therapy: A Retrospective Cohort Study by Chung-Wu Lin, Kai-Ming Chang, Wen-Hui Ku and Kuo-Jang Kao in Breast Cancer: Basic and Clinical Research

Supplemental Material

sj-xlsx-7-bcb-10.1177_11782234251363665 – Supplemental material for Tumor Immune Response as a Biomarker for Metastasis-Free Survival of Breast Cancer and Immune Checkpoint Inhibition Therapy: A Retrospective Cohort Study

Supplemental material, sj-xlsx-7-bcb-10.1177_11782234251363665 for Tumor Immune Response as a Biomarker for Metastasis-Free Survival of Breast Cancer and Immune Checkpoint Inhibition Therapy: A Retrospective Cohort Study by Chung-Wu Lin, Kai-Ming Chang, Wen-Hui Ku and Kuo-Jang Kao in Breast Cancer: Basic and Clinical Research

Supplemental Material

sj-xlsx-8-bcb-10.1177_11782234251363665 – Supplemental material for Tumor Immune Response as a Biomarker for Metastasis-Free Survival of Breast Cancer and Immune Checkpoint Inhibition Therapy: A Retrospective Cohort Study

Supplemental material, sj-xlsx-8-bcb-10.1177_11782234251363665 for Tumor Immune Response as a Biomarker for Metastasis-Free Survival of Breast Cancer and Immune Checkpoint Inhibition Therapy: A Retrospective Cohort Study by Chung-Wu Lin, Kai-Ming Chang, Wen-Hui Ku and Kuo-Jang Kao in Breast Cancer: Basic and Clinical Research

Footnotes

Acknowledgements

The authors wish to thank Ms. Lih-Chian Wu, Ms. Yu-Yen Bao, Ms. Pai-Gene Chen, and Ms. Li-Une Lin for their excellent technical and administrative support.

ORCID iDs

Chung-Wu Lin

Kuo-Jang Kao

Ethics Approval

The Institutional Review Board of Koo Foundation Sun Yat-Sen Cancer Center approved this study, approval number 20131209B.

Consent for Publication

The study was exempt from obtaining informed consent by the IRB.

Consent to Participate

Written informed consents were obtained from 998 patients of the validation series. All participants were adult females, no minors, aged 18 years or less, were included.

Author Contributions

Chung-Wu Lin: Formal analysis; Investigation; Methodology; Writing—original draft; Writing—review & editing.

Kai-Ming Chang: Data curation; Investigation; Methodology; Software; Writing—original draft.

Wen-Hui Ku: Investigation; Methodology; Writing—original draft.

Kuo-Jang Kao: Conceptualization; Formal analysis; Methodology; Supervision; Writing—original draft; Writing—review & editing.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported by grants from Department of Health and Welfare, Republic of China (MOHW103-TD-B-111-11, MOHW104-TDU-B-212-124-007). This study is supported in part by the Institutional Molecular Medicine Development Fund donated by Dr. CC Chen.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Availability of Data and Material

Gene expression profiles for training are deposited as GSE20685. Public data sets were used for confirmation: GSE7390, GSE2603, GSE2990, GSE2034 and GSE11121. Gene expression profiles for validation are deposited as GSE272291.

Supplemental Material

Supplemental material for this article is available online.

References

Gao

Liu

Jiang

JY.

Predictive and prognostic role of tumour-infiltrating lymphocytes in breast cancer patients with different molecular subtypes: a meta-analysis. BMC Cancer. 2020;20:1150.

Pellegrino

Hlavata

Migali

, et al. Luminal breast cancer: risk of recurrence and tumor-associated immune suppression. Mol Diagn Ther. 2021;25:409-424.

Denkert

von Minckwitz

Darb-Esfahani

, et al. Tumour-infiltrating lymphocytes and prognosis in different subtypes of breast cancer: a pooled analysis of 3771 patients treated with neoadjuvant therapy. Lancet Oncol. 2018;19:40-50.

Kao

Chang

Hsu

Huang

AT.

Correlation of microarray-based breast cancer molecular subtypes and clinical outcomes: implications for treatment optimization. BMC Cancer. 2011;11:143.

Perou

Sørlie

Eisen

, et al. Molecular portraits of human breast tumours. Nature. 2000;406:747-752.

Malkova

Sharoyko

Zhukova

Gubal

Orlova

RV.

Laboratory biomarkers of an effective antitumor immune response. Cancer Treat Res Commun. 2021;29:100489.

Charoentong

Finotello

Angelova

, et al. Pan-cancer immunogenomic analyses reveal genotype-immunophenotype relationships and predictors of response to checkpoint blockade. Cell Rep. 2017;18:248-262.

Coschi

Juergens

RA.

Overcoming resistance mechanisms to immune checkpoint inhibitors: leveraging the anti-tumor immune response. Curr Oncol. 2023;31:1-23.

Desmedt

Piette

Loi

Wang

, et al. Strong time dependence of the 76-gene prognostic signature for node-negative breast cancer patients in the TRANSBIG multicenter independent validation series. Clin Cancer Res. 2007;13:3207-3214.

10.

Minn

Gupta

Siegel

Bos

, et al. Genes that mediate breast cancer metastasis to lung. Nature. 2005;436:518-524.

11.

Sotiriou

Wirapati

Loi

Harris

, et al. Gene expression profiling in breast cancer: understanding the molecular basis of histologic grade to improve prognosis. J Natl Cancer Inst. 2006;98:262-272.

12.

Wang

Klijn

Zhang

Sieuwerts

, et al. Gene-expression profiles to predict distant metastasis of lymph-node-negative primary breast cancer. Lancet. 2005;365:671-679.

13.

Schmidt

Böhm

von Törne

Steiner

, et al. The humoral immune system has a key prognostic impact in node-negative breast cancer. Cancer Res. 2008;68:5405-5413.

14.

Xin

Shen

Zheng

, et al. Immune checkpoint inhibitors plus neoadjuvant chemotherapy in early triple-negative breast cancer: a systematic review and meta-analysis. BMC Cancer. 2021;21:1261.

15.

Brown

Loi

Immune checkpoint inhibition in the treatment of early stage triple negative breast cancer: 2021 update. Breast. 2022;62:S29-S33.

16.

Nicolini

Ferrari

Carpi

Immune checkpoint inhibitors and other immune therapies in breast cancer: a new paradigm for prolonged adjuvant immunotherapy. Biomedicines. 2022;10:2511.

17.

Hubbell

Liu

Mei

Robust estimators for expression analysis. Bioinformatics. 2002;18:1585-1592.

18.

von Elm

Altman

Egger

, et al. Strengthening the reporting of observational studies in epidemiology (STROBE) statement: explanation and elaboration. Plos Med. 2007;4:e296.

19.

Fehrenbacher

Spira

Ballinger

, et al. Atezolizumab versus docetaxel for patients with previously treated non-small-cell lung cancer (POPLAR): a multicentre, open-label, phase 2 randomised controlled trial. Lancet. 2016;387:1837-1846.

20.

Ayers

Lunceford

Nebozhyn

, et al. IFN-γ-related mRNA profile predicts clinical response to PD-1 blockade. J Clin Invest. 2017;127:2930-2940.

21.

Teschendorff

Miremadi

Pinder

Ellis

Caldas

An immune response gene expression module identifies a good prognosis subtype in estrogen receptor negative breast cancer. Genome Biol. 2007;8:R157.

22.

Cortes

Cescon

Rugo

, et al. Pembrolizumab plus chemotherapy versus placebo plus chemotherapy for previously untreated locally recurrent inoperable or metastatic triple-negative breast cancer (KEYNOTE-355): a randomised, placebo-controlled, double-blind, phase 3 clinical trial. Lancet. 2020;396:1817-1828.

23.

Rizzo

Ricci

AD.

Biomarkers for breast cancer immunotherapy: PD-L1, TILs, and beyond. Expert Opin Investig Drugs. 2022;31:549-555.

24.

Tawbi

Schadendorf

Lipson

, et al. Relatlimab and nivolumab versus nivolumab in untreated advanced melanoma. N Engl J Med. 2022;386:24-34.

25.

Huertas-Caro

Ramirez

Gonzalez-Torres

Sanabria-Salas

Serrano-Gómez

SJ.

Immune lymphocyte infiltrate and its prognostic value in triple-negative breast cancer. Front Oncol. 2022;12:910976.

26.

Fernandez-Martinez

Pascual

Singh

, et al. Prognostic and predictive value of immune-related gene expression signatures vs tumor-infiltrating lymphocytes in early-stage ERBB2/HER2-positive breast cancer: a correlative analysis of the CALGB 40601 and PAMELA trials. JAMA Oncol. 2023;9:490-499.

27.

Helmink

Reddy

Gao

, et al. B cells and tertiary lymphoid structures promote immunotherapy response. Nature. 2020;577:549-555.

28.

Blomberg

Spagnuolo

Garner

, et al. IL-5-producing CD4+ T cells and eosinophils cooperate to enhance response to immune checkpoint blockade in breast cancer. Cancer Cell. 2023;41:106-123.

29.

Virassamy

Caramia

Savas

, et al. Intratumoral CD8+ T cells with a tissue-resident memory phenotype mediate local immunity and immune checkpoint responses in breast cancer. Cancer Cell. 2023;41:585-601.

30.

Lee

Han

Choi

SY.

Molecular subtypes of triple-negative breast cancer: understanding of subtype categories and clinical implication. Genes Genomics. 2020;42:1381-1387.

31.

Thompson

Leon-Ferre

Sinnwell

, et al. Luminal androgen receptor breast cancer subtype and investigation of the microenvironment and neoadjuvant chemotherapy response. NAR Cancer. 2022;4:zcac018.

32.

Jézéquel

Lasla

Gouraud

, et al. Mesenchymal-like immune-altered is the fourth robust triple-negative breast cancer molecular subtype. Breast Cancer. 2024;31:825-840.

33.

Zhu

Tse

Wang

, et al. Immune gene expression profiling reveals heterogeneity in luminal breast tumors. Breast Cancer Res. 2019;21:147.

34.

Pardoll

DM.

The blockade of immune checkpoints in cancer immunotherapy. Nat. Rev Cancer. 2012;12:252-264.

35.

Doroshow

Bhalla

Beasley

, et al. PD-L1 as a biomarker of response to immune-checkpoint inhibitors. Nat Rev Clin Oncol. 2021;18:345-362.

36.

Sivapiragasam

Ashok Kumar

Sokol

, et al. Predictive biomarkers for immune checkpoint inhibitors in metastatic breast cancer. Cancer Med. 2021;10:53-61.

37.

Mariam

Kamath

Schveder

McLeod

Rotroff

DM.

Biomarkers for response to anti-PD-1/Anti-PD-L1 immune checkpoint inhibitors: a large meta-analysis. Oncology (Williston Park). 2023;37:210-219.

38.

Salgado

Loi

Tumour infiltrating lymphocytes in breast cancer: increasing clinical relevance. Lancet Oncol. 2018;19:3-5.

39.

Turpin

Peltonen

Rannikko

, et al. Patient-derived tumor explant models of tumor immune microenvironment reveal distinct and reproducible immunotherapy responses. Oncoimmunology. 2025;14:2466305.

40.

Zou

Zhang

Liu

Chen

Tang

Intratumoral microbiota in orchestrating cancer immunotherapy response. J Transl Int Med. 2025;12:540-542.

41.

Identifying immune-related biomarkers to predict the efficacy of cancer immunotherapy (IMMUNO-SUP). ClinicalTrials.gov. ID NCT03595813. https://clinicaltrials.gov/study/NCT03595813

42.

Shao

Xie

Chen

, et al. Inetetamab for injection in combination with vinorelbine weekly or every three weeks in HER2-positive metastatic breast cancer: a multicenter, randomized, phase II clinical trial. J Transl Int Med. 2024;12:466-477.

43.

Wolf

Yau

Wulfkuhle

, et al. Redefining breast cancer subtypes to guide treatment prioritization and maximize response: predictive biomarkers across 10 cancer therapies. Cancer Cell. 2022;40:609-623.

44.

Onkar

Carleton

Lucas

, et al. The great immune escape: understanding the divergent immune response in breast cancer subtypes. Cancer Discov. 2023;13:23-40.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.04 MB

7.60 MB

0.03 MB

0.02 MB

0.04 MB

0.03 MB

0.02 MB

0.01 MB