Sage Journals: Discover world-class research

Abstract

BACKGROUND:

The clinicopathological significance of spatial tumor-infiltrating lymphocytes (TILs) subpopulations is not well studied due to lack of high-throughput scalable methodology for studies with large human sample sizes.

OBJECTIVE:

Establishing a cyclic fluorescent multiplex immunohistochemistry (mIHC/IF) method coupled with computer-assisted high-throughput quantitative analysis to evaluate associations of six TIL markers (CD3, CD8, CD20, CD56, FOXP3, and PD-L1) with clinicopathological factors of breast cancer.

METHODS:

Our 5-plex mIHC/IF staining was shown to be reliable and highly sensitive for labeling three biomarkers per tissue section. Through repetitive cycles of 5-plex mIHC/IF staining, more than 12 biomarkers could be detected per single tissue section. Using open-source software CellProfiler, the measurement pipelines were successfully developed for high-throughput multiplex evaluation of intratumoral and stromal TILs.

RESULTS:

In analyses of 188 breast cancer samples from the Nashville Breast Health Study, high-grade tumors showed significantly increased intratumoral CD3 $+$ CD8 $+$ cytotoxic T lymphocyte density ( $P=$ 0.0008, false discovery rate (FDR) adjusted $P=$ 0.0168) and intratumoral PD-L1 expression ( $P=$ 0.0061, FDR adjusted $P=$ 0.0602) compared with low-grade tumors.

CONCLUSIONS:

The high- and low-grade breast cancers exhibit differential immune responses which may have clinical significance. The multiplexed imaging quantification strategies established in this study are reliable, cost-efficient and applicable in regular laboratory settings for high-throughput tissue biomarker studies, especially retrospective and population-based studies using archived paraffin tissues.

Keywords

Multiplex immunohistochemistry quantitative imaging analysis breast cancer tumor-infiltrating lymphocytes

Abbreviations

CI	confidence interval
CTLs	cytotoxic T lymphocytes
DFS	disease-free survival
ER	estrogen receptor
FOXP3	forkhead box protein 3
HER2	human epidermal growth factor
	receptor 2
HR	hazard ratios
mIHC/IF	fluorescence multiplex
	immunohistochemistry
NBHS	Nashville Breast Health Study
NK	natural killer cells
OS	overall survival
PD-L1	programmed death ligand 1
PR	progesterone receptor
RT	room temperature
TILs	tumor infiltrating lymphocytes
iTILs	intratumoral TILs
sTILs	stromal TILs
TMA	tissue microarray
TNM	tumor-node-metastasis
Tregs	T regulatory cells
TSA	tyramide signal amplification

1. Introduction

Tumor-infiltrating lymphocytes (TILs), important components of the tumor microenvironment, represent a local immune response against cancer. Recent studies indicate that TILs have a predictive value for immunotherapy and prognostic roles for various types of neoplasms, including triple-negative and HER2-positive breast cancers [1, 2]. Common clinical application of TILs assessment is based on pathologists’ visual scoring of hematoxylin and eosin (H&E) stained tissue slides, recommended by the International Immuno-Oncology Biomarker Working Group as biomarkers in routine histopathological reporting for breast cancer patients [3, 4]. The standardized TILs assessment guidelines are of value to researchers and practicing pathologists, owing to their ease of application in clinical trials and in daily histopathology practice, with improved consistency. However, some inherent limitations cannot be fully addressed through standardization and training. For example, H&E-based visual assessment cannot evaluate intratumoral TILs (iTILs), which have cell-to-cell contact with tumor cells; cannot differentiate various TIL subpopulations, which may differentially impact tumor progression and response to treatment [5, 6]; and have some degree of inter-reader variability so that it is considered insufficiently reproducible for clinical application [7, 8]. Conventional immunohistochemistry (IHC) as a gold standard for identifying tissue biomarkers is widely used to examine TIL subsets and immune cell markers [9, 10, 11, 12]. However, IHC only labels one or a limited number of markers per tissue section; therefore, it is difficult to apply IHC for tumor immune profiling, colocalized biomarker examination and automated image quantitative assessment [13]. Due to the lack of high-throughput scalable methodology for studies with large sample sizes, the clinicopathological significance of spatial TIL subpopulations in breast cancer molecular subtypes remains to be elucidated.

Multiplex immunohistochemistry (mIHC) technologies have emerged as powerful investigative tools to circumvent the limitations of H&E-based and conventional IHC methods. mIHC technologies allow the simultaneous detection of multiple markers on a single section to provide objective quantitative data for greater insight into disease heterogeneity, molecular profiling and biological mechanisms driving diseases [13]. Studies have shown high concordance of mIHC with conventional IHC [14] and significantly higher diagnostic accuracy in predicting clinical responses to anti-PD-1/PD-L1 therapy [15]. Moreover, mIHC conserves limited tissues when study samples are of low availability, such as rare donors, small biopsy samples or limited tissue sections collected in large population-based studies. In recent years, various mIHC techniques have been developed, such as chromogenic mIHC [16], tyramide signal amplification (TSA)-based mIHC [17, 18, 19], metal-based mIHC [20, 21], and DNA barcoding-based mIHC [22]. Each mIHC platform has some advantages, but their utilization requires certain constraints, such as time-consuming, low sensitivity, high cost, requirement of dedicated instrumentation, or lack of high-throughput image analysis [13], making it impractical for most lab settings and for large population studies.

The limitations described above motivated us to conduct this study, with aims to establish a reliable and sensitive high-plex fluorescent mIHC method, to develop an automated quantitative imaging method using open-source imaging software for high-throughput evaluation of spatial distribution of immune cell markers, and to apply the quantitative mIHC/IF approach in a pilot study to evaluate associations of TILs with clinicopathological parameters of breast cancer patients. The established mIHC/IF workflow provides a robust, cost-efficient modality, which can be conducted in regular IHC settings for deep phenotyping of the tumor and its microenvironment, especially applicable for retrospective or population-based biomarker studies using formalin-fixed paraffin-embedded (FFPE) archived tissues.

2. Materials and methods

2.1 Specimens

The human breast cancer samples evaluated in this study were from the Nashville Breast Health Study (NBHS). The NBHS is a case-control study conducted in Nashville, Tennessee, which included 2,694 breast cancer patients diagnosed between February 1, 2001 and December 31, 2011, identified through the Tennessee State Cancer Registry and five major hospitals providing medical care for breast cancer patients. The recruitment of study participants has been described elsewhere [23, 24]. Approval for this study was obtained from institutional review boards of Vanderbilt University Medical Center and other participating institutions, and all participants provided informed consent prior to enrollment in this study. In current study, we randomly selected four tissue microarray (TMA) blocks from the NBHS tissue repository, which included tumor tissue samples from 208 patients. From each donor block, representative cancer areas of three spots in the peripheral, central and middle regions were punched with a 1 mm needle and transferred into a recipient TMA block using a semi-automated tissue microarrayer. Each TMA block contained 170 tissue cores (13X13), including 52 cases of breast cancer and 14 internal control cores (eight normal human tissues of liver, spleen, kidney, lung, lymph node, prostate, brain and placenta and six breast normal and cancer cell lines: MCF7, MCF10A, MDA-MB-231, MDA-MB-361, MDA-MB-435S, and MDA-MB-453). TMA slides were sectioned at 5 $\mu$ m, covered with a thin layer of paraffin, and stored in a vacuum cabinet at 4 ${}^{\circ}$ C until use.

2.2 Fluorescent multiplex immunohistochemistry (mIHC/IF)

Two panels of 5-plex mIHC/IF staining protocols were developed to evaluate both TIL subpopulations and their spatial localizations in breast tumor tissues. The first panel included CD3 (for total T cells), CD8 (for cytotoxic T lymphocytes, i.e., CTLs), and forkhead box protein 3 (FOXP3, for regulatory T cells, i.e., Tregs). The second panel included CD20 (for B cells), CD56 (for NK cells), and the immune check-point molecule programmed death ligand 1 (PD-L1). Additionally, DAPI (for nuclei counterstain) and pan-cytokeratin (panCK, for tissue segmentation) were included in both panels. The primary antibodies of immune cell markers used in this study are commercially available and well validated [18, 19, 25, 26], i.e., rabbit anti-CD3 (Clone# SP7, #M3070, 1:1000 dilution. Spring Biosciences, CA, USA), rabbit anti-CD8 (#M5394, 1:500 dilution, Spring Biosciences), rabbit anti-PD-L1 (#13684S, 1:5000 dilution. Cell Signaling Technology, MA, USA), rabbit anti-CD56 (#156R-95, 1:3200 dilution. Cell Marque, CA, USA), mouse anti-CD20 (#GA604, 1:10 dilution. Agilent, CA, USA), and mouse anti-FOXP3 (#AB20034, 1:1000 dilution. Abcam, Cambridge, UK). The mIHC methods were established using tyramide signal amplification (TSA) technology, which allows for signal amplification of low abundance targets and facilitates heat-mediated removal of primary/secondary antibodies without disruption of TSA-fluorescence signals [17]. To optimize mIHC/IF staining protocols, we constructed a control TMA block with 13 human tumor and normal tissues, including breast, lung and colorectal cancers, and one each of spleen, prostate and placenta tissues. The staining methods are summarized below.

After deparaffinization and antigen retrieval using a high-pressure cooker with citrate buffer (pH 6.0), followed by blocking steps with 3% hydrogen peroxide solution and 5% normal goat serum, slides were subsequently incubated with the first primary antibody (mouse anti-FOXP3 or mouse anti-CD20) for 1 hour at room temperature (RT), Polymer-HRP goat anti-mouse secondary antibody (DAKO EnVision kits, Agilent, CA, USA) for 30 minutes at RT, and TSA-Cy5 solution (PerkinElmer Opal 3-plex Kit, NEL791001KT, 1:100 dilution, PerkinElmer, MA, USA) for 5 minutes at RT. Next, slides were incubated with the second primary antibody (rabbit anti-CD3 or rabbit anti-CD56) at 4 ${}^{\circ}$ C overnight. Slides were then incubated with Polymer-HRP goat anti-rabbit secondary antibody (DAKO EnVision kits), followed by CF488*TSA (1:200 dilution. Biotium, CA, USA) for 10 minutes at RT. We then performed antibody stripping with a laboratory microwave (Model# BP090, Microwave Research and Applications, Inc., IL, USA), which included placing slides in a plastic jar filled with citrate buffer, bringing the liquid to a boiling point (1 minute 35 seconds at high power), microwaving for an additional 15 minutes at low power, and allowing slides to cool to RT. The above blocking steps were then repeated. Slides were subsequently incubated with the third primary antibody (rabbit anti-CD8 or rabbit anti-PD-L1) for 1 hour at RT, Polymer-HRP goat anti-rabbit secondary antibody for 30 minutes at RT, and TSA-Cy3 solution (PerkinElmer Opal 3-plex Kit, NEL791001KT, 1:100 dilution) for 5 minutes at RT. After blocking steps as above, the slides were incubated with mouse anti-pan cytokeratin (#NBP2-29429, 1:50 dilution. Novus Biologicals, CO, USA) at 4 ${}^{\circ}$ C overnight, followed by Cy7 ${}^{\@setsize{\scriptsize}{8pt}{\viipt}{\@viipt}\textregistered}$ goat anti-mouse IgG (H $+$ L) (#16856, 1:50 dilution. AAT Bioquest Inc., CA, USA) for 1 hour at RT. Thorough washing of slides in PBS was conducted between each step. The slides were then counterstained with DAPI (1 $\mu$ g/mL for 10 minutes), cover-slipped and stored in a slide box at 4 ${}^{\circ}$ C until image acquisition.

2.3 Fluorescence quenching and cyclic mIHC/IF

Conventional immunofluorescence imaging is typically limited to 4–6 channels, above which, crosstalk between fluorophores becomes a problem for regular laboratory settings. To increase the multiplicity of immunofluorescence imaging, various fluorescence quenching methods have been developed and reported as cyclic immunofluorescence technology [27, 28, 29]. In this study, a chemical bleaching method, based on a U.S. patent (No: US 7,741,046 B2) [30], was established using TSA-cyanine dyes in our mIHC/IF protocols. Stock solutions were 1 M sodium bicarbonate (NaHCO ${}_{3}$ , FW 84.01) and 30% hydrogen peroxide. The NaHCO ${}_{3}$ solution was adjusted to pH 10 using sodium hydroxide. The fresh working solution was made by mixing 1 volume of 1 M NaHCO ${}_{3}$ , 3 volumes of distilled water, and 1 volume of 30% hydrogen peroxide. Slides were placed in the bleaching solution for 3 hours at RT under a table lamp, followed by a thorough washing with PBS. After staining and image acquisition of the first 5-plex mIHC/IF panel, a microwave treatment and chemical inactivation of fluorophores were conducted, followed by the next staining cycle using the same slide (Fig. 1A).

Figure 1.

Cyclic mIHC/IF staining of selected TIL markers. A. The cyclic 5-plex mIHC/IF detection scheme. B. Example images of sample breast cancers stained by 5-plex mIHC/IF of the first panel (a, CD3, CD8 and FOXP3) and second panel (b, CD20, CD56 and PD-L1). Total T cells (CD3 $+$ signal in green), CTLs (CD3 $+$ CD8 $+$ signal in red) and Tregs (CD3 $+$ FOXP3 $+$ signal in pink) are present mainly in the stromal area and less within the tumor area. NK cells (CD56 $+$ signal in green) are present mainly in the stromal area, and B cells (CD20 $+$ signal in red) are present in both the stromal and intratumoral areas. PD-L1 protein expression (in red) mainly appears as cytoplasmic-membrane staining diffusely in breast cancer cells in this image. C. Fluorescence inactivation and cyclic mIHC/IF staining that are repeated for four cycles of 5-plex mIHC stains and obtained very close staining results for all biomarkers.

Figure 2.

Measurement pipeline for spatial quantitative analysis of CD3 $+$ total T cells, CD3 $+$ CD8 $+$ CTLs and CD3 $+$ FOXP3 $+$ Tregs in breast cancer tissue. The pipeline is summarized as following steps: Step 1. Using “LoadImages” module to input images automatically and consecutively into the pipeline. Five grayscale 16-bit images per group from a designated input folder are loaded in sequential order (A). One TMA core of breast cancer is used as an example. (B). Step 2. Using “IdentifyPrimaryObjects” module to identify objects of cell nuclei, epithelial/cancer cells, and positive signals of FITC (CD3), Cy3 (CD8) and Cy5 (FOXP3). Step 3. Using “MaskObjects” module to keep CD8 $+$ CD3 $+$ and FOXP3 $+$ CD3 $+$ objects which define CTL cells and Tregs, respectively. Those Cy3 $+$ or Cy5 $+$ objects outside CF488 $-$ CD3 $+$ region are removed (A). Using “MeasureObjectsSizeShape” to measure several area and shape features of identified nuclei, epithelial cells, and positive objects of CD3 $+$ total T cells, CD3 $+$ CD8 $+$ CTL cells and CD3 $+$ FOXP3 $+$ Tregs; using “FilterObjects” modules to remove the objects which area values are less than mean-STD. (B). Using “RelateObjects” module to remove positive signals that are not associated with nuclei. Since cytoplasmic or membranous biomarkers are close yet outside nuclei, the “ExpandOrShrinkObjects” module is added to expand nuclei by 3 pixels followed by “RelatedObjects” module. (C). Step 4. Using “ExpandOrShrinkObjects” module to expand filtered epithelium (step 3B) by 25 pixels (9 um), then using “MaskObjects” modules to exclude expanded epithelium area and measure cell nuclei, CD3 $+$ total T cells, CD3 $+$ CD8 $+$ CTL cells and CD3 $+$ FOXP3 $+$ Tregs within stromal area. Step 5. Using “MaskObjects” modules to keep expanded epithelium area and measure cell nuclei, CD3 $+$ total T cells, CD3 $+$ CD8 $+$ CTL cells and CD3 $+$ FOXP3 $+$ Tregs of intra or peritumoral areas. Step 6. Using “CalculationMath” and “ExportToSpreadsheet” modules to produce spatial quantitative data of iTILs and sTILs and export data to an Excel file in the designated export folder.

2.4 Image acquisition and automated quantification

The mIHC stained slides were imaged using an Olympus IX-81 fluorescence fully motorized and automated multispectral slide analysis system, with filter sets specific for DAPI, GFP, CY3, CY5, and CY7. Three TMA cores per case were imaged at X 100 amplification, and exposure times were determined for each antibody stain (typically, DAPI $=$ 20 ms, GFP $=$ 200 ms, CY3 $=$ 300 ms, CY5 $=$ 750 ms, and CY7 $=$ 750 ms). Before biomarker staining, we performed DAPI/panCK double staining, followed by a pre-scan of sample tissues, to capture non-specific autofluorescence signals in the FFPE tissues. The registered images were later used to subtract autofluorescence from specific biomarker signals of stained slides. Grayscale images from the 5-plex mIHC were processed automatically to overlay each biomarker with tissue segmentation markers (pan-cytokeratin and DAPI) and stored as RGB images (16-bit tiff format) using a simple image processing pipeline with the open-source CellProfiler v3.1.5 [31]. A visual check of image quality and editing was performed by a study pathologist. Based on proposed guidelines for semi-quantification of stromal TILs in breast cancer [3], TILs around ductal carcinoma in situ and normal lobules, and those in tumor zones with crush artifacts, necrosis, and regressive hyalinization were excluded using the open-source ImageJ software. The established high-throughput spatial quantification pipelines using open-source CellProfiler for automated imaging quantification of biomarkers are described below.

For the first panel of CD3, CD8 and FOXP3 stains, the five grayscale images were input into a single integrated pipeline to quantify three biomarkers, summarized in Fig. 2. Total T cells, CTLs and Tregs were defined as CD3 $+$ , CD3 $+$ CD8 $+$ , and CD3 $+$ FOXP3 $+$ , respectively. For the second panel of CD20, CD56 and PD-L1 stains, three measurement pipelines were established for each biomarker. The 16-bit RGB images of each biomarker were loaded into its own pipeline. The main pipeline steps are summarized in Supplementary Fig. 1, taking PD-L1 as an example. TILs within the tumor cell nest (intratumoral score for iTILs) and stromal area (stromal score for sTILs) were evaluated separately. TILs within 9 $\mu$ m (25 pixels in our imaging system) to cancer cells were counted as iTILs. TIL intensity was recorded as positive count (percentage of intratumoral or stromal TIL cell count over total cell count in tumor or stromal areas) and positive density (number of TILs per 100 $\mu$ m ${}^{2}$ tumor or stroma cell nuclear area). Therefore, in addition to a total quantitative score for each biomarker, the spatial distributions of immune cell phenotypes in both intratumoral and stromal regions were also quantified. All pipelines were established using TMA controls and sample images that typically stained as negative, weak, moderate and strong positive intensity as training sets. Then, randomly selected 50 sample images were used as validation sets for necessary threshold adjustment of the “IdentifyPrimaryObjects” module until quantitative scores and pathologist’s visual scores were well matched. After imaging quantification, the study pathologist performed a quality check for about 5% of images, especially those with very high scores. Any improperly scored images with high background were rescored with a second pipeline of increased thresholding in an advanced setting of the “IdentifyPrimaryObjects” module.

2.5 Statistical analysis

Mean and standard deviation were applied to describe distributions for continuous variables and percentages of each category for categorized variables. Correlations between immune cell markers in human breast cancer tissue were examined using Spearman correlation coefficients. Cox proportional hazards models were employed to evaluate associations of selected biomarkers with clinicopathologic characteristics and overall survival (OS) of breast cancer. Tests for trends were conducted by entering the order of categorized variables as continuous variables in regression models. Multivariable Cox models included adjustments for tumor-node-metastasis (TNM) stage, tumor grade, ER and PR status, and race. All analyses were performed using Statistical Analysis Software (SAS Office Analytics, version 14.3; SAS Institute Inc., Cary, NC), with a two-sided significance threshold of $P<$ 0.05.

3. Results

3.1 Establishment of cyclic mIHC/IF and automated image quantification

The basic steps of TSA-based cyclic mIHC/IF stain are illustrated in Fig. 1A. The 5-plex staining protocols for two panels of immune cell markers were successfully established, with excellent signal-to-noise ratio (Fig. 1B). Staining protocols were optimized and validated by comparing conventional standard IHC with single fluorescent staining for each biomarker to produce specific, non-interfering, and balanced labeling for three immune cell markers of interest per slide. The sensitivity of TSA-based mIHC/IF is generally 5 to 10 times higher than standard single stains for CD3, CD8, CD20, and FOXP3, but 32 and 100 times higher for CD56 and PD-L1, respectively (Supplementary Fig. 2). After one cycle of sequential 5-plex staining, we tested chemical bleaching methods to inactivate fluorescence signals in order to perform the next cycle of 5-plex staining using the same slide. We found that TSA-FITC signal could not be fully inactivated by Lin’s (activating Alex fluorophores) [28], Gerdes’s, or Riordan’s (activating cyanine fluorophores) methods [27, 29]. Therefore, we replaced TSA-FITC with CF488*TSA (Biotium), a cyanine dye equivalent to TSA-FITC, and used a simple inactivation method based on a U.S. patent [30]; all cyanine fluorophores (Cy3, Cy5, Cy7, and CF488) could be totally inactivated in 3 hours. A very weak DAPI signal might remain, which did not affect the next biomarker staining cycle. We repeated four cycles of 5-plex mIHC stains and obtained the same staining results for all biomarkers (Fig. 1C).

After image acquisition of the stained slides, positive cells within the intratumoral, stromal and total areas of breast cancer tissues were quantified by established measurement pipelines using CellProfiler. Validation sets of 50 sample images were semi-quantitatively scored by an experienced study pathologist to compare quantitative data scored using the measurement pipelines. We found that high-quality images, defined as having contrast, bright details with a dark background, and being properly focused (Fig. 1B), could be correctly quantified by the default setting of the “IdentifyPrimaryObjects” module. However, for images with unfavorable signal-to-noise ratios, the advanced setting of the “IdentifyPrimaryObjects” module, with careful threshold adjustment, was needed to correctly differentiate true positive signals from nonspecific backgrounds for all training sets of images. In this study, our established pipelines using training sets were verified until the pathologist’s visual score and automated quantitative data reached a good consistency (Supplementary Fig. 3). The verified pipelines enabled more than 1000 images to be automatically analyzed at one time. Biomarker assessment was conducted for three TMA cores per case, then averaged for further statistical analysis. Based on our experience, a workflow of high-plex, high-throughput mIHC/IF imaging quantification workflow was established and illustrated in Fig. 3.

Figure 3.

The workflow of high-plex, high-throughput spatial quantitative analysis of tissue TIL biomarkers.

3.2 Associations of clinicopathological parameters with prognosis of breast cancer patients

Among four TMAs, a total of 188 (89.4%) cases had evaluable tissue cores and were included in our analysis (Table 1). The mean age at diagnosis was 54 years. The majority (77.7%) were diagnosed with early-stage (0-II) disease; the remaining had advanced-stage (III-VI) disease, with a 5.5 times higher risk of death. Most patients had either grade II (33.0%) or grade III (48.4%) disease. Most tumors were ER positive (71.8%), PR positive (53.2%), and HER2 negative (75.0%). Approximately 20% of patients had triple-negative breast cancer. A total of 23 patients (12.2%) died during an average of 5.45 years of follow-up.

Table 1
Characteristics of breast cancer patients in current study from the NBHS

Characteristics		Patients $N=$ 188			HR (95% CI) ${}^{*}$
Age at diagnosis	Mean $\pm$ SD	53.98 $\pm$ 10.52		per SD	1.15 (0.75–1.74)
No. of deaths		22	(11.83)
TNM stage
	0–I	83	(44.62)		ref
	II	61	(32.80)		3.11 (1.06–9.15)
	III–IV	19	(10.22)		6.61 (1.99–22.0)
	Unknown	23	(12.37)	P for trend	0.0014
Grade
	1	35	(18.82)		ref
	2	61	(32.80)		1.36 (0.35–5.28)
	3	90	(48.40)		1.60 (0.45–5.72)
				P for trend	0.4625
Race
	Black	112	(60.22)		ref
	White	65	(34.95)		1.83 (0.77–4.33)
	Other	9	(4.84)
ER
	Negative	52	(28.00)		ref
	Positive	134	(72.00)		0.62 (0.25–1.49)
PR
	Negative	87	(46.77)		ref
	Positive	99	(53.23)		0.69 (0.30–1.60)
HER2
	Negative	139	(74.73)		ref
	Positive	47	(25.27)		1.53 (0.63–3.74)
TNBC
	No	150	(80.65)		ref
	Yes	36	(19.35)		1.04 (0.35–3.12)

${}^{*}$ Adjusted for age at diagnosis.

3.3 Correlations between immune cell markers in human breast cancer tissue

Table 2
Correlation between immune cell markers in human breast cancer tissue ${}^{*}$

Breast cancer $n=$ 178	Biomarkers	CD3				CD8				FOXP3				CD20				CD56
		$r$		$P$		$r$		$P$		$r$		$P$		$r$		$P$		$r$		$P$
Total	CD8	0.	924	$<$ 0.	0001
	FOXP3	0.	799	$<$ 0.	0001	0.	7443	$<$ 0.	0001
	CD20	$-$ 0.	0083	0.	9121	0.	0016	0.	9834	$-$ 0.	0793	0.	2925
	CD56	0.	0143	0.	85	0.	0168	0.	8248	$-$ 0.	0298	0.	6941	$-$ 0.	0654	0.	3869
	PD-L1	0.	4332	$<$ 0.	0001	0.	3934	$<$ 0.	0001	0.	3497	$<$ 0.	0001	0.	1526	0.	042	0.	0542	0.	4739
Stromal	CD8	0.	912	$<$ 0.	0001
	FOXP3	0.	7405	$<$ 0.	0001	0.	6586	$<$ 0.	0001
	CD20	0.	2203	0.	0031	0.	2318	0.	0018	$-$ 0.	0061	0.	9352
	CD56	0.	0314	0.	6779	0.	0215	0.	7761	$-$ 0.	0314	0.	6779	0.	1845	0.	0139
	PD-L1	0.	4159	$<$ 0.	001	0.	4001	$<$ 0.	0001	0.	2513	0.	0007	0.	4736	$<$ 0.	0001	0.	0647	0.	3925
Intratumoral	CD8	0.	8358	$<$ 0.	0001
	FOXP3	0.	7399	$<$ 0.	0001	0.	6755	$<$ 0.	0001
	CD20	$-$ 0.	089	0.	2373	$-$ 0.	0905	0.	2297	$-$ 0.	0954	0.	205
	CD56	$-$ 0.	0019	0.	9801	$-$ 0.	02	0.	7916	0.	0069	0.	9275	$-$ 0.	1176	0.	1191
	PD-L1	0.	3974	$<$ 0.	0001	0.	3739	$<$ 0.	0001	0.	3468	$<$ 0.	0001	0.	1164	0.	1216	0.	0193	0.	799

${}^{*}$ Spearman correlation coefficients, using positive cell density data for this analysis.

Table 2 presents correlations between immune cell markers. The positive staining of CD8 and FOXP3 were significantly correlated with CD3 ( $r=$ 0.799–0.924, $P<$ 0.0001), both in intratumoral and stromal areas. PD-L1 positivity was moderately but significantly correlated with total CD3, CD8, FOXP3 and stromal CD20 ( $r=$ 0.350–0.474, $P<$ 0.0001). CD56 $+$ cell count was not significantly correlated with other selected markers.

3.4 Associations of spatial TIL subsets with clinicopathological parameters of breast cancer patients

Associations of selected biomarkers, as well as CD8 $+$ /FOXP3 $+$ ratio, with clinicopathological parameters and OS of breast cancer patients were evaluated. Compared with grade I breast cancer, grade II and III breast cancers had significantly higher intratumoral density of CD3 $+$ CD8 $+$ CTLs ( $P=$ 0.0008, false discovery rate (FDR) adjusted $P=$ 0.0168, grade II vs. I, $P=$ 0.0002, grade III vs. I, $P=$ 0.0165) and higher intratumoral PD-L1 expression ( $P=$ 0.0061, FDR adjusted $P=$ 0.0602, grade II vs. I, $P=$ 0.009, grade III vs. I, $P=$ 0.002). Other biomarkers were not significantly associated with clinicopathological parameters after FDR corrections (Table 3 and Supplementary Table 1). None of the biomarkers showed significant associations with OS in this study (Supplementary Table 2).

4. Discussion

In this study, we established a cyclic mIHC/IF method for highly multiplexed biomarker detection of FFPE tissues and developed image analysis pipelines for high-throughput automated quantification of spatial distributions of immune cell markers in tumor stromal and intratumoral compartments of breast cancer. The analyses of pilot breast cancer samples showed that higher breast cancer grades (II and III) were significantly associated with higher density of intratumoral CD3 $+$ CD8 $+$ CTLs and higher expression of intratumoral PD-L1. Although our protocol was developed based on cancer immune responses, the established quantitative mIHC/IF workflow can be readily extended to tissue imaging research outside of oncology and should benefit many basic and translational research studies.

Of several alternatives of mIHC technology and platforms, we integrated TSA, an enzyme-tyramide mediated signal amplification technology, into our mIHC/IF protocol to capitalize on the advantages of TSA, such as a much higher sensitivity than standard IHC for detection of low-abundance biomarkers, a very favorable signal-to-noise ratio, which is especially important for studies using archived FFPE tissues, the feasibility of

Table 3

Association between immune cell markers and clinicopathological parameters in breast cancer patients ${}^{*}$

Biomarker

positive

density

Event/ total

Tumor histologic grade

ER status

PR status

HER2 status

Grade I

Grade II

Grade III

P

value

Negative

Positive

P

value

Negative

Positive

P

value

Negative

Positive

P

value

CD3

+

22/186

Stromal

19.90 (1.27)

25.03 (1.19)

22.56 (1.16)

0.7241

19.70 (1.26)

24.15 (1.13)

0.479

25.58 (1.18)

20.57 (1.17)

0.403

22.53 (1.12)

23.61 (1.22)

0.840

Intratumoral

3.37 (1.29)

8.01 (1.20)

6.06 (1.17)

0.022

6.74 (1.28)

5.66 (1.14)

0.572

6.40 (1.20)

5.56 (1.18)

0.612

6.10 (1.13)

5.48 (1.24)

0.670

Total

7.75 (1.25)

11.91 (1.18)

11.66 (1.15)

0.2576

11.89 (1.24)

10.50 (1.13)

0.651

12.78 (1.17)

9.42 (1.16)

0.220

11.14 (1.11)

10.13 (1.21)

0.669

CD3

+

CD8

+

22/186

Stromal

2.19 (1.51)

4.54 (1.35)

3.63 (1.28)

0.3525

3.02 (1.48)

3.79 (1.24)

0.652

5.41 (1.34)

2.44 (1.31)

0.078

3.34 (1.22)

4.27 (1.41)

0.540

Intratumoral

0.30 (1.47)

1.88 (1.33)

0.96 (1.27)

0.0008

{}^{**}

1.07 (1.45)

0.92 (1.23)

0.754

0.98 (1.32)

0.94 (1.29)

0.936

0.89 (1.20)

1.20 (1.39)

0.432

Total

1.13 (1.37)

2.84 (1.25)

2.46 (1.21)

0.0507

2.46 (1.35)

2.14 (1.18)

0.715

2.79 (1.25)

1.82 (1.23)

0.216

2.22 (1.16)

2.25 (1.30)

0.956

CD3

+

FOXP3

+

22/186

Stromal

0.62 (1.49)

2.22 (1.33)

1.96 (1.27)

0.0257

1.23 (1.46)

1.85 (1.23)

0.392

2.40 (1.32)

1.18 (1.30)

0.105

1.68 (1.21)

1.56 (1.39)

0.859

Intratumoral

0.22 (1.43)

0.75 (1.30)

0.58 (1.24)

0.0192

0.43 (1.41)

0.57 (1.21)

0.520

0.88 (1.29)

0.33 (1.26)

0.014

0.55 (1.18)

0.45 (1.35)

0.551

Total

0.52 (1.35)

1.12 (1.24)

1.27 (1.20)

0.0441

0.91 (1.33)

1.09 (1.17)

0.624

1.54 (1.23)

0.72 (1.21)

0.021

1.11 (1.15)

0.83 (1.28)

0.306

CD20

22/186

Stromal

58.80 (2.17)

26.79 (1.75)

17.86 (1.60)

0.4494

63.25 (2.08)

17.63 (1.50)

0.169

34.70 (1.71)

19.11 (1.66)

0.476

33.36 (1.44)

11.22 (1.90)

0.147

Intratumoral

78.70 (1.79)

81/06 (1.52)

73.23 (1.42)

0.9832

56.47 (1.73)

86.64 (1.36)

0.540

121.55 (1.50)

50.57 (1.46)

0.165

88.31 (1.32)

50.38 (1.62)

0.322

Total

161.20 (1.44)

122.65 (1.30)

105.88 (1.25)

0.6412

107.96 (1.41)

125.15 (1.21)

0.737

154.86 (1.29)

95.28 (1.27)

0.221

135.51 (1.19)

83.53 (1.35)

0.174

CD56

22/186

Stromal

27.66 (2.38)

11.09 (1.87)

21.39 (1.70)

0.6116

17.56 (2.29)

18.34 (1.59)

0.967

18.45 (1.84)

17.81 (1.76)

0.970

16.99 (1.51)

21.87 (2.06)

0.766

Intratumoral

72.97 (1.62)

115.93 (1.42)

91.13 (1.34)

0.7138

221.17 (1.58)

67.25 (1.29)

0.0427

49.08 (1.40)

170.01 (1.37)

0.0189

123.08 (1.25)

43.45 (1.49)

0.0256

Total

71.47 (1.46)

132.15 (1.31)

139.90 (1.26)

0.3101

178.20 (1.43)

103.88 (1.22)

0.238

78.57 (1.30)

178.69 (1.28)

0.0465

140.48 (1.19)

78.47 (1.37)

0.1116

PD-L1

23/180

Stromal

20.98 (1.39)

32.08 (1.88)

130.34 (1.69)

0.1300

209.09 (2.28)

35.72 (1.58)

0.092

42.47 (1.83)

79.68 (1.77)

0.504

51.36 (1.51)

89.75 (2.06)

0.510

Intratumoral

54.35 (1.65)

274.56 (1.44)

378.17 (1.35)

0.0061

{}^{***}

338.97 (1.61)

207.21 (1.30)

0.414

213.81 (1.42)

263.06 (1.39)

0.703

207.28 (1.27)

361.61 (1.52)

0.254

Total

83.03 (1.56)

225.50 (1.38)

446.21 (1.31)

0.0086

285.80 (1.52)

253.54 (1.27)

0.823

256.17 (1.36)

268.05 (1.34)

0.925

242.72 (1.23)

330.64 (1.45)

0.476

CTL/treg ratio

22/186

Stromal

1.53 (1.35)

1.85 (1.22)

1.66 (1.18)

0.8474

1.64 (1.31)

1.73 (1.16)

0.879

2.06 (1.21)

1.41 (1.21)

0.221

1.65 (1.14)

1.87 (1.26)

0.646

Intratumoral

0.92 (1.37)

1.66 (1.25)

1.50 (1.20)

0.2880

1.35 (1.33)

1.46 (1.18)

0.823

1.25 (1.23)

1.62 (1.23)

0.432

1.35 (1.15)

1.69 (1.29)

0.443

Total

2.15 (1.24)

2.05 (1.17)

1.80 (1.14)

0.7296

1.85 (1.23)

1.98 (1.12)

0.797

2.02 (1.16)

1.87 (1.15)

0.741

1.83 (1.11)

2.33 (1.19)

0.238

${}^{*}$ Adjusted for age at diagnosis, race, TNM stage, tumor grade, and ER/PR positivity. The quantitative data for all markers presented as mean (SE). ${}^{**}$ The false discovery rate (FDR) adjusted $P=$ 0.0168, grade II vs. I, $P=$ 0.0002, grade III vs. I, $P=$ 0.0165. ${}^{***}$ The FDR adjusted $P=$ 0.0602, grade II vs. I, $P=$ 0.009, grade III vs. I, $P=$ 0.002.

repetitive antibody stripping for the sequential mIHC staining due to covalently bound TSA-fluorophores to targeted epitopes, low cost and ease of performance in regular IHC settings to provide quantitatively reproducible batch-to-batch multiplexed data [17]. Furthermore, we established a chemical quenching method to inactivate TSA-cyanine fluorophore signals for repetitive cycles of 5-plex mIHC staining on the same slides, i.e., cyclic mIHC/IF. We conducted four cycles for two panels of 5-plex mIHC/IF staining using a TMA control slide, showing consistent results of all selected biomarkers without increased nonspecific background, loss of target antigens or tissue integrity. We were able to perform additional mIHC/IF on the same tissue sections, as a previous study reported that by iterative staining of a single slide, up to 60 or more biomarkers [27] could be assessed. It is important to note that obtaining satisfying staining results requires diligent optimization regarding antigen retrieval, balanced primary and secondary antibody concentrations, complete antibody stripping and fluorophore inactivation, as well as proper ways of preventing tissue detachment from glass slides [32] and storing unstained tissue sections [33].

Before applying automated imaging quantification, two important issues need to be addressed to ensure high quality images for subsequent analysis. The first is the removal of autofluorescence, which may prevent the clean visualization of specific signals of fluorophores Cy3 and CF488/Cy2/FITC, particularly using FFPE tissues [34]. In this study, we applied a “two-step method”: before the first mIHC/IF cycle, an automated multichannel scanning was performed on DAPI/panCK stained slides, and at the end of mIHC/IF staining and image acquisition, the autofluorescence background was subtracted using the registered pre-scanning images. A fully motorized and automated multispectral fluorescence microscope analysis system was needed for this purpose. Other autofluorescence removal methods may also be used, according to laboratory settings [35]. The second issue was editing images to exclude areas of ductal carcinoma in situ, normal lobules, and other areas unsuitable for analysis in order to fulfill international guidelines on TIL assessment in breast cancer [3, 4]. In this study, more than 80% of images did not need editing, although editing can be efficiently performed by a visual check using ImageJ software. The majority of images showed good quality. A small amount of slightly blurred (out-of-focus) images were found, which did not lower the specific signal scores. However, some images did show unfavorable signal-to-noise ratios, which may have been due to prolonged autoexposure time for the images with weak positive signals. The key to correctly quantifying these images was using the advanced setting on the “IdentifyPrimaryObjects” module, in which thresholds were carefully verified using both training and validation sets to differentiate true positive signals from nonspecific background. A post-analytical quality check of 5% of images, especially those having very high scores, was the final phase of imaging quantification, to correct potential errors from low quality images before releasing data for statistical analyses.

Computer-aided imaging quantitation methods have been demonstrated to produce highly similar data to pathologists’ visual evaluation, with advantages of improved accuracy (a greater degree of objectivity and reproducibility), efficiency and multiplexed evaluation by mapping two or more biomarkers [36, 37]. In this study, the measurement pipelines using CellProfiler were developed for each target. In the integrated pipeline for spatial quantifications of CD3, CD8 and FOXP3 (Fig. 2), the “MaskObjects” module was used to define CTLs by masking CD3 $+$ and CD8 $+$ signals, and Tregs by masking CD3 $+$ and FOXP3 $+$ signals (multiplexed evaluation). To exclude potential nonspecific staining signals, several image processing modules were built into the pipelines: “MeasureObjectsSizeShape” followed by “FilterObjects” modules were used to remove objects for which area values were less than mean-STD, and the “RelateObjects” module was used to remove positive signals not associated with nuclei. We used the “ExpandOrShrinkObjects” module to expand filtered panCK $+$ tumor areas by 25 pixels (9 um), followed by the “MaskObjects” module to include or exclude expanded panCK $+$ areas for stromal or intratumoral TIL measurements. For high quality images with favorable signal-to-noise ratios, pipelines with the default setting of “IdentifyPrimaryObjects” module yielded objective, reproducible quantification data that generalized well across data sets. If images showed a background with unfavorable signal-to-noise ratios, a modified pipeline with advanced settings of the “IdentifyPrimaryObjects” module was applied to carefully determine threshold values for intensity as a means of defining “positive” or “negative” cells for each marker. If images with non-uniformities in illumination or out-of-focus existed, the correct illumination modules and quality control pipelines [38] should be applied.

In the current study, higher densities of intratumoral CTLs were observed in higher grade (II, III) than lower grade breast cancers, consistent with other studies [11, 12, 39], suggesting a differential immune response between high- and low-grade breast cancers. Intriguingly, a significantly elevated PD-L1 was observed in high-grade breast cancers and was correlated with infiltration of total T, CTLs, Tregs, and stromal B cells ( $P<$ 0.0001). Similar findings have been reported by other study groups [40, 41]. PD-L1 can be expressed by all human cells. The PD-1/PD-L1 pathway plays a subtle role in maintaining peripheral T-lymphocyte tolerance and seems to be one of the major immune escape mechanisms, inversely correlated with survival in breast cancer [42, 43]. The heterogenic expression of PD-L1 in breast cancer is reported to be positively associated with poor prognosis, with clinicopathological features such as young patients’ age, large tumor size, high proliferative rate, and aggressive molecular subtypes (HER2 $+$ and triple-negative breast cancers), in addition to high histologic grade [41]. Elevated PD-L1 expression partially explains unfavorable features of high-grade tumors. There is evidence that low- and high-grade breast cancers exhibit distinct molecular profiles at genomic, transcriptomic, and immunohistochemical levels [44, 45, 46]. The current study provides evidence that the adaptive immune responses between high- and low-grade tumors may also be different. Using a few immune biomarkers, multivariate analyses, and taking other well-known factors into account are required to determine the true values of immune-related biomarkers [47]. Interestingly, we found that correlations between PD-L1 and most of the immune cell characterizations in this study (except FOXP3) were higher in stromal than intratumoral components. These results are consistent with the infiltration direction of inflammatory cells and gradient immune responses; i.e., leukocytes migrate from peripheral interstitial blood vessels to the tumor site. FOXP3-positive Tregs can regulate a variety of immune cell functions [48]. In this study, we found that the correlation between PD-L1 and FOXP3 expression in intratumoral tissue was higher than that in stromal tissue, which is likely to represent that the Tregs response mainly plays a role when immune cells migrate into the tumor cell nest.

In conclusion, this study successfully established a highly sensitive, multiplex and high-throughput mIHC/IF imaging quantitative evaluation workflow for spatial quantifications of tissue immune cell markers. The mIHC/IF quantitative imaging techniques are applicable in regular IHC settings and are cost-efficient and readily extended to benefit many basic and translational research studies, especially for retrospective and population-based biomarker studies using FFPE tissues. The pilot study provides evidence that immune responses between high- and low-grade breast cancers may be different. Further larger population studies of comprehensive molecular and immune response profiling to elucidate critical roles of immune-related biomarkers in cancer prognosis and treatment response are warranted.

Author contributions

Conception: T.S., S.W. and Q.C.

Interpretation or analysis of data: H.C., T.S., S.W., A.B.F. and Q.C.

Preparation of the manuscript: T.S., S.W., A.B.F., E.T.M. and S.H.

Revision for important intellectual content: Q.C., W.Z. and X.O.S.

Supervision: W.Z., X.O.S. and Q.C.

Data availability

The data and image measurement pipelines with details of modules and settings used in this study are available from corresponding authors on reasonable request.

Funding

NBHS is supported by the NIH grant R01CA100374. This work was supported, in part, by UL1TR002243 (VICTR IDs: VR19880 and VR53783) from the National Center for Advancing Translational Sciences. NHBS data and tissue sample collection, TMA construction, fluorescent mIHC staining, and imaging quantitative analysis were conducted in the Survey and Biospecimen Shared Resource, which is supported in part by the Vanderbilt-Ingram Cancer Center (P30CA68485).

Ethics approval and consent to participate

The Nashville Breast Health Study (NBHS) was approved by the institutional review boards of Vanderbilt University Medical Center and the other participating institutions, and all participants provided informed consent prior to enrollment in this study.

Supplementary data

The supplementary files are available to download from http://dx.doi.org/10.3233/CBM-220071.

sj-docx-1-cbm-10.3233_CBM-220071.docx - Supplemental material

Supplemental material, sj-docx-1-cbm-10.3233_CBM-220071.docx

Footnotes

Acknowledgments

Dr. Robert Coffey’s lab at VUMC kindly provided technical support for the establishment of 5-plex fluorescent mIHC and TMA image capturing. We also thank Ms. Nabela Hamm for tissue processing of the NBHS samples, Dr. Mary Shannon Byers, Ms. Rachel Mullen, and Ms. Kathleen Harmeyer for assistance in editing and submitting this manuscript.

Conflict of interest

All the authors declare that they have no conflict of interest.

References

Savas

Salgado

Denkert

et al., Clinical relevance of host immunity in breast cancer: From TILs to the clinic, Nat Rev Clin Oncol 13 (2016), 228–241.

Stanton

S.E.

and Disis

M.L.

, Clinical significance of tumor-infiltrating lymphocytes in breast cancer, J Immunother Cancer 4 (2016), 59.

Hendry

Salgado

Gevaert

et al., Assessing Tumor-infiltrating Lymphocytes in Solid Tumors: A Practical Review for Pathologists and Proposal for a Standardized Method From the International Immunooncology Biomarkers Working Group: Part 1: Assessing the Host Immune Response, TILs in Invasive Breast Carcinoma and Ductal Carcinoma In Situ, Metastatic Tumor Deposits and Areas for Further Research, Adv Anat Pathol 24 (2017), 235–251.

Dieci

M.V.

Radosevic-Robin

Fineberg

et al., Update on tumor-infiltrating lymphocytes (TILs) in breast cancer, including recommendations to assess TILs in residual disease after neoadjuvant therapy and in carcinoma in situ: A report of the International Immuno-Oncology Biomarker Working Group on Breast Cancer, Semin Cancer Biol 52 (2018), 16–25.

Pages

Galon

Dieu-Nosjean

M.C.

et al., Immune infiltration in human tumors: A prognostic factor that should not be ignored, Oncogene 29 (2010), 1093–1102.

Verma

Hanby

A.M.

Horgan

et al., Levels of different subtypes of tumour-infiltrating lymphocytes correlate with each other, with matched circulating lymphocytes, and with survival in breast cancer, Breast Cancer Res Treat 183 (2020), 49–59.

O’Loughlin

Andreu

Bianchi

et al., Reproducibility and predictive value of scoring stromal tumour infiltrating lymphocytes in triple-negative breast cancer: A multi-institutional study, Breast Cancer Res Treat 171 (2018), 1–9.

Kos

Roblin

Kim

R.S.

et al., Pitfalls in assessing stromal tumor infiltrating lymphocytes (sTILs) in breast cancer, NPJ Breast Cancer 6 (2020), 17.

West

N.R.

Kost

S.E.

Martin

S.D.

et al., Tumour-infiltrating FOXP3 (+) lymphocytes are associated with cytotoxic immune responses and good clinical outcome in oestrogen receptor-negative breast cancer, Br J Cancer 108 (2013), 155–162.

10.

Ali

H.R.

Provenzano

Dawson

S.J.

et al., Association between CD8+ T-cell infiltration and breast cancer survival in 12,439 patients, Ann Oncol 25 (2014), 1536–1543.

11.

Miyoshi

Shien

Ogiya

et al., Associations in tumor infiltrating lymphocytes between clinicopathological factors and clinical outcomes in estrogen receptor-positive/human epidermal growth factor receptor type 2 negative breast cancer, Oncol Lett 17 (2019), 2177–2186.

12.

Liu

Lang

Zhao

et al., CD8 (+) cytotoxic T cell and FOXP3 (+) regulatory T cell infiltration in relation to breast cancer survival and molecular subtypes, Breast Cancer Res Treat 130 (2011), 645–655.

13.

Tan

W.C.C.

Nerurkar

S.N.

Cai

H.Y.

et al., Overview of multiplex immunohistochemistry/immunofluorescence techniques in the era of cancer immunotherapy, Cancer Commun (Lond) 40 (2020), 135–153.

14.

Yeong

Tan

Chow

Z.L.

et al., Multiplex immunohistochemistry/immunofluorescence (mIHC/IF) for PD-L1 testing in triple-negative breast cancer: A translational assay compared with conventional IHC, J Clin Pathol 73 (2020), 557–562.

15.

Stein

J.E.

Rimm

D.L.

et al., Comparison of Biomarker Modalities for Predicting Response to PD-1/PD-L1 Checkpoint Blockade: A Systematic Review and Meta-analysis, JAMA Oncol 5 (2019), 1195–1204.

16.

Morrison

L.E.

Lefever

M.R.

Behman

L.J.

et al., Brightfield multiplex immunohistochemistry with multispectral imaging, Lab Invest 100 (2020), 1124–1136.

17.

Stack

E.C.

Wang

Roman

K.A.

et al., Multiplexed immunohistochemistry, imaging, and quantitation: A review, with an assessment of Tyramide signal amplification, multispectral imaging and multiplex analysis, Methods 70 (2014), 46–58.

18.

Parra

E.R.

Uraoka

Jiang

et al., Validation of multiplex immunofluorescence panels using multispectral microscopy for immune-profiling of formalin-fixed and paraffin-embedded human tumor tissues, Sci Rep 7 (2017), 13380.

19.

Halse

Colebatch

A.J.

Petrone

et al., Multiplex immunohistochemistry accurately defines the immune context of metastatic melanoma, Sci Rep 8 (2018), 11158.

20.

Angelo

Bendall

S.C.

Finck

et al., Multiplexed ion beam imaging of human breast tumors, Nat Med 20 (2014), 436–442.

21.

Levenson

R.M.

Borowsky

A.D.

and Angelo

, Immunohistochemistry and mass spectrometry for highly multiplexed cellular molecular imaging, Lab Invest 95 (2015), 397–405.

22.

Goltsev

Samusik

Kennedy-Darling

et al., Deep profiling of mouse splenic architecture with CODEX multiplexed imaging, Cell 174 (2018), 968–981.

23.

Cui

Deming-Halverson

S.L.

Shrubsole

M.J.

et al., Use of nonsteroidal anti-inflammatory drugs and reduced breast cancer risk among overweight women, Breast Cancer Res Treat 146 (2014), 439–446.

24.

Han

M.R.

Deming-Halverson

Cai

et al., Evaluating 17 breast cancer susceptibility loci in the Nashville breast health study, Breast Cancer 22 (2015), 544–551.

25.

Steele

K.E.

Tan

T.H.

Korn

et al., Measuring multiple parameters of CD8+ tumor-infiltrating lymphocytes in human cancers by image analysis, J Immunother Cancer 6 (2018), 20.

26.

Jasinski-Bergner

Stoehr

Bukur

et al., Clinical relevance of miR-mediated HLA-G regulation and the associated immune cell infiltration in renal cell carcinoma, Oncoimmunology 4 (2015), e1008805.

27.

Gerdes

M.J.

Sevinsky

C.J.

Sood

et al., Highly multiplexed single-cell analysis of formalin-fixed, paraffin-embedded cancer tissue, Proc Natl Acad Sci U S A 110 (2013), 11982–11987.

28.

Lin

J.R.

Fallahi-Sichani

and Sorger

P.K.

, Highly multiplexed imaging of single cells using a high-throughput cyclic immunofluorescence method, Nat Commun 6 (2015), 8390.

29.

Riordan

D.P.

Varma

West

R.B.

et al., Automated analysis and classification of histological tissue features by multi-dimensional microscopic molecular profiling, PLoS One 10 (2015), e0128975.

30.

Melinda Larsen, Anup Sood Gerdes

M.J.

, Michael Christopher Montalto, Zhengyu Pang and Fiona Ginty, Sequencial analysis of biological samples, General Electric Company, Niskayuna NY US, (US 7,741,046 B2), 2010.

31.

Kamentsky

Jones

T.R.

Fraser

et al., Improved structure, function and compatibility for CellProfiler: Modular high-throughput image analysis software, Bioinformatics 27 (2011), 1179–1180.

32.

Gambella

Porro

Pigozzi

et al., Section detachment in immunohistochemistry: Causes, troubleshooting, and problem-solving, Histochem Cell Biol 148 (2017), 95–101.

33.

Shrubsole

M.J.

Ness

R.M.

et al., Immunohistochemical expressions of Ki-67, cyclin D1, beta-catenin, cyclooxygenase-2, and epidermal growth factor receptor in human colorectal adenoma: A validation study of tissue microarrays, Cancer Epidemiol Biomarkers Prev 15 (2006), 1719–1726.

34.

Davis

A.S.

Richter

Becker

et al., Characterizing and diminishing autofluorescence in formalin-fixed paraffin-embedded human respiratory tissue, J Histochem Cytochem 62 (2014), 405–423.

35.

Woolfe

Gerdes

Bello

et al., Autofluorescence removal by non-negative matrix factorization, IEEE Trans Image Process 20 (2011), 1085–1093.

36.

Rizzardi

A.E.

Johnson

A.T.

Vogel

R.I.

et al., Quantitative comparison of immunohistochemical staining measured by digital image analysis versus pathologist visual scoring, Diagn Pathol 7 (2012), 42.

37.

Meyerholz

D.K.

and Beck

A.P.

, Principles and approaches for reproducible scoring of tissue stains in research, Lab Invest 98 (2018), 844–855.

38.

Bray

M.A.

and Carpenter

A.E.

, Quality control for high-throughput imaging experiments using machine learning in cellprofiler, Methods Mol Biol 1683 (2018), 89–112.

39.

Goff

S.L.

and Danforth

D.N.

, The role of immune cells in breast tissue and immunotherapy for the treatment of breast cancer, Clin Breast Cancer 21 (2021), e63–e73.

40.

Sabatier

Finetti

Mamessier

et al., Prognostic and predictive value of PDL1 expression in breast cancer, Oncotarget 6 (2015), 5449–5464.

41.

Bertucci

and Goncalves

, Immunotherapy in Breast Cancer: The Emerging Role of PD-1 and PD-L1, Curr Oncol Rep 19 (2017), 64.

42.

Okazaki

and Honjo

, PD-1 and PD-1 ligands: From discovery to clinical application, Int Immunol 19 (2007), 813–824.

43.

Huang

Ran

Shao

et al., Prognostic and clinicopathological value of PD-L1 expression in primary breast cancer: A meta-analysis, Breast Cancer Res Treat 178 (2019), 17–33.

44.

Lee

C.H.

Tan

P.H.

et al., A molecular signature of the Nottingham prognostic index in breast cancer, Cancer Res 64 (2004), 2962–2968.

45.

Sotiriou

Wirapati

Loi

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

46.

Ignatiadis

and Sotiriou

, Understanding the molecular basis of histologic grade, Pathobiology 75 (2008), 104–111.

47.

Rakha

E.A.

Reis-Filho

J.S.

Baehner

et al., Breast cancer prognostic classification in the molecular era: The role of histological grade, Breast Cancer Res 12 (2010), 207.

48.

Wing

J.B.

Tanaka

and Sakaguchi

, Human FOXP3 (+) Regulatory T Cell Heterogeneity and Function in Autoimmunity and Cancer, Immunity 50 (2019), 302–316.

Multiplex immunohistochemistry and high-throughput image analysis for evaluation of spatial tumor immune cell markers in human breast cancer

Abstract

BACKGROUND:

OBJECTIVE:

METHODS:

RESULTS:

CONCLUSIONS:

Keywords

Abbreviations

1. Introduction

2. Materials and methods

2.1 Specimens

2.2 Fluorescent multiplex immunohistochemistry (mIHC/IF)

2.3 Fluorescence quenching and cyclic mIHC/IF

2.5 Statistical analysis

3. Results

3.1 Establishment of cyclic mIHC/IF and automated image quantification

Table 1 Characteristics of breast cancer patients in current study from the NBHS

Table 2 Correlation between immune cell markers in human breast cancer tissue *

4. Discussion

Author contributions

Data availability

Funding

Ethics approval and consent to participate

Supplementary data

sj-docx-1-cbm-10.3233_CBM-220071.docx - Supplemental material

Footnotes

Acknowledgments

Conflict of interest

References

Table 1
Characteristics of breast cancer patients in current study from the NBHS

Table 2
Correlation between immune cell markers in human breast cancer tissue ${}^{*}$