Immunologic heterogeneity of tumor infiltrating lymphocyte composition in primary melanoma

Abstract

Tumor infiltrating lymphocytes (TILs) in primary melanomas are thought to represent the host anti-tumor immune response, but controversy exists over whether TILs offer independent prognostication of survival. We studied a cohort of 1,241 primary melanoma patients to assess the association of absent, non-brisk, and brisk TIL grade with survival outcomes. We tested whether quantitative TIL counts using immunohistochemical lymphocyte markers CD3, CD45, and FOXP3 add prognostic value to TIL grading compared to histology alone in 15% of the cohort. To assess for inter-group immunologic heterogeneity among TIL grades, we investigated differential expression of 594 immunoregulatory genes in 67 primary melanomas. On histologic evaluation of 1,241 primary melanomas, TILs were graded as absent (n=388, 31%), non-brisk (n=330, 27%), and brisk (n=523, 42%). Patients with brisk TILs had improved recurrence-free survival (RFS) (p=0.025) and overall survival (OS) (p=0.006) compared to patients with non-brisk and absent TILs, for which there were no differences in RFS (p=0.40) or OS (p=0.41). TIL quantitation by immunohistochemistry did not improve prognostication compared to TIL grading on hematoxylin and eosin stained sections. Melanomas with non-brisk and absent TILs share similar immunoregulatory gene expression profiles. In contrast, melanomas with brisk TILs demonstrate upregulation of T-cell activation pathways and inhibition of upstream immune checkpoint regulators. The presence of TILs in primary melanomas represents a heterogeneous group and caution in prognostic interpretation is warranted. Melanomas with brisk TILs are defined by an immunostimulatory gene expression profile and improved prognosis compared to melanomas with non-brisk or absent TILs.

1. INTRODUCTION

Melanoma incidence in the United States is rising. There were 73,000 new cases of melanoma diagnosed in 2015 and over 9,000 deaths (1). The American Joint Committee on Cancer’s (AJCC) melanoma staging system is the mainstay of clinical prognostication in melanoma and is dependent upon primary tumor thickness, ulceration, mitoses, and nodal status (2-6). Despite the AJCC’s survival estimates by stage, clinical behavior of melanoma is often unpredictable. Additional prognostic markers are warranted to more precisely identify early stage, high risk melanoma patients in jeopardy of recurrence and early death (5, 7). This will aid in modification of clinical surveillance strategies and improve patient selection for adjuvant therapy administration.

Tumor-infiltrating lymphocytes (TILs) have been identified as a favorable prognostic marker across multiple cancers including breast, ovarian, and colon cancer and may serve as a surrogate indicator of the strength of the host anti-tumor immune response (8-10). While multiple studies have assessed the prognostic value of TILs in primary melanoma (11-19), pooled results from the literature remain controversial. Several studies show no association of TIL presence and improved melanoma prognosis (13, 15-17), while the studies that do show a significant correlation (11, 12, 14, 18, 19) are difficult to draw conclusions from due to heterogeneity of patient populations, research methods, and TIL grading systems used. TIL assessment also remains a semi-quantitative measure that may be challenging to consistently reproduce. To date, no cancer staging algorithms incorporate immunologic markers.

Importantly, there is also a lack of understanding of the diverse composition of the TIL infiltrate, comprised of effector and suppressor T cells, B cells, natural killer cells, macrophages, dendritic cells, and myeloid-derived suppressor cells, each of which impart competing immunostimulatory or immunosuppressive effects within the tumor microenvironment. For example, tumor-associated T cells may be immunosuppressive (i.e. T regulatory cells) or anergic and thus their presence does not ensure an effective anti-tumor immune response. Therefore, the functional impact of TILs within the tumor microenvironment, rather than their mere presence or absence, is an important consideration that has not been significantly addressed in the literature.

The primary objectives of this study are to identify patterns of immunologic heterogeneity that exist among progressing levels of TIL intensity within primary melanomas and how these profiles impact patient prognosis. We test whether quantitative immunohistochemistry (IHC) compared to TIL grading by histology alone allows for more precise prognostic predictions. We also compare immunoregulatory gene expression profiles of primary melanomas to explore whether an immunologic signature may functionally explain survival differences amongst melanomas of each TIL grade.

2. MATERIALS AND METHODS

2.1 Study Population

A cohort of patients with primary cutaneous melanoma presenting to New York University (NYU) Langone Medical Center from 2002-2013 were enrolled on the Interdisciplinary Melanoma Cooperative Group (IMCG) biospecimen database protocol. This protocol is approved by the NYU institutional review board (IRB), prospectively enrolls melanoma patients presenting to surgical and medical oncologists at the NYU Perlmutter Cancer Center, and banks patient tissue specimens for research purposes, which are linked to the clinicopathologic database. Patient demographics, and clinical, pathologic, and follow-up information are prospectively recorded and updated on a regular basis as per the standards outlined in our protocol. Informed consent for use of clinical data and tissue was obtained from all patients at the time of enrollment. Patients with multiple primary melanomas were excluded from this study.

2.2 Histopathological Review

Histopathologic features from hematoxylin and eosin (H&E) stained sections for all melanoma cases were determined by the IMCG pathologist (FD) including: thickness, ulceration, mitoses, histologic subtype, regression, vascular invasion, perineural invasion, microsatellites, and solar elastosis. TILs were graded as brisk (present throughout the vertical growth phase (VGP) or infiltrating the entire base of the VGP), non-brisk (present in one or more foci of the VGP), or absent (no lymphocytes are in contact with the VGP but may be present in perivascular or fibrotic areas), as previously defined (12). Patients who underwent sentinel lymph node (SLN) biopsy had SLN status recorded as positive or negative.

2.3 Immunohistochemical Analysis

Primary melanomas representing 15% of the entire cohort were randomly selected to undergo immunohistochemical staining for CD3, a pan T-cell marker and T-cell co-receptor, and CD45, a transmembrane glycoprotein expressed on leukocytes. Polyclonal rabbit anti-human CD3 and monoclonal mouse anti-human CD45 (Dako North America, Inc., Carpinteria, CA) were used for TIL identification. Monoclonal mouse anti-human Foxp3 clone 236A/E7 (eBioscience, San Diego, CA) was used to detect T regulatory cells (Tregs) in the same cohort.

Formalin fixed paraffin embedded (FFPE) primary melanomas were deparaffinized, rehydrated, and subjected to heat-induced epitope retrieval in citrate buffer in a microwave. Sections were quenched with hydrogen peroxidase, incubated in blocking solution (Vectastain Elite ABC Kit; rabbit IgG for CD3; mouse IgG for both CD45 and Foxp3, Vector Laboratories, Burlingame, CA), and then incubated in primary antibody diluted in blocking solution at a concentration of 1:500 for 1 hour at room temperature and at 4° C overnight. Sections were washed and incubated in a 1:200 dilution of biotinylated antibody solution (goat anti-rabbit for CD3; horse anti-mouse for both CD45 and Foxp3) for 1 hour and incubated with Vectastain ABC reagent. Avidin-biotinylated horseradish peroxidase complexes were visualized with diaminobenzidine (DAB substrate kit, Vector Laboratories, Burlingame, CA). Slides were counterstained with hematoxylin, dehydrated, and mounting medium and coverslips were applied.

The IMCG pathologist was blinded to patient clinicopathologic data. He quantitatively scored CD3 and CD45 expression as the absolute number of intratumoral TILs stained in the densest infiltrative area in one high power field (HPF) and the absolute number of peritumoral TILs stained in the dermis immediately beyond the tumor-dermis interface. Cut-off for quantitative scoring was set at TIL counts of greater than 50 cells. Foxp3 expression was scored by identifying the densest area of antibody staining and quantifying the number of Foxp3-positive cells in a single HPF within the lymphocytic infiltrate. Percentage of Foxp3-stained cells within the lymphocytic infiltrate was estimated.

2.4 Gene Expression

RNA was isolated from macrodissection of 67 FFPE primary melanoma sections using the Rneasy FFPE Kit (Qiagen, Valencia, CA) per manufacturer protocol (20). After RNA was quantified and subjected to quality control measures, gene expression analysis was conducted using Nanostring technology per manufacturer protocol (21). The selected genes consisted of a customized NCounter® GX Human Immunology Kit (Nanostring Technologies, Seattle, WA) comprised of 594 immunoregulatory genes, related to 24 immunology-related gene networks compiled from the Gene Ontology Consortium list of immunologically important genes.

2.4.1 Pathway Analysis

Gene expression data were analyzed using Ingenuity Pathway Analysis (IPA) software (Qiagen, Redwood City, CA) to identify immunologic profiles characteristic of melanomas of each TIL grade (22). Information from the core analysis function including top canonical pathways, upstream regulators, regulator effects, and gene networks with annotated disease states and functions were integrated to identify the most influential immunoregulatory pathways that defined each TIL grade.

2.5 Statistical Analysis

TIL grade and association with clinicopathologic characteristics was assessed using the Chi square test and the non-parametric test. Specifically, the p-values in Table 1 were calculated as follows: if the variable had categorical data, the chi-squared test was performed; if the variable had continuous data, the Kruskal-Wallis Rank Sum Test for median and the Analysis of Variance (ANOVA) for mean was performed. Survival curves stratified by TIL grade and quantitative IHC values were made using the Kaplan-Meier method and compared using the log-rank test. Hazard ratios and 95% confidence intervals (CIs) for recurrence-free survival (RFS), overall survival (OS), and post-recurrence survival were estimated using Cox regression models. Multivariate analyss controlling for primary tumor thickness, ulceration, and mitoseswas performed to assess the independent impact of TIL grade on survival outcomes. Sensitivity and specificity were calculated to determine the ability of quantitative CD3 and CD45 TIL counts to reflect the initial TIL grade as determined on H&E examination. Gene expression data was stratified by TIL grade using the t-test. Fold changes were calculated and a p-value<0.05 was used in IPA.

Table 1.

Characteristics of 1,241 primary cutaneous melanomas by TIL grade

	Absent (n=388)		Non-brisk (330)		Brisk (523)		p-value
	No.	%	No.	%	No.	%
Gender
Female	183	47.2	142	43	230	43.9	0.49
Male	205	52.8	188	57	293	56.1
Age							0.1
Median	60.6		61.1		58.6
Mean	58.8		60.1		57.4
Thickness
≤ 1mm	238	61.3	153	46.4	323	61.8	0.002
> 1mm	150	38.7	177	53.6	200	38.2
Ulceration							<0.001
Absent	338	87.3	245	74.2	437	83.6
Present	49	12.7	85	25.8	86	16.4
Unknown	1		0		0
Mitoses							<0.001
Absent	198	51.2	117	35.9	220	42.1
Present	189	48.8	209	64.1	303	57.9
Unknown	1		4		0
Histologic Subtype							<0.001
Nodular	64	17.1	85	27.6	109	22.8
Superficial Spreading	243	65	171	56.3	319	66.7
Other	67	17.9	48	16.1	50	10.5
Unclassified	14		26		45
SLN biopsy							<0.001
Not performed	228	59.1	146	44.6	274	53
Performed	158	40.9	181	55.4	243	44
SLN biopsy							0.49
Negative	124	90.7	290	88.7	470	91.1
Positive	34	9.3	37	11.3	46	8.9
Other	2		3		6

3. RESULTS

3.1 Non-brisk TIL grade was associated with the worst pathologic characteristics

Clinicopathological features for 1,241 primary melanoma patients stratified by TIL grade are shown in Table 1. TILs were graded as absent (n= 388, 31%), non-brisk (n= 330, 27%), and brisk (n= 523, 42%) on H&E. Non-brisk TIL grade had the thickest tumors (mean = 2.11 mm, p=0.002) and increased rates of ulceration (p<0.001), mitoses (p<0.001), and nodular histology (p<0.001) compared to melanomas with absent TIL grade and brisk TIL grade. Although SLN biopsy was performed most often for patients with non-brisk TIL-graded melanomas (p<0.001), TIL grade did not correlate with SLN biopsy positivity (p=0.49) (Table 1).

3.2 Brisk TIL grade is associated with the best survival outcomes

Brisk TIL grade was associated with the best RFS (p=0.025) (Figure 1a), OS (p=0.0057) (Figure 1b), and post-recurrence survival (p=0.046) compared to non-brisk and absent TIL grades. Surprisingly, there were no differences in RFS (p=0.4) or OS (p=0.41) for patients when comparing only non-brisk vs. absent TIL-graded melanomas. This demonstrates that brisk TIL grade distinctly separates out from non-brisk and absent TIL grade as a positive prognostic marker. Median follow-up time for the cohort was 3.9 years. Multivariate analysis was performed to identify whether the poor survival outcomes of patients with non-brisk TIL-graded melanomas compared to those with brisk TIL grade were driven by unfavorable pathologic features alone. However, brisk TIL grade’s association with improved OS was independent of thickness, the most important primary melanoma prognostic feature (p=0.036) (Table 2). Additionally, brisk versus non-brisk TIL grade continued to be significantly prognostic for improved OS, independent of pathologic features incorporated into current AJCC staging including thickness, ulceration, and mitoses (p=0.049). (Table 2).

Figure 1.

Kaplan-Meier survival curves for 1,241 primary melanoma patients with a median follow-up time of 3.9 years showing significant (a) recurrence-free survival and (b) overall survival differences comparing absent, non-brisk, and brisk TIL grade.

Table 2.

Univariate and multivariate analyses for recurrence-free and overall survival comparing brisk to non-brisk TIL grade

Recurrence-Free Survival
UNIVARIATE
	Baseline		HR	95% CI	p-value
TIL GRADE	non-brisk	brisk	0.56	0.37-0.85	0.007
MULTIVARIATE
	Baseline		HR	95% CI	p-value
TIL GRADE	non-brisk	brisk	0.73	0.47-1.14	0.173
THICKNESS			1.14	1.11-1.18	<0.001
	Baseline		HR	95% CI	p-value
TIL GRADE	non-brisk	brisk	0.78	0.50-1.21	0.268
THICKNESS			1.08	1.05-1.12	<0.001
ULCERATION	absent	present	3.69	2.28-5.97	<0.001
MITOSES	absent	present	4.94	2.10-11.65	<0.001
Overall Survival
UNIVARIATE
	Baseline		HR	95% CI	p-value
TIL GRADE	non-brisk	brisk	0.54	0.37-0.79	0.002
MULTIVARIATE
	Baseline		HR	95% CI	p-value
TIL GRADE	non-brisk	brisk	0.66	0.44-0.97	0.036
THICKNESS			1.13	1.10-1.16	<0.001
	Baseline		HR	95% CI	p-value
TIL GRADE	non-brisk	brisk	0.67	95% CI	0.049
THICKNESS			1.08	1.05-1.12	<0.001
ULCERATION	absent	present	2.89	1.85-4.50	<0.001
MITOSES	absent	present	1.75	1.03-2.98	0.037

3.3 CD3 and CD45 expression was highly sensitive for TIL detection, but did not add to prognostication

To identify whether a more sensitive measurement of TILs impacts prognosis, primary melanomas were stained for CD3 (n=210, 18% absent, 29% non-brisk, 53% brisk) and CD45 (n=196, 16% absent, 30% non-brisk, 54% brisk) to quantify intratumoral and peritumoral TILs. Representative staining of melanomas with absent, non-brisk, and brisk TILs and quantitative averages of total CD3 and CD45-expressing TILs (intratumoral + peritumoral) are illustrated in Figure 2a-c. Total CD3 and CD45 counts were sensitive for TIL detection (0.92 and 0.98, respectively) but not specific (0.299 and 0.146, respectively) to TIL grade as determined on H&E. For example, there were cases in which TILs graded as absent on H&E were found to be quantifiable by IHC. Despite this increased sensitivity, quantitative TIL detection by IHC did not impact survival outcomes (Figure 2d and 2e). Using a count of 50 TILs as an upper limit cut-off for defining high versus low TIL count, there were no significant survival differences for intratumoral, peritumoral, or total CD3 or CD45 TIL quantitation. When comparing only melanomas with non-brisk versus absent TIL grade, higher intratumoral CD3 and CD45 expression in melanomas with non-brisk TILs was associated with improved RFS (CD3: p=0.001; CD45: p=0.003) and OS (CD3: p=0.002; CD45: p=0.02).

Figure 2.

(a-c): Demonstrates primary melanomas with (a) absent, (b) non-brisk, and (c) brisk TIL grade on H&E, with corresponding CD3 and CD45 expression, all at a power of 20X. Insets in each of the H&E panels show the entire primary melanoma as viewed on H&E at a power of 4X. Of note, CD3 staining in panel (a) shows non-specific staining. (d-e): Overall survival curves for (d) quantitative intratumoral CD3 counts and (e) quantitative intratumoral CD45 counts. TIL counts greater than 50 were used as an upper limit. “High” represents counts > 50. “Low” represents counts ≤ 50.

3.4 Expression of Foxp3+ Tregs did not significantly differ by TIL grade

Foxp3 expression was measured in the same cohort stained for CD3 and CD45 (n=184, 24% absent, 30% non-brisk, and 46% brisk TIL grades) to identify whether Tregs may contribute to a more immunosuppressed tumor microenvironment in melanomas with non-brisk TILs. Although the quantity of Foxp3+ Tregs increased with melanomas of increasing TIL grade, there was no difference in the percentage of Foxp3+ Tregs (p=0.13) by TIL grade. Increased percentage of Foxp3+ cells in the lymphocytic infiltrate did not impact OS (p=0.23) (Figure 3).

Figure 3.

(a) H&E and corresponding FOXP3 expression at a power of 10X and 40X. (b) Overall survival curves for percentage of FOXP3 positive cells within the entire lymphocytic infiltrate. 10% was taken as the median. “High” represents > 10%. “Low” represents ≤ 10%.

3.5 Melanomas with present TILs represent an immunologically heterogeneous group

A subset of 67 primary melanomas graded as having absent (n=16), non-brisk (n=25), and brisk (n=26) TILs were evaluated for differential immunoregulatory gene expression profiles. Only 8/579 (1%) immunoregulatory genes were differentially expressed in melanomas with non-brisk versus absent TILs, suggesting these melanomas are related immunologically. In contrast, 180/579 (31%) genes were differentially expressed between the brisk and non-brisk TILs group and 297/579 (51%) genes were differentially expressed between the brisk and absent TILs group, emphasizing that melanomas with brisk TILs represent a distinct immunologic entity.

The top canonical pathways enriched in melanomas with brisk versus non-brisk TILs included pathways involved in immune cell function (T helper cell differentiation, dendritic cell maturation), T cell co-stimulation (iCOS, CD28, and CD40 signaling), cytokine signaling (IL-2), and other signaling pathways including interferon, T cell receptor, Toll-like receptor, and JAK/STAT signaling (Figure 4). Melanomas with brisk TILs also demonstrated activation of upstream regulators including cytokines (IL1B, IL2, IL21, and IFNG), co-stimulatory transmembrane receptors (CD40), and the T cell receptor complex. Interestingly, immune checkpoint regulators such as PD1 and CTLA4 were more inhibited in melanomas with brisk TILs compared to those with non-brisk TILs. Melanomas with brisk TILs had gene expression patterns involved pathways enriched for T lymphocyte activation, development, and migration. Upregulated genes common to these pathways include CD28, IFNG, CD2, ZAP70, and TNF. The immunologic profile of melanomas with brisk TILs separated out from melanomas graded as non-brisk or absent TILs. The brisk TIL group is distinct functionally and prognostically from non-brisk TILs, representing a threshold within the “present” (brisk and non-brisk) TILs category.

Figure 4.

Top canonical pathways enriched in melanomas with brisk compared to non-brisk TILs using Integrative Pathway Analysis. Bar length represents statistical significance as displayed as the –log of p-value which is calculated by Fisher’s exact test. Orange squares represent the number of genes in pathway meeting the cutoff criteria in our dataset divided by the total number of genes involved in the pathway. Orange bars indicate predicted pathway activation. Gray bars indicate pathways where a prediction cannot be made.

4. DISCUSSION

There is an unresolved debate in the literature as to whether TIL grade is a robust enough prognostic feature to be included in AJCC staging. In our cohort, brisk TIL grade correlates with improved RFS and OS compared to non-brisk and absent TIL grades. Surprisingly, there are no distinct differences in survival outcomes for patients with non-brisk or absent TIL-graded melanomas. Melanomas with non-brisk TILs are associated with the most unfavorable pathologic features. These results raise the question of whether non-brisk TIL grade may be suggestive of a partially failed host immune response, particularly because patients with a non-brisk TIL grade had an equivalent prognosis to patients without any lymphocytic infiltrate. Through gene expression profiling of each TIL grade, we identified that that melanomas with brisk TIL grade harbor an immunologically activated tumor microenvironment in comparison to melanomas with non-brisk TIL grade, despite both groups having a TIL “presence.”

There are several limitations in the current literature reporting on the prognostic value of TILs that we attempt to address in our study. First, there is a lack of standardization of TIL grading methods. Although Clark’s method for TIL grading (absent, non-brisk, brisk) is used in many publications, other studies score TILs on a scale of 0 to 3, consider the presence of “slight” TILs as “absent,” or only consider TILs as “present” or “absent” (11, 12, 14). Second, heterogeneity of patient selection criteria and often small sample size may also impact the data given that TILs have a higher incidence in thin melanomas (15, 19). For example, Rao et al. studied 293 high-risk, thick (> 4 mm) primary melanomas in which TIL grade did not significantly correlate with survival outcomes (23). In contrast, Thomas et al. did find survival differences between TIL grades in a cohort of 3,330 primary invasive melanomas where the majority were thin (<1 mm) (18) and carried a higher percentage of present (brisk or non-brisk) TILs. Third, inconsistent grouping of TIL grades and loss of prognostic significance in multivariate analysis poses additional limitations in interpreting the impact of TILs. Barnhill et al. examined a majority of thin melanomas (< 1.7 mm) however no survival differences were shown between TIL grades, possibly because TIL “presence” was not stratified as brisk or non-brisk (13). Conversely, Burton et al. found 5 year OS differences between patients with brisk (95%) and non-brisk (84%) TIL grades, however statistical significance was lost in multivariate analysis and absent TIL grade was not addressed (16). Similarly, Mandala et al. found 5 year OS differences for present (94.8%) vs. absent (90.4%) TILs in assessment of 1,251 primary melanomas, but the significance of the survival difference was not confirmed in Cox regression models (17).

We addressed these limitations by conducting an analysis of TIL grade in a large prospectively enrolled cohort of primary melanoma patients representative of the general population. Interestingly, patients with non-brisk TIL grade consistently had the most unfavorable pathologic features as well as OS outcomes no better than patients who had no lymphocytic infiltrate at all. While RFS outcomes by TIL grade were dependent on primary tumor thickness, the OS outcomes for brisk TIL grade were independent of the AJCC staging parameters of primary tumor thickness, ulceration, and mitoses, suggesting that other host- and tumor-related functional factors explain the worse outcomes seen in patients with non-brisk TILs.

Immune measures, specifically TIL grade, are not standard components of melanoma pathology reports because they have not yet been shown to definitively impact survival or alter clinical management. In parallel, associations between lymphocyte marker expression and melanoma prognosis have been investigated, but results are inconsistent (24-29). In our cohort, CD3, CD45, and Foxp3 expression do not add prognostic value over examining H&E sections alone. One reason for this lack of correlation is that H&E TIL grading evaluates the entire surface area of the primary tumor, whereas our quantitative TIL grading by IHC focuses on only one high power field, designed to be an easy and quick measure for any pathologist to perform.

However, a more global analysis using gene expression profiling did identify immunologic heterogeneity among melanomas with present lymphocytic infiltrates (brisk and non-brisk). Melanomas with absent and non-brisk TIL grade had the most similar immunoregulatory gene expression patterns and were immunologically distinguishable from melanomas with brisk TIL grade. This is particularly interesting because many studies classify TILs as only “present” or “absent.” Instead, our data emphasize that presence of brisk TILs is prognostically superior and immunologically distinct from a group of lower intensity “present” (non-brisk) TILs, and certainly from absent TILs. We highlight a clear prognostic and functional distinction between melanomas with brisk and non-brisk TIL grade, which historically have been grouped together as “present” TILs.

In the pathway analysis of the primary melanoma gene expression data, melanomas with brisk TIL grade had upregulated T cell signaling, activation, and migration pathways compared to non-brisk TIL-graded melanomas. Melanomas with brisk TIL grade also demonstrated increased expression of genes involved in immune co-stimulatory signaling pathways including CD40 on antigen presenting cells and iCOS and CD28 on T cells. Zap70, also upregulated in the brisk TILs dataset, functions in T cell receptor-mediated signal transduction in combination with LCK and FYN (30). This is further independently validated in the Cancer Genome Atlas (TCGA) analysis in which melanoma patients with high lymphocyte score and elevated LCK expression in their tumors demonstrated improved survival (31). Identification of distinct immunologic gene expression profiles between melanomas with brisk and non-brisk TIL grades emphasizes that the “presence” of TILs alone is not enough to implicate them as immunologically functional.

The strengths of our study include a large sample size with comprehensive pathologic evaluation and protocol-driven follow-up data. The 4-year median follow-up time also allows ample time for detection of recurrent melanomas, many of which tend to recur within 3-5 years after initial diagnosis. Furthermore, our study population is largely generalizable to the broader population. For example, despite the fact that our institution largely represents a tertiary referral center and often attracts patients with more advanced stages of melanoma, 79% of the primary melanomas studied were less than or equal to 2 mm thick. Thin melanomas have been shown to carry a higher incidence of brisk TIL grading (15, 18) and this finding may account for the higher percentage of brisk TIL grading in our cohort, in which thin melanomas (< 2mm) have an increased incidence of brisk TILs (44%) compared to thick melanomas (> 4mm) with brisk TILs (31%). While one limitation may be that we performed immune marker expression studies and gene expression analysis on a subset of cases rather than the entire cohort, these cases were selected at random and are representative of the overall study cohort. Given the limited amount of tissue available for most primary melanomas, we included a substantial number of cases in our analyses without compromising valuable tissue in our biorepository that is in use for other investigations.

“Immunoscores” are being proposed as prognostic tools in colon cancer, melanoma, and other malignancies (32, 33). Identifying ratios and combinations of immune cell markers with or without use of gene expression profiling can highlight immunoregulatory pathways that impact survival differences. This may lead to identification of new immunologic prognostic markers that could be incorporated into melanoma staging algorithms, whereas examining each marker alone is less revealing. The immunologic heterogeneity inherent in melanomas of varying TIL grades provides clues towards factors driving the prognostic impact of TIL grade over more static measures documenting only the “presence” or “absence” of TILs.

We conclude that the presence of TILs in primary melanomas represents a heterogeneous clinical group and caution in prognostic interpretation is warranted. We challenge the current dogma that the presence of TILs, compared to an absent lymphocytic infiltrate, correlates with improved outcomes. In our cohort, melanomas with non-brisk TIL grade appear to be more closely related to melanomas with absent TIL grade, both prognostically and immunologically. Therefore, TIL presence does not uniformly translate into a heightened host anti-tumor immune response, although brisk TIL grade is a positive prognostic factor in primary melanomas and is a distinct group from non-brisk TIL grade in terms of both survival outcomes and immunoregulatory gene expression profiles. We emphasize that recognition of the underlying properties of the host immune system is crucial in accounting for differences in patient survival among TIL grades.

ABBREVIATIONS

AJCC

American Joint Committee on Cancer

CI

confidence interval

FFPE

formalin-fixed paraffin embedded

H&E

hematoxylin and eosin

HPF

high power field

IHC

immunohistochemistry

IMCG

Interdisciplinary Melanoma Cooperative Group

IPA

Integrative Pathway Analysis

IRB

Institutional Review Board

NYU

New York University

OS

overall survival

RFS

recurrence-free survival

SLN

sentinel lymph node

TIL

tumor infiltrating lymphocyte

Treg

T regulatory cell

VGP

vertical growth phase