Prevalence of CD8+ cytotoxic lymphocytes in human neoplasms

Purpose Immune checkpoint inhibitors have recently been approved by the US FDA as first and/or second line therapy in a subset of cancer types. Recent evidence suggests that the quantity of tumor infiltrating lymphocytes (TILs) influences the likelihood of response to immune checkpoint inhibitors. Here, we set out to assess the density of CD8+ lymphocytes in a wide range of different cancer types and subtypes. Methods The density of CD8+ lymphocytes was compared across different cancer types using tissue microarrays (TMAs) composed of up to 50 tumor samples each from 84 different cancer types and subtypes. In total 2652 cancers and 608 normal tissues were successfully analyzed by CD8 immunohistochemistry followed by automated image analysis of digitized slides. Results We found that the median CD8+ lymphocyte counts ranged from 6 cells/mm2 in pleomorphic adenoma up to 1573 cells/mm2 in Hodgkin’s lymphoma. The CD8 counts were generally lower in normal tissues compared to cancer tissues. Blood vessels of the spleen were the only non-lymphatic tissue staining positive for CD8. Tumor types approved for checkpoint inhibitor therapy, including malignant melanoma (81), muscle invasive urothelial carcinoma (119), small cell lung cancer (120), clear cell renal cell cancer (153), squamous cell carcinoma (189) and adenocarcinoma of the lung (328) as well as Hodgkin’s lymphoma (1573) were all ranking among the upper half of our list. Comparably high CD8 densities (median cells/mm2) were also found in several rare and aggressive cancer types including Merkel cell carcinoma (70), angiosarcoma (95), anaplastic thyroid cancer (156) and embryonal carcinoma of the testis (186). In 73 of the 84 analyzed cancer types, the highly variable CD8 counts occasionally exceeded the average CD8 count of tumors for which checkpoint inhibitors have been approved. Conclusion These data support the concept that among most tumor types at least some individual cancers may benefit from treatment with immune checkpoint inhibitors. Electronic supplementary material The online version of this article (10.1007/s13402-020-00496-7) contains supplementary material, which is available to authorized users.

inhibitors and the cancer types for which they are being utilized will increase markedly in the coming years [9].
Despite striking successes of these therapies in many patients, as yet the majority of treated individuals only shows a partial response [10]. Predictive tests that identify suited patients are urgently needed. An increasing number of publications has demonstrated that the number of tumor infiltrating lymphocytes (TILs) is one relevant factor determining the potential response of a cancer to checkpoint inhibitors [11]. As these medications combat cancer by stimulating cytotoxic lymphocytes, it is thought that the presence of more lymphocytes may increase the likelihood for a successful application of immune checkpoint inhibitors. Even in absence of specific therapy, tumors with high lymphocyte content such as medullary breast cancers, seminomas or colorectal carcinomas are often characterized by a particularly good prognosis [11][12][13]. Many studies describing (TILs) have shown associations with tumor phenotype, patient outcome or response to therapy [14][15][16]. Altogether, these data indicate that the number of lymphocytes in cancer tissues is of clinical importance.
Given the growing interest in the lymphocyte content of tumors, we utilized tissue microarrays (TMAs) composed of up to 50 tumors each from 84 different cancer types and subtypes to compare the density of CD8 + lymphocytes. The data provide a systematic overview with respect to the degree of cytotoxic T cell involvement in different cancers.

Patients and tissues
Formalin fixed paraffin embedded tissue samples from 3659 patients representing more than 80 different tumor types and subtypes and more than 70 different normal tissue types were retrieved from the archives of the Institute of Pathology at the University Medical Center Hamburg-Eppendorf. A pathologist identified representative cancerous and normal human tissue areas to assemble two different types of tissue microarrays (TMAs) from these samples: The first, a multitumor (TMA), contained 4-50 samples each from 84 different human tumor types and subtypes, as shown in Fig. 1. The samples of this first TMA were distributed among 8 different TMA blocks, each containing between 454 and 532 samples. The tissue cores were selected for a high tumor cell content on hematoxylin & eosin stained tissue sections of the donor tissue blocks, but not for particular features connected to the lymphocyte content such as the presence or density of infiltrating lymphocytes. The second TMA was composed of normal appearing tissues taken from organs removed for non-tumor reasons and encompassed 8 samples each of 76 different normal tissue types (608 samples on one slide), as shown in Supplementary Fig. 1. The exact composition of the study samples is shown in Table 1 and Supplementary Table 1. For both TMAs, a single 0.6 mm tissue punch was taken from Fig. 1 Distribution of CD8 + cell density (cell/mm 2 ) across 84 different human tumor entities. In total 3339 tumor samples, represented by the black dots, were analyzed. The vertical bars indicate the median density per entity. Ca, carcinoma; SCC, squamous cell carcinoma each tissue sample. The TMA construction process was described previously [17]. The use of archived tissues has been approved by local laws (HmgKhG §12) and the local ethics committee (Ethics Commission Hamburg, WF-049/09). All studies have been carried out in compliance with the Helsinki Declaration.

Immunohistochemistry
Freshly cut TMA sections were all stained for CD8 in one run on the same day. The slides were deparaffinized, rehydrated, exposed to heat-induced antigen retrieval for 15 min at 98°C in

Statistics
JMP Pro 12 software package (SAS Institute Inc., NC, US) and R version 3.5.1 (The R foundation) [18,19] were used to plot the data, to calculate the median values and to perform analysis of variance (ANOVA).

Results
A total of 2652 (72%) of 3659 tumor samples and 608 (100%) of 608 normal tissue samples were interpretable in our TMA analysis. The remaining 1007 spots were excluded due to missing tissue or the absence of unequivocal cancer cells in the TMA spot. The number of interpretable spots for each tumor type ranged from 3 to 48 samples (mean: 32 ± 13). The vast majority, i.e., 2602 (98%) of the 2652 tissue spots, contained CD8 + lymphocytes, although there were 50 (2%) samples that completely lacked CD8 + T cells. These were from 31 different tumor entities. All data are summarized in Table 1 and Fig. 1. The median number of CD8 + lymphocytes was highly variable and ranged from 6 cells/mm 2 in pleomorphic adenoma (Fig. 2b) up to 1572 cells/mm 2 in Hodgkin's lymphoma (Fig. 2f)

Discussion
The data from this study provide a comprehensive overview of the density of tumor infiltrating CD8 + lymphocytes across a large number of different human tumor types. Numerous studies have recently quantified immune cell infiltrations in and around tumors. Most of these have used semi-quantitative methods to describe their data such as cell counts per highpower field or per predefined tissue area, positive pixel counts per area, semi-quantitative estimation of TIL numbers or the fraction of TILs in relation to the total intra-tumoral stroma [20][21][22][23][24][25][26]. Data on lymphocytic infiltration in cancers cannot always easily be compared between studies because objective metric data are often lacking. In this project, we described all data as "cells/mm 2 " in order to generate as comparable data as possible. An automated approach was developed that eliminates differences in tissue quantity occurring in TMA studies. Most spots are typically intact in TMA studies, but some are incomplete (Fig. 3). This may be due to specific properties of the arrayed tissue or may be caused by technical issues during embedding of tissue cores in a TMA or during sectioning. The validity of our approach is supported by our finding that tumor types among the top positions in our ranking order are morphologically characterized by high lymphocyte densities, such as seminoma, Warthin's tumor, medullary breast cancer or thymoma. Other cancers with particularly high lymphocyte densities are those occurring in areas with a high density of mucosal lymphocytes, such as cervical or stomach cancer.
Several earlier CD8 density studies in cancers employing automated image analysis or manual cell counting to quantitate TILs in terms of "cells per mm 2 " yielded results that were comparable to those from our study. For example, an assessment of 220 gastric cancers by automated image analysis showed a median count of 436 CD8 + cells/mm 2 , which fits well with our results (mean: 413 CD8 + cells/mm 2 in 62 intestinal and diffuse stomach cancers) [27]. Another study that counted T cells manually in 102 cervical squamous cell cancers reported a median of 441 CD8 + T cells/mm 2 , which is very similar to the median value of 468 CD8 + T cells/mm 2 in our 43 patients [20]. One earlier study by Steele et al. [28] has used "cells/mm 2 " to compare the density of CD8 + cells in 8 different cancer types in a study analyzing the largest possible area per cancer on large sections. Several relevant observations were comparable between our and their study ( Table 2), such as relatively low densities of CD8 + cells in kidney and prostate cancers and a high density in non-small cell lung cancer. The largest discrepancy was seen in pancreatic cancers, in which Steele at al. found 316 cells/mm 2 and we found 93 cells/mm 2 on average. It is a peculiarity of pancreatic cancer that it contains a high amount of tumor stroma and that the lymphocytes mostly reside in the stroma. It cannot be excluded that the selection process of tissues for TMA manufacturing has resulted in an overrepresentation of cancer cell-rich tissues with less stroma, as these are easier to target during the array manufacturing process.
Cancer types for which treatment regimens with immune checkpoint inhibitors are already approved, such as malignant melanoma, muscle invasive urothelial carcinoma, clear cell This supports the concept that a high number of tumor infiltrating lymphocytes goes along with an increased likelihood for response to immune checkpoint inhibitors. The median content of CD8 + cells was 248 cells/mm 2 for these tumor types. That 73 of our 84 analyzed tumor types had -at least occasionally -CD8 + cell counts that were higher than 248 cells/mm 2 may be viewed as an argument for the applicability of immune checkpoint inhibitors -and probably other forms of immunotherapy -to a wide variety of different types of cancer. However, the CD8-positive cell density was found to be highly variable in most analyzed cancer types. All analyzed cancer types had at least one tumor with a CD8 + cell count below 282/mm 2 . This is in line with clinical observations that only a fraction of individual cancers from a certain tumor type will eventually respond favorably to treatment [29].
We identified several rare and aggressive tumor types that frequently exhibited a high density of CD8 + cells. These included anaplastic thyroid cancer, anal carcinoma, embryonal carcinoma of the testis, Merkel cell carcinoma, angiosarcoma and squamous cell carcinoma of the vagina. These cancers are candidates for a thorough assessment of the efficacy of immune checkpoint inhibitors. A clinically relevant effect of the lymphocyte content has already been suggested for some of these tumors, including reports on successful therapy attempts with checkpoint inhibitors. For Merkel cell carcinoma, a prognostic role of the associated immune cell count has been described in several studies [30,31]. Furthermore, an objective response rate of more than 50% in a cohort of 26 Merkel cell carcinoma patients treated with anti-PD-1 therapy [32] has led to several ongoing phase I (NCT02488759, NCT02584829), II (NCT02196961) and III (NCT03271372) studies for this tumor entity. Also a patient with anaplastic thyroid cancer has recently reached complete radiographic and clinical remission after anti-PD-1 therapy [33]. Combination therapies can also be efficient in these tumor types as demonstrated by the success of combining monoclonal antibodies targeting PD-1 and CTLA-4 in angio-and other sarcoma types [34]. Moreover, a prognostic role of the CD8 + cell count was recently suggested in angiosarcoma [35].
The premise that the number of lymphocytes in a cancer is linked to the likelihood of a positive response to immune Fig. 3 Area measurements of representative TMA cores that show an uneven or incomplete shape. Magnification 100×; TMA spot size 0.6 mm. Note that the tissue spots may have a larger area as theoretically expected from a 0.6 mm tissue punch (0.283 mm 2 ). This is because the tissue expands when mounted on a slide checkpoint inhibitors is widely accepted [36]. Many studies have separately assessed lymphocyte counts in the tumor center and in its periphery [13,[37][38][39]. Some studies have recently suggested that the lymphocyte content at the tumor periphery may be more important clinically than the lymphocyte content in the tumor center. For example, in a study on 46 patients with metastatic melanoma, Tumeh et al. found that response to anti-PD-1 therapy was associated with higher CD8 + cell densities within the invasive margin [37]. In another study on metastatic colorectal cancer, the density of immune cells at the invasive margin predicted response to conventional chemotherapy [40]. However, quantification of TILs at the tumor margin is hampered by a "non-standardized" structure of the tumor periphery. For several tumor types such as bladder cancer, colorectal cancer and stomach cancer, different growth patterns have been described [41][42][43]. TIL quantification is easy in tumors with a solid growth and a clearly defined tumor margin, but is much more difficult in cancers showing insular or diffuse growth patterns with irregular borders. In contrast, lymphocyte quantification in the tumor center is independent of tumor growth patterns and thus simpler and perhaps more reproducible to measure than in the periphery. The high concordance between large sections and TMA CD8 counts in 17 small cell bladder carcinomas and in 47 urothelial cancers demonstrates that representative lymphocyte quantification is possible for tumor centers on a 0.6 mm tissue core (Fig. 4).
Tissue microarrays with small 0.6 mm tissue cores were used in this study. When we introduced the TMA technology in 1998, several authors criticized the small diameter of the TMA spots. However, almost 20 years later, every known association between immune cell quantification and patient outcome [13,44,45] or between molecular markers and clinico-pathological features [46][47][48][49] have been successfully reproduced in TMA studies. In an earlier study on p53 staining, we even showed that the 0.6 mm TMA format provided better predictive power compared to the large section format technology [46]. Contrary to initial expectations, tissue heterogeneity does not negatively influence the ability of TMA spots to detect associations between molecular markers and tumor features. This is because of the high number of samples that are typically used in such analyses, as well as the high degree of standardization of the experimental conditions that cannot be achieved in standard slide-by-slide analyses [50][51][52][53].
Our analysis also included 76 different types of normal tissues, which apart from lymphatic tissues, exhibited much lower CD8 + lymphocyte counts than the tumors. Spleen blood vessels were the only non-lymphoid cell type staining unequivocally positive for CD8. CD8 expression in splenic vessels was first described by Buckley et al. in 1985 [54]. Further studies revealed that only the alpha chain of the CD8 molecule is expressed by the venous sinus-lining cells in the spleen's red pulp (also known as littoral cells) and that this expression is unique to human spleen [55].
In summary, the data presented in this study provide a comprehensive overview of CD8 positivity across human normal tissues and cancers. Cancer types with high densities of CD8 + immune cells may be the best candidates for studies assessing the efficacy of immune checkpoint therapies.
Funding Information Open Access funding provided by Projekt DEAL.

Compliance with ethical standards
Conflict of interest The authors declare that they have no conflict of interest.
Ethical approval The usage of archived tissues has been approved by local laws (HmgKhG §12) and the local ethics committee (Ethics commission Hamburg, WF-049/09). All studies have been carried out in compliance with the Helsinki Declaration.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.