The correlations between IL-17 vs. Th17 cells and cancer patient survival: a systematic review

Abstract

Both IL-17 and Th17 cells have been ascribed tumor promoting as well as tumor suppressing functions. We reviewed the literature on correlations between IL-17 versus Th17 cells and survival in human cancer, following the PRISMA guidelines. Serum, formalin-fixed, paraffin-embedded (FFPE) tissue and peripheral blood samples were most frequently studied. High IL-17 quantities were correlated with poor prognosis, whereas high Th17 cell frequencies were correlated with improved prognosis. Since Th17 cells are a subpopulation of IL-17⁺ cells and had a different correlation with prognosis than total IL-17, we substantiate that a distinction should be made between Th17 and other IL-17⁺ cells.

Keywords:

• cancer
• IL-17
• survival
• tumor microenvironment
• Th17

Abbreviations:

• FFPE, formalin-fixed, paraffin-embedded
• PBMC, peripheral blood mononuclear cells
• RORγ, receptor-related orphan receptor gamma
• TNM, tumor node metastasis stage

Introduction

Interleukin-17 (IL-17) was characterized in 1993 and originally named cytotoxic T lymphocyte-associated-8 (CTLA-8).¹ IL-17 was more recently renamed IL-17A and has five family members: IL-17B–F.² Only IL-17F shows some homology and overlapping functions with IL-17A. The main functions of IL-17 are the attraction of neutrophils and stimulation of inflammation.³ The T-helper 17 (Th17) cell, one of the predominant producers of IL-17 that was characterized in 2005,⁴ is essential to protect the host against pathogens that are not handled well by Th1 and Th2 cells.⁵ This pro-inflammatory cell type plays a dominant role in a variety of autoimmune diseases.³ Antibodies targeting IL-17 and its receptor are now used in clinical trials to treat autoimmune diseases like psoriasis, rheumatoid arthritis, and Crohn's disease.⁶ Since IL-17 can also be produced by innate immune cell types including both lymphoid-derived (e.g., γδT cells, invariant natural killer T cells, and innate lymphoid cells)^7,8 and myeloid-derived cells (e.g., neutrophils, macrophages, and mast cells),⁹ it may bridge the activities of the innate and adaptive immune system.¹⁰

Much less studied is the role of IL-17 in cancer. Both tumor-suppressing and tumor-promoting functions have been ascribed to the IL-17 protein and Th17 cells.¹¹ This ambiguity about the function of IL-17 and Th17 cells in cancer has limited the potential for targeting the molecule or using cell-based immunotherapy. Part of the ambiguity may have arisen because different aspects of the IL-17 response are studied. Total protein amount or cells expressing IL-17 protein have been measured in serum and tumor-associated fluids by ELISA and in FFPE tissue by immunohistochemistry, respectively. The effect of Th17 cells has been analyzed mainly in peripheral blood, but also in tumor-associated fluids, FFPE and fresh frozen tissue by flow cytometry, immunohistochemistry or RT-PCR. A review on Th17 cells in cancer by Wilke et al. in 2011 already noted that correlations between the IL-17 protein and survival may be different from correlations with the Th17 cell population.¹²

To systematically study the correlations between the IL-17 protein and Th17 cells and survival in human cancer, we investigated all publications in NCBI PubMed, Ovid Embase, and Web of Science addressing this subject. The aim of our study was to identify the correlations between both IL-17 protein and Th17 cells and prognosis in cancer. The studies were classified by the sample type used to study IL-17 or Th17 cells: serum, FFPE tissue, peripheral blood, tumor-associated fluids, and fresh frozen tissue. Subsequently, the effect on survival was analyzed for each of the sample types studied. The implications for further research of IL-17 and Th17 cells are discussed.

Results

Study design and selection criteria

Of the 2,643 publications identified through database searching on IL-17 or Th17 and cancer, 56 studies met the inclusion criteria (Fig. 1). The main reasons for a publication to be excluded were: being a conference abstract (23%), an animal study (24%), no study on cancer (27%), or not reporting on survival data (19%). Two articles were excluded due to lack of other references on the same method and survival analysis. One article reported on an IL-17 polymorphism analysis,¹³ while the other studied RNA levels of Th17 cell expressed retinoic acid receptor-related orphan receptor gamma (RORγt).¹⁴ Neither of these studies found a correlation with survival. An overview of the included studies sorted by sample type and clinical outcome is shown in Table 1.

Table 1.1. Correlation between IL-17 quantification in serum. All studies describing a correlation between a measurement of IL-17 in serum by ELISA and overall or disease-free survival are shown. The analyses were sorted by clinical outcome, cancer type and study size. Sample size indicates the number of patients on which the correlation between the IL-17 measurement and survival was reported. The column ‘Correlation’ indicates whether a high measurement was correlated with poor or improved survival. A multivariate analysis including both the IL-17 measurement as well as another variable also containing this IL-17 measurement (e.g. a ratio) was not included in our analysis since the potential effect might be lost by correcting for it. An asterisk behind a reference number indicates that the study was not included in the quantitative analyses in Table 2 and Fig. 2. Whether or not the correlation found was independently correlated with survival when corrected for clinico-pathological parameters in a multivariate Cox regression analysis is also indicated. A multivariate analysis including both the IL-17 measurement as well as another variable also containing this IL-17 measurement (e.g., a ratio) was not included in our analysis since the potential effect might be lost by correcting for it. Measurement deviances are indicated under ‘Notes’. CLL = chronic lymphocytic leukemia; AC = adenocarcinoma; CRC = colorectal carcinoma; HCC = hepatocellular carcinoma; NA = not applicable or not mentioned in the article; NSCLC = non-small cell lung carcinoma.

Cancer type	Sample size (n)	Outcome	Correlation	Multivariate Cox p < 0.05	Notes	Ref #
NSCLC	128	OS	Poor	Yes		25
HBV-related HCC	105	OS, DFS	Poor	DFS: Yes OS: NA		26
Leukemia treated with myelo-ablative conditioning and SCT	95	DFS	Poor	Yes		31
Gastric carcinoma	85	OS	Poor	Yes		17
CRC	80	DFS	Poor	NA		24
Acute leukemia	93	OS	Improved	No		32
CLL	294	OS	No correlation		sample type: plasma	33
CLL	84	OS	No correlation			34
Pancreatic AC	62	OS	Trend toward poor			36
Multiple myeloma	50	OS	No correlation		peripheral blood, bone marrow	35

Figure 1. PRISMA Flow diagram. Database searching identified 2642 publications on IL-17 or Th17 and cancer. One publication on this topic of our group published in this issue of OncoImmunology was added manually. Using the search term “tumor” caused studies on tumor necrosis factor to be selected regardless of whether the study was on cancer. Although tumor necrosis factor was excluded as a major topic, many publications that were not on cancer had to be excluded manually. All inclusion criteria were met by 56 articles. Five studies were excluded from analysis for using different sample types or methods than all other articles. Another 32 studies were excluded from meta-analysis because not enough studies were available on sample types other than tissue analyzed for IL-17 expression or insufficient data were provided. Figure adapted from Moher et al.⁸⁰

Studies reporting on survival analysis or risk of recurrence were included regardless of the outcome of the study. A potential publication bias was caused by excluding articles that reported on correlations with other clinico-pathological parameters but not survival. This bias was minimized by screening all articles that reported on correlations with clinico-pathological parameters for having performed a survival analysis. The survival criterion enabled us to focus on studies that are relevant for the potential targeting of IL-17 or Th17 cells in a clinical setting.

Generally, a random or consecutive group of patients was analyzed for relatively objective measures (see Tables S1–4) Although most studies did not provide details on the sample selection method, the majority of the studied did specify that they used a continuous variable or categorized IL-17 or Th17 cell numbers in groups based on the presence, mean or median to analyze the effect by Kaplan–Meier and Cox regression analyses. Potential risks of bias identified in categorizing IL-17 or Th17 expression were optimal cut-off values chosen arbitrarily^15,16 or using a minimum p value,^17-23 receiver operating characteristic (ROC) curve^24-27 or regression tree analysis.²⁸ Furthermore, one study compared the six long (>3 y) vs. short (<1.5 y) surviving patients.²⁹ Another study reported that post-chemotherapy samples were used when no pretreatment samples were available for immunohistochemistry.³⁰ A final potential risk factor was observed in a study of leukemia patients treated with allogeneic stem cell transplantation after myeloablative conditioning, which included donors that varied from related to unrelated and different prophylaxis regimens to prevent graft-vs.-host disease.³¹ Additional study details and concerns are listed per sample type in Tables S1–4. Clinico-pathological characteristics of the different studies per measurement method are provided in Tables S5–8.

High IL-17 serum levels are correlated with poor survival

Serum, paraffin tissue, peripheral blood mononuclear cells (PBMCs) and occasionally tumor-associated fluids or fresh frozen tissue were used to measure IL-17 protein or RNA and Th17 cells. Since the cell source and related activity measured may differ in different sample types, we sorted and analyzed the studies by sample type. The amount of IL-17 protein in serum was measured by ELISA (Table 1.1). Since total protein quantity was measured, the IL-17 could have been derived from Th17 cells but also from innate immune cell types. Five studies out of ten reported that a high amount of serum IL-17 protein was correlated with poor survival.^17,24-26,31 One study showed a correlation between high IL-17 and improved survival in leukemia.³² Four studies did not observe a significant correlation between high serum IL-17 levels and survival,^33-36 although one group did find a trend toward poor prognosis (p = 0.05).³⁶ Overall, a high amount of IL-17 protein in serum has predominantly been correlated with poor survival (Table 2).

Table 1.2. Correlation between IL-17+ cells in tissue and survival. A representation of studies on tumor infiltrated IL-17+ cells quantified by immunohistochemistry on FFPE tissue slides or tissue microarrays. If ‘intratumoral’ is indicated, peritumoral cells were scored as well. CT = chemotherapy; NS = not significant; RT = radiotherapy; SCC = squamous cell carcinoma

Cancer type	Sample size (n)	Outcome	Correlation	Multivariate Cox p < 0.05	Notes	Ref #
HCC	323	OS, DFS	Poor	No		40
HCC	300	OS, DFS	Poor (intratumoral)	NA		18
HCC	150	OS, DFS	Poor (intratumoral)	No	peritumoral IL-17^+ cells correlated with improved survival	41
HCC	108	OS, DFS	Poor (intratumoral)	Yes	hot-spot areas scored	19
HCC	56	OS, DFS	Poor DFS	NA	both intra- and peritumoral cells scored	42
HCC	43	OS, DFS	Poor	Yes		43
Intrahepatic cholangiocarcinoma	123	OS	Poor (intratumoral)	Yes		44
CRC	104	DFS	Poor	NA	both tumor center and invasive margin scored	20
CRC	102	OS	Poor	NA		45
CRC	52	OS	Poor	Yes		16
NSCLC	102	OS	Poor	NA		46
NSCLC	52	OS, DFS	Poor	Yes	same research group as ref^16	15
Breast carcinoma	207	OS, DFS	Poor DFS	Yes	scores in tumor center and front hot-spots averaged	30
Gastric carcinoma	112	OS	Poor	Yes	mainly mast cells were IL-17+	50
Cervical SCC	109	OS	Poor (TNM stage I)	Yes	NS in all TNM stages	37
Gallbladder carcinoma	104	OS, DFS	Poor OS	No		47
Laryngeal SCC	71	DFS	Poor	No		49
Pancreatic AC	46	OS	Poor	NA	condensed expression areas scored	48
Gastric AC	192	OS	Improved	Yes		51
Esophageal SCC	181	OS	Improved	No		52
Cervical carcinoma	153	DFS	Improved	Yes	densest lymphocytic infiltrates scored	21
Recurrent ovarian carcinoma	47	OS, DFS	Improved DFS	No		28
Pancreatic ductal AC treated with vaccine, CT, RT	12	OS	Improved	NA	lymphoid aggregates scored	29
HCC	132	OS, DFS	No correlation		densest lymphocytic infiltrates scored	22
Nasopharyngeal carcinoma	106	OS	No correlation			53
Epithelial ovarian carcinoma	104	OS	No correlation		consecutive hot-spot areas scored	23
Giant cell tumors of bone	74	DFS	No correlation			54
Esophageal SCC	215	OS	High IL-17^+ cells in tumor muscularis propria correlated with improved OS	Yes	correlation between IL-17^+ cells in tumor nests and survival not studied	39*
CRC	78	OS, DFS	Improved	Yes	mainly tumor cells positive	55*
Stage IV glioblastoma	41	OS	Improved	Yes	mainly tumor cells positive	56*

Table 1.3. Correlation between Th17 cells and survival. Analyses of Th17 quantification on FFPE tissue, peripheral blood PBMCs and fresh frozen samples. Bref = brefeldin A; FC = flow cytometry; GVHD = graft-versus-host disease; Iono = ionomycin; Mon = monensin; PMA = phorbol 12-myristate 13-acetate

Cancer type	Sample	Sample size (n)	Measurement	Outcome	Correlation	Multivariate Cox p < 0.05?	Notes	Ref #
Cervical SCC	FFPE	51	IHC CD3^+ IL-17^+ cells	OS	Improved	Yes		37
Differentiated thyroid carcinoma	FFPE	266	IHC CD4^+ IL-17^+ cells	DFS	Trend toward improved			59
HCC	PBMC	150	FC CD4^+ IL-17^+	DFS, OS	Poor	Yes		41
Gastric carcinoma	PBMC + PMA/iono/ mon 4u	32	FC CD4^+ IL-17^+IFNG^-	OS	Poor	NA		62
Acute leukemia	PBMC + PMA/iono/ mon 5h	93	FC CD3^+CD4^+ IL-17^+ cells	OS	Improved	Yes		32
CLL	PBMC + 5h PMA/iono/ mon	66	FC CD3^+CD4^+ IL-17^+ cells	OS	Improved	No	Same group as ref^34	38
End-stage melanoma treated with ipilimumab (aCTLA4)	PBMC + 2.5h PMA/iono/ 2h mon	47	FC CD4^+CD8^-IL-17^+ increase 6m after/1wk before therapy	DFS	Improved	NA		61
HCC treated with transarterial chemoembolization	PBMC + activation mix at -5h and +30d	30	FC CD4^+ IL-17^+ cells	OS	Improved (+30d measurement)	Yes	NS for -5h measurement	60
CLL	PBMC + PMA/iono/mon 5h	150	FC CD3^+CD4^+ IL-17^+ cells	OS	No correlation			33
hematological melagnancies treated with allogeneic HSCT	PBMC	30	FC CD3^+CD4^+ CD8^-IL-17^+ cells	DFS	No correlation	NA	60% of patients got GVHD	64
HCC	PBMC	26	FC CD4^+ IL-17^+ cells	DFS	No correlation		IL-17 ELISA <detection limit	63
Stage IV melanoma with anti-tumor antigen T cell response	Melan-A reactive CD3^+CD4^+ PBMC + 12d peptide mix + 12h antigen/mon	38	FC IL-17 present/absent	OS	Poor	Yes	IL-17 present in n = 3 ; CD4 related with poor OS	65*
Acute myeloid leukemia	BMMC + 5h PMA/iono/bref	98	FC CD3^+CD8-IL17^+	OS	Poor	NA	no correlation for PBMCs (n = 30 )	66*
CRC	normal biopsies ∼10 cm from tumor center	19	qRT-PCR IL-17	OS	No correlation			57
Ovarian carcinoma	fresh frozen tumor tissue	17	agarose gel RT-PCR IL-17 present/absent	OS	No correlation			58

Table 1.4. Correlation between IL-17 and Th17 cells in tumor-associated fluids and survival sorted by measurement type. MPE = malignant pleural effusion

Cancer type	Sample	Sample size (n)	Measurement	Outcome	Correlation	Multi-variate Cox p < 0.05?	Notes	Ref #
Lung carcinoma	MPE supernatant	78	ELISA IL-17	OS	Poor	Yes		27
Ovarian carcinoma	tumor ascites	85	ELISA IL-17	OS	Improved	Yes		69
Lung carcinoma	MPE	30	FC CD4^+ IL-17^+ cells	OS	Improved	Yes		67
Lung AC with pleura metastasis	MPE + 2h PMA/iono + 4h Bref/mon	24	FC CD3^+CD4^+ IL-17^+ cells	OS	Improved	NA		68
Gastric carcinoma	peritoneal lavage	114	qRT-PCR IL-17	OS	Improved (curative resections; n=79 )	Yes	NS for all patients (n=114 )	70

Table 2. Correlations per measurement type. The number of analyses per sample and measurement type of IL-17 protein or Th17 cells showing a correlation with improved or poor prognosis or no effect is indicated. The final column denotes the ratio of the number of analyses showing a correlation with improved prognosis over the number of analyses showing a correlation with poor prognosis, as an indication of the factor difference. A white box indicates a correlation with improved survival, a dark gray box a correlation with poor survival and a light gray box no clear correlation

Target	Sample type	Measurement method	#analyses improved prognosis	#analyses poor prognosis	#analyses no effect	total # analyses	Factor difference
IL-17	Serum	ELISA	1	5	4	10	0.2
	FFPE tissue	IHC	5	18	4	27	0.3
	Tumor-associated fluid	ELISA	1	1	0	2	1.0
		Total	7	24	8	39	0.3
Th17	FFPE tissue	IHC Th17	1	0	1	2	NA
	Peripheral blood	Flow cytometry	4	2	3	9	2.0
	Tumor-associated fluid	Flow cytometry	2	0	0	2	NA
		RT-PCR	1	0	0	1	NA
	Fresh frozen tissue	RT-PCR	0	0	2	2	NA
		Total	8	2	6	16	4.0

A high number of IL-17⁺ cells in tissue is correlated with poor survival

The total number of IL-17⁺ cells was quantified on cancer tissue FFPE whole slides or tissue microarrays using immunohistochemistry. This type of analysis allows for quantification of the total number of IL-17⁺ cells within the tumor microenvironment. IL-17 is expressed by different types of tumor infiltrating immune cells in cancer, predominantly neutrophils and mast cells.^37-39 The total number of IL-17⁺ cells was correlated with poor prognosis in 18 out of 27 studies (Table 1.2).^{15,16,18-20,30,37,40-50} Five studies reported on a correlation between a high number of IL-17⁺ cells and improved survival.^{21,28,29,51,52} It is important to note that in two of these five studies, the IL-17⁺ cells were scored in areas with the densest lymphocytic infiltrate, one of which was on pancreatic ductal adenocarcinoma patients who had received immunotherapy (the correlation between IL-17 and survival was based on 12 patients).^21,29 Four studies did not observe a significant correlation between total IL-17⁺ cells in the tumor and survival.^22,23,53,54 Again the scoring in 2 of these 4 studies had been performed in hot-spot or dense lymphocytic infiltrate areas, while only 3 of the 18 studies reporting on a negative correlation had focused on hot-spots. Three more studies did not focus on IL-17⁺ tumor-infiltrating immune cells and are included with their reported correlations in Table 1 for completeness, but not in the quantitative analyses.^39,55,56

Collectively, 18 studies reported on a significant correlation between high IL-17 and poor prognosis, over 3.5 times more than the studies showing a correlation with improved prognosis (n = 5, Table 2). To visualize the overall correlation, forest plots are shown for the hazard ratio of a high number of IL-17⁺ cells on overall (OS) (Fig. 2A) and disease-free survival (DFS) (Fig. 2B). Of the 22 studies reporting on overall survival, 7 were excluded from the meta-analysis due to insufficient Cox regression data. Of the 16 studies reporting on DFS, 4 were excluded due to insufficient Cox regression data.

Figure 2. Forest plots for IL-17⁺ cells in tissue. Schematic quantitative analyses of the studies on the number of IL-17⁺ cells in FFPE tissue are shown by forest plots. Cox regression hazard ratios and 95% confidence intervals for the correlation between a high number of IL-17⁺ cells and overall survival (A) and disease-free survival (B) were obtained from the articles or via personal communication with the authors. An asterisk (*) indicates that a multivariate Cox regression analysis was used because a univariate analysis was not provided. A dagger (†) indicates that part of the data were obtained via e-mail. The cut-off value used to divide the IL-17⁺ cell frequency in a high and low group is indicated for comparison. The center of the random effects model represents the pooled hazard ratio, while the 95% confidence interval is represented by the diamond horizontal borders.

Correlation between IL-17 RNA expression in fresh frozen tissue and survival inconclusive

Two studies have analyzed IL-17 RNA expression in fresh frozen samples using RT-PCR (Table 1.3). Both studies, on small study populations, did not find associations with survival. One study analyzed IL-17 expression in tumor adjacent normal appearing biopsies (∼10 cm from the tumor center) from 19 colorectal cancer patients.⁵⁷ The other study in 17 ovarian cancer patients only measured the presence of PCR products on agarose gel.⁵⁸ Insufficient data were available to conclude on an association between IL-17 RNA expression in fresh frozen tissue and survival.

A high number of Th17 cells in tissue is correlated with improved survival

The total number of Th17 cells can be quantified using a combination of a T-cell marker and IL-17 in FFPE slides. Using immunohistochemistry, our group has shown that a high number of Th17 cells was correlated with improved prognosis in squamous cervical cancer,³⁷ while another study found a trend toward improved DFS (p = 0.06) in differentiated thyroid cancer (Table 1.3).⁵⁹ We did not include analyses on the ratio of the number of IL-17⁺ cells over the number of CD3⁺ or CD4⁺ T cells, because we do not regard this as a measure for Th17 cells since IL-17 is also produced by other cell types.

A high number of Th17 cells in peripheral blood is correlated with improved survival

Flow cytometry was used to quantify the Th17 cell frequency among PBMCs, usually defined as CD4⁺IL-17⁺ cells (seeTable 1.3 for details). A high number of Th17 cells was correlated with improved survival in four studies.^32,38,60,61 Two studies found a correlation with poor survival,^41,62 while three studies did not find a significant correlation.^33,63,64

Two studies focused on a different aspect of the Th17 response and are included in the overview in Table 1 for completeness, but not in the quantitative analyses.^65,66 Notably, while two studies reported on a correlation between a high number of Th17 cells and poor prognosis, twice as many studies (n = 4) reported on a correlation with improved prognosis (see Table 2).

A high number of Th17 cells in tumor-associated fluids is correlated with improved survival

Tumor-associated fluids have infrequently been studied for the correlation between IL-17 or Th17 numbers and survival (Table 1.4). Two studies have analyzed the number of Th17 cells in lung cancer malignant pleural effusion by flow cytometry.^67,68 Both found a significant correlation between a high number of Th17 cells and improved overall survival.

One group has studied the correlation between high IL-17 protein levels in lung cancer malignant pleural effusion measured by ELISA and described a correlation with poor survival.²⁷ Another study found a correlation between high IL-17 protein levels in ovarian carcinoma ascites and improved survival.⁶⁹ Finally, a study in gastric cancer patients showed a significant correlation between high IL-17 RNA expression measured by qRT-PCR and improved survival in patients treated with curative resection.⁷⁰

Collectively, of the five studies on tumor-associated fluids, the studies quantifying Th17 cells using flow cytometry (n = 2) and qRT-PCR (n = 1) found a correlation with improved prognosis. Of the two studies quantifying IL-17 using ELISA, one found a correlation with improved, and one with poor prognosis.

Differences between cancer types

While functional differences between IL-17 and Th17 cells may be due to the cellular source of IL-17 and the accompanying immune response, this might also depend on the cancer type. In studies on liver cancer (n = 13), a negative (n = 9), or no significant (n = 3) correlation was found between high IL-17 or Th17 cells and prognosis, except for the study of hepatocellular cancer treated with transarterial chemoembolization. All studies on colorectal cancer (n = 6) also found a correlation between high IL-17 and poor prognosis (n = 4) or no significant correlation (n = 1), except for one study that reported on IL-17 being expressed mainly by tumor rather than tumor infiltrating immune cells. The studies on non-small cell lung cancer (n = 3) reported a significant correlation between IL-17 and poor prognosis as well.

In contrast, all analyses described in six leukemia studies (n = 8) showed a significant correlation between PBMC Th17 cells or serum IL-17 and improved prognosis (n = 3) or no effect (n = 4), except for one study of serum IL-17 in patients that received stem cell transplantation after myeloablative conditioning. This might indicate that the immune response in haematological malignancies may differ from solid tumors. Of the studies on ovarian cancer (n = 4), two described a correlation between high IL-17 and improved survival. The other two groups did not find a significant correlation with disease-specific survival, but one of the studies described a correlation between high IL-17 and improved progression-free survival.

These findings indicate that there may be context specific effects on the IL-17 or Th17 cell immune response, although the number of studies per cancer type was too limited to determine whether the cancer type or sample type is more important for the effect on survival.

Discussion

The clinical impact of Th17 cells has remained unresolved in cancer.⁷¹ The aim of this review was to identify the correlations between a high amount of IL-17 protein or high number of Th17 cells in human cancer and patient survival. Following an extensive electronic database search, publications were manually selected without format or language restrictions. Survival analyses were studied in the full article if any analysis regarding prognosis was mentioned in the abstract, minimizing the risk of publication bias. Although the risk of bias in included studies was limited, all studies used different cut-off levels to divide IL-17 or Th17 expression in a high and low expression group due to a lack of established cut-off levels. This study limitation makes it difficult to compare different studies directly.

The sample type studied proved to be crucial for the correlation with clinical outcome. This may partly be explained by a difference in cell source. Some tumor microenvironments may be more favorable for Th17 cells, while others may be more readily infiltrated by IL-17 producing neutrophils. Additionally, the method used determines whether Th17 cells, IL-17 protein, or all IL-17 producing cells are measured. A high amount of IL-17 protein, predominantly produced by neutrophils and mast cells in cancer^37-39 and measured in serum, FFPE tissue and tumor-associated fluids, was over three times more frequently correlated with poor than with improved prognosis. A meta-analysis could only be performed for IL-17 in FFPE tissue due to the limited number of studies on the other sample types. The forest plots clearly showed that a high number of IL-17⁺ cells was correlated with an increased hazard ratio, despite the use of a range of cut-off values, which might depend on the type of cancer and analysis. In contrast, a high number of Th17 cells measured in FFPE tissue, peripheral blood or tumor-associated fluids was four times more frequently correlated with improved than with poor prognosis. Since, IL-17 RNA can generally not be quantified in neutrophils^37,72 the data obtained by RT-PCR analyses most likely represent IL-17 produced by Th17 cells. The PCR measurements in tumor-associated fluids and fresh frozen tissue are thus regarded as an indicator of the Th17 cell frequency. Because of limited data available, we could not conclude on an association between IL-17 RNA expression and survival.

Th17 cells seemed to primarily have a tumor-suppressing effect, whereas IL-17 was generally associated with poor outcome. IL-17 has been shown to be produced by only a small Th17 cell population.^37-39 The tumor-promoting function can be explained by the role of IL-17 in inducing angiogenesis⁷³ and recruiting neutrophils.⁷⁴ Neutrophils have been reported to convert to a tumor promoting phenotype and to induce angiogenesis.⁷⁵ The immune cells capable of producing IL-17 include neutrophils as well as other cell types,^7-9 which may determine an important part of the clinical outcome. The tumor suppression by Th17 cells is probably due to different properties than the secretion of IL-17. Th17 cells might stimulate the Th1 and cytotoxic T cell tumor targeting immune responses.⁷⁶ Additionally, Th17 cells have been shown to have memory stem cell-like properties and the ability to differentiate to Th1/Th17 cells that produce interferon-gamma.⁷⁷ Th17 cells may thus either directly or indirectly suppress tumorigenesis.

The type of IL-17 response is thus likely to be context dependent. Liver cancer, colorectal cancer, and non-small cell lung cancer seem to be correlated with an unfavorable IL-17 response.^{15,16,18–20,24-26,40-46} Leukemia on the other hand might provide an environment favorable for Th17 cells to suppress tumor growth.^32,38 Similarly, ovarian cancer might attract a tumor suppressing IL-17 response, as the majority of studies found a correlation with a favorable outcome.^28,69 A possible explanation might be that different microenvironments favor infiltration of or differentiation toward more or less tumor-promoting immune cell phenotypes. Not only Th17 cells, but also innate cell types capable of producing IL-17 may be correlated with improved prognosis, as we and others have shown for mast cells.^37,39 Although we cannot discriminate whether the cancer type or method used is more important for the correlations found, it is likely that both are important for the cell source studied and thus clinical outcome.

Based on the findings described in the current review, cancer patients with high total IL-17 protein levels might benefit from anti-IL-17 treatment, blocking the tumor promoting response. Adoptive transfer of Th17 cells might be another promising treatment. The feasibility of both approaches needs to be investigated further. Human Th17 cells can be induced by a combination of IL-1β, IL-6, and IL-23, although the exact conditions required are still under debate.⁷⁸ The functions of differentially obtained Th17 cell populations should be determined by functional studies. Animal models may be helpful to clarify this, as the induction of Th17 cells by IL-6 and TGF-β is clearer in mice than in humans, although the described effects on survival are still contradictory.^12,79

We conclude that while IL-17 primarily promotes tumorigenesis, the subpopulation of IL-17 producing Th17 cells seems to have a tumor-suppressing effect. Future research should use methodology that makes a distinction between soluble IL-17 protein, Th17 cells, and other IL-17⁺ cells. This will help to determine whether IL-17 and Th17 cells should be targeted or used in a clinical setting.

Methods

Study design

A systematic search in the NCBI PubMed, Ovid Embase, and Web of Science bibliographic databases was conducted without language restriction in collaboration with information specialist JML following the PRISMA guidelines.⁸⁰ Since IL-17 is frequently studied in autoimmune diseases and together with tumor necrosis factor, these terms were excluded as major topic. The full search terms are provided in Table S9.

Selection criteria

All original research studies reporting on an IL-17A or Th17 cell measurement and OS or DFS in human cancer published until 10 September 2014 were included. Articles that did not meet all inclusion criteria (e.g., conference abstracts; studies on IL-17 producing γδTCR or CD8⁺ cytotoxic T cells or only describing remission or progression-free survival) were excluded. All relevant references in reviews were manually checked for presence in the systematic search.

Data collection and analysis

Survival analyses, generally Kaplan–Meier survival curves, a log rank test and Cox regression analyses, but in some cases a Spearman's rank correlation,^33,68 Wilcoxon signed-rank test²⁹ or Satterwaithe t-test⁶¹ were obtained. All included articles were reviewed by both SP and ESJ. If extracted data did not match, the data were discussed until a consensus was reached. Cox regression hazard ratios and confidence intervals were used to perform a meta-analysis. The authors of all articles that did not provide sufficient information on study details, clinico-pathological data or Cox regression analyses were contacted via e-mail. Due to the number of studies with sufficient data per sample type studied, only the analyses of the number of IL-17⁺ cells in FFPE tissue were suitable for meta-analysis. To keep the data as comparable as possible, scores in the tumor center were used in the cases where data were reported on different scores (e.g., invasive margin, peritumor).

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Acknowledgments

We thank Dr. J. Cao for extracting data from the Chinese articles. We also acknowledge Dr. I.T.A. Peters for providing critical feedback on the manuscript. We are very grateful for the data provided by authors via e-mail.

Funding

This work was supported by grant UL2010–4801 from the Dutch Cancer Society.