Identification of macrophage-related DEGs in ccRCC using TCGA-KIRC database
TCGA-KIRC data, including 538 tumors and 72 NATs, were downloaded using the TCGAbiolinks R-package (18). For data preprocessing, 167 tumor samples were excluded based on the standard of < 60% tumor purity using the TCGAtumor_purity function of the package. The remaining data were further normalized using the GC-content method (19) and genes with expression levels < 25% of the average levels were excluded. After data preprocessing, the expression data of 13,125 genes from 363 tumor samples and 72 NATs were used for DEGs analysis using the R-package Limma. Using a threshold of the absolute value of log fold change > 0.1 and an adjusted P-value < 1e-10, it was discovered that 3,323 genes were differentially expressed between tumors and NATs (Fig. 1). Moreover, we selected the five most cited (22–23) macrophage biomarkers (i.e., CD14, CD68, CD86, CD163, and colony stimulating factor 1 receptor [CSF1R]) and calculated their Pearson’s correlation coefficient with the discovered DEGs. Thus, a set of correlated DEGs was generated for each of the five macrophage biomarkers. For the detection of genes closely related to TAMs, correlated DEGs with a Pearson’s correlation coefficient > 0.4 and P-value < 1e-10 were considered strongly correlated DEGs (An additional table file shows this in more detail [see Additional file 3]). The Venn diagram intersection analysis revealed that five candidate DEGs were closely related to all selected macrophage biomarkers, namely SH3 and cysteine rich domain 3 (STAC3), galectin 9 (LGALS9), triggering receptor expressed on myeloid cells 2 (TREM2), Fc fragment of IgE receptor Ig (FCER1G), and paired immunoglobin like type 2 receptor alpha (PILRA) (Figs. 2A, B). For validation of our findings, the web-based tool TIMER 2.0 (24) was employed to calculate the correlation score between the abundance of macrophage infiltrates in ccRCC and the expression levels of each candidate DEG using TCGA-KIRC dataset. For each entered candidate DEG, ≥ 13 of 15 algorithms showed a positive correlation (P < 0.05, ρ > 0) (Fig. 2C).
FCER1G was abundantly expressed in ccRCC and contributed to poor prognosis in patients receiving anti-PD1 treatment
To determine their clinical importance, patients with ccRCC from TCGA-KIRC database were divided into two groups according to the median expression levels of each candidate DEG. A Kaplan–Meier survival curve was drawn to examine the difference in OS between the two groups. Among the five identified macrophage-related DEGs, only STAC3 (P = 0.00073) and FCER1G (P = 0.0068) showed a significant (P < 0.05) prognostic difference between the groups (Fig. 3A). Of note, patients with different STAC3 expression levels exhibited a more significant diversity in OS. However, STAC3 showed the lowest average expression in ccRCC tissue among all five DEGs (Fig. 3B). Therefore, we selected FCER1G as the most pronounced macrophage-related DEG with prognostic value for further exploration of its function in ccRCC. The literature review revealed that the differentially expression of FCER1G has been previously reported and has prognostic importance in TCGA-KIRC database (25, 26). Concerning the close relationship of FCER1G with tumor immunity, we downloaded the bulk RNA-seq and clinical data of the CheckMate 025 clinical trial, based on which Nivolumab (anti-PD1) was approved by the Food and Drug Administration for the treatment of ccRCC, to further investigate the prognostic value of FCER1G (20). In this dataset, 181 and 130 patients with ccRCC were treated with Nivolumab and Everolimus (MTOR inhibitor), respectively. The best cut-off value of FCER1G for each group was derived by the ROC curve (Data not shown). The ROC curve revealed that the best cut-off value for FCER1G expression in Nivolumab arm was 33.34. For Everolimus arm, The ROC curve showed that the best cut-off value for FCER1G expression was 34.27. The Kaplan–Meier survival curve showed that following the administration of Nivolumab, patients with a higher FCER1G expression levels were associated with an inferior prognosis (Fig. 3C). In contrast, this tendency was not evident in those treated with Everolimus (Fig. 3D). Considering that anti-PD1 treatment mainly targets CD8+ T cells (27), it is reasonable to presume that the expression levels of FCER1G in ccRCC may be correlated with CD8+ T cells state.
FCER1G and macrophage markers exerted a synergistic effect on predicting the survival of patients with ccRCC, and high FCER1G expression was related to suppression of T lymphocytes
A study reported that FCER1G may be correlated with the immune microenvironment of ccRCC (28). However, there are no studies investigating the exact role of FCER1G in ccRCC tumor immunity. According to the above-mentioned results, we investigated the potential synergistic effect of FCER1G expression and macrophage presence in ccRCC. We employed the median expression levels of FCER1G and macrophage markers to classify patients from TCGA-KIRC database into four expression groups: FCER1Ghigh, Markerhigh; FCER1Ghigh, Markerlow; FCER1Glow, Markerhigh and FCER1Glow, Markerlow. In the groups using CD68 (Fig. 4A) and CD163 (Fig. 4B) as classification markers, the combined indicator showed better survival stratification in the Kaplan–Meier survival method. The best stratification was observed with the combination of CD68 and FCER1G (Fig. 4B). Patients with low expression of both FCER1G and CD68 had the longest OS, whereas those with high expression of both FCER1G and CD68 were linked to the worst clinical prognosis. To further investigate the underlying mechanisms of FCER1G in ccRCC, we performed GSEA by mapping the gene phenotype plotted according to the FCER1G expression levels in TCGA-KIRC with the data obtained from the Gene Ontology and Kyoto Encyclopedia of Genes and Genomes database. Only gene sets with false discovery rate and nominal P-values < 0.05 were considered significantly enriched (An additional table file shows this in more detail [see Additional file 5]). Most gene sets negatively enriched by high FCER1G expression were related to immune response (Fig. 5A). Notably, gene sets related to T cell function were the most common among all immune response gene sets (Fig. 5B). The results of the GSEA suggested that high FCER1G expression in ccRCC is functionally correlated with the suppression of T lymphocytes. This finding is also consistent with the survival curve of patients treated with Nivolumab in the CheckMate 025 trial.
Prognostic value of FCER1G and CD68 in retrospectively collected ccRCC samples
To further validate our findings in clinical setting, we collected samples from 350 patients with ccRCC (Tables 1) (Additional table files show this in more detail [see Additional file 1 and 2]) who underwent either open or laparoscopic surgery from 2010 to 2018 in Shanghai Changhai Hospital. Samples stained with hematoxylin-eosin were examined, and nine pieces of TMA were produced (TMA-30 and TMA-2020 NO.1–8, described in Methods). IHC staining was performed to determine the expression of FCER1G and CD68 in ccRCC using successive sections (thickness: 4 µm) (Fig. 6A). Next, the IHC-score (described in Methods) of each sample was evaluated. Higher expression of both FCER1G and CD68 was detected in tumors compared with NATs (Fig. 6B). Meanwhile, the expression levels of FCER1G and CD68 were strongly correlated in ccRCC tissues (Fig. 6C), with a Pearson’s correlation coefficient of 0.618. Although the correlation of expression was strong, we observed that the expression pattern of FCER1G and CD68 differed in a considerable portion of samples (Fig. 6A, 6C lower left); hence, the types of cells that FCER1G and CD68 represent varied. Owing to the long duration of the follow-up and completeness of OS and PFS, TMA-30 was used as the training group to determine the best cut-off value for the FCER1G and CD68 IHC-score. Using 5-year OS as primary outcome, the ROC curve showed that the best IHC-score cut-off value for FCER1G was 77.5, with an AUC of 0.7026. For CD68, the best IHC-score cut-off value was 147.5, with an AUC of 0.8237 (Fig. 7) (Tables 2, 3). Subsequently, a Kaplan–Meier survival curve was drawn for TMA-2020 NO.1–8 to evaluate the prognostic value of FCER1G or CD68. Patients with high expression levels of FCER1G (Fig. 8A) or CD68 (Fig. 8B) were associated with inferior OS and PFS. Univariate and multivariate Cox regression analyses were performed to further determine whether FCER1G and CD68 were independent risk factors for evaluating OS and PFS in patients with ccRCC. Both FCER1G and CD68 were associated with shorter OS and PFS in Univariate Cox regression analysis. Following multivariable adjustment (i.e., age, sex, Fuhrman grade, TNM stage, and T category), CD68 was identified as an independent risk factor for the PFS of patients with ccRCC (Tables 4, 5).
Table 2
Correlation between FCER1G expression and clinical characteristics of patients with ccRCC in TMA-2020 NO.1–8 (n = 321).
Characteristics
|
FCER1G expression in TMA-2020 NO.1–8
|
Total
(n = 321)
|
P-value
|
High expression
(n = 177)
|
Low expression
(n = 144)
|
Age
|
|
|
|
0.8798
|
< 60 years
|
103
|
85
|
188
|
|
≥ 60 years
|
74
|
59
|
133
|
|
Sex
|
|
|
|
0.0663
|
Male
|
116
|
108
|
224
|
|
Female
|
61
|
36
|
97
|
|
Fuhrman grade
|
|
|
|
< 0.0001
|
1–2
|
158
|
103
|
261
|
|
3–4
|
19
|
41
|
60
|
|
TNM stage
|
|
|
|
0.0046
|
I–II
|
161
|
114
|
275
|
|
III–IV
|
15
|
30
|
45
|
|
NA
|
1
|
0
|
1
|
|
T category
|
|
|
|
0.0008
|
T1a–T1b
|
152
|
100
|
252
|
|
T2a–T4
|
24
|
44
|
68
|
|
NA
|
1
|
0
|
1
|
|
Overall survival
|
|
|
|
0.0291
|
−
|
141
|
108
|
249
|
|
+
|
5
|
12
|
17
|
|
NA
|
31
|
24
|
55
|
|
Progression-free survival
|
|
|
|
0.0010
|
−
|
138
|
98
|
236
|
|
+
|
8
|
22
|
30
|
|
NA
|
31
|
24
|
55
|
|
ccRCC, clear cell renal cell carcinoma; FCER1G, Fc fragment of IgE receptor Ig; NA, not applicable; TMA, tissue microarray |
Table3.Correlation between CD68 expression and clinical characteristics of patients with ccRCC in TMA-2020 NO.1-8 (n=321).
Characteristics
|
CD68 expression in TMA-2020 NO.1-8
|
Total
(n=321)
|
P-value
|
High expression
(n=238)
|
Low expression
(n=83)
|
Age
|
|
|
|
0.4993
|
<60 years
|
142
|
46
|
188
|
|
≥60 years
|
96
|
37
|
133
|
|
Sex
|
|
|
|
0.0914
|
Male
|
160
|
64
|
224
|
|
Female
|
78
|
19
|
97
|
|
Fuhrman grade
|
|
|
|
<0.0001
|
1–2
|
207
|
54
|
261
|
|
3–4
|
31
|
29
|
60
|
|
TNM stage
|
|
|
|
0.0024
|
I–II
|
213
|
62
|
275
|
|
III–IV
|
24
|
21
|
45
|
|
NA
|
1
|
0
|
1
|
|
T category
|
|
|
|
0.0117
|
T1a–T1b
|
196
|
56
|
252
|
|
T2a–T4
|
41
|
27
|
68
|
|
NA
|
1
|
0
|
1
|
|
Overall survival
|
|
|
|
0.0282
|
−
|
191
|
58
|
249
|
|
+
|
9
|
8
|
17
|
|
NA
|
38
|
17
|
55
|
|
Progression-free survival
|
|
|
<0.0001
|
−
|
186
|
49
|
|
|
+
|
14
|
16
|
|
|
NA
|
38
|
17
|
55
|
|
ccRCC, clear cell renal cell carcinoma; NA, not applicable; TMA, tissue microarray
Table 4
Univariate and multivariate Cox regression analyses of patient characteristics with overall survival.
Characteristics
|
Univariate
|
|
Multivariate
|
HR (95% CI)
|
P-value
|
|
HR (95% CI)
|
P-value
|
Age (< 60 vs. ≥60 years)
|
2.1 (0.72–6.00)
|
0.18
|
|
|
|
Sex (female vs. male)
|
1.6 (0.44–5.70)
|
0.48
|
|
|
|
Fuhrman grade (1–2 vs. 3–4)
|
4.5 (1.6–13.0)
|
0.005
|
|
4.027 (1.3926–11.6430)
|
0.0101
|
TNM stage (1–2 vs. 3–4)
|
4.8 (1.6–14.0)
|
0.0049
|
|
3.092 (0.9729–9.8270)
|
0.0557
|
T category (T1a–1b vs. T2a–T4)
|
6.1 (2.9–13.0)
|
1.5E-06
|
|
5.0043 (1.8022–13.8960)
|
0.002
|
CD68 (low vs. high)
|
4.3 (1.5–12.0)
|
0.0069
|
|
2.5303 (0.7912–8.0920)
|
0.11758
|
FCER1G (low vs. high)
|
3.4 (1.1–11.0)
|
0.04
|
|
1.5414 (0.4111–5.7790)
|
0.52102
|
CI, confidence interval; FCER1G, Fc fragment of IgE receptor Ig; HR, hazard ratio |
Table 5
Univariate and multivariate Cox regression analyses of patient characteristics with progression-free survival
Characteristics
|
Univariate
|
|
Multivariate
|
HR (95% CI)
|
P-value
|
|
HR (95% CI)
|
P-value
|
Age (< 60 vs. ≥60 years)
|
1.5 (0.69–3.10)
|
0.32
|
|
|
|
Sex (female vs. male)
|
1.5 (0.6–3.7)
|
0.38
|
|
|
|
Fuhrman grade (1–2 vs. 3–4)
|
3.9 (1.8–8.2)
|
0.0005
|
|
3.621 (1.6718–7.8450)
|
0.0011
|
TNM stage (1–2 vs. 3–4)
|
6.6 (3–14)
|
1.7E-06
|
|
4.886 (2.1503–11.1010)
|
0.00015
|
T category (T1a–1b vs. T2a–T4)
|
5.5 (3.3–9.1)
|
1E-10
|
|
3.7106 (1.6951–8.1230)
|
0.00104
|
CD68 (low vs. high)
|
4.9 (2.3–11.0)
|
0.000047
|
|
2.9592 (1.2744–6.8710)
|
0.0116
|
FCER1G (low vs. high)
|
4 (1.7–9.5)
|
0.0016
|
|
1.7784 (0.6478–4.8820)
|
0.26382
|
CI, confidence interval; FCER1G, Fc fragment of IgE receptor Ig; HR, hazard ratio |
Combination of CD68 and FCER1G expression resulted in a better prognostic stratification in patients with ccRCC
To further test the possible synergistic effect of FCER1G and CD68 expression in predicting the prognosis of ccRCC, 321 patients in TMA-2020 were classified into four groups using the cut-off values determined from the ROC curve: FCER1Ghigh, CD68high; FCER1Ghigh, CD68low; FCER1Glow, CD68high and FCER1Glow, CD68low. The Kaplan–Meier survival curve showed a significant difference in survival between different groups in terms of OS (Fig. 9A) and PFS (Fig. 9B). The group with high expression of both FCER1G and CD68 showed the worst prognosis (Fig. 9C). Furthermore, we examined the prognostic accuracy of FCER1G and CD68 expression versus that of established indicators in patients with ccRCC. Concordance index analysis was used in the validation cohort (n = 321), which demonstrated that the integration of CD68 and FCER1G expression into the established prognostic indicators exhibited a higher concordance index value than any of these indicators alone (Table 6). In the same TNM stage group, patients with high expression of both CD68 and FCER1G (double high) showed worse OS (Fig. 10A) and PFS (Fig. 10B) compared with the remaining patients (non-double high). This may assist physicians in distinguish high-risk patients following surgery for ccRCC. Finally, we constructed nomograms to predict OS and PFS in patients with ccRCC at 3 and 5 years (Fig. 11A). Calibration plots of the nomograms for the prediction of 3- and 5-year OS (Fig. 11B) and PFS (Fig. 11C) are presented below.
Table 6
Concordance index analysis
Characteristics
|
Overall survival
|
|
Progression-free survival
|
Validation cohort (n = 321)
|
|
Validation cohort (n = 321)
|
Fuhrman grade (1–2 vs. 3–4)
|
0.4970 (0.4055–0.5884)
|
|
0.4893 (0.3917–0.5869)
|
TNM stage (1–2 vs. 3–4)
|
0.5265 (0.4153–0.6377)
|
|
0.5217 (0.3989–0.6444)
|
T category (T1a–1b vs T2a–T4)
|
0.5419 (0.4502–0.6335)
|
|
0.5438 (0.4439–0.6437)
|
CD68 (low vs. high)
|
0.5616 (0.4802–0.6431)
|
|
0.5420 (0.4546–0.6295)
|
CD68 + Fuhrman grade
|
0.5236 (0.4557–0.5917)
|
|
0.5194 (0.4480–0.5908)
|
CD68 + TNM stage
|
0.5450 (0.4743–0.6157)
|
|
0.5404 (0.4653–0.6156)
|
CD68 + T category
|
0.5462 (0.4795–0.6128)
|
|
0.5451 (0.4748–0.6155)
|
FCER1G (low vs. high)
|
0.5959 (0.5221–0.6696)
|
|
0.5906 (0.5141–0.6669)
|
FCER1G + Fuhrman grade
|
0.5537 (0.4922–0.6153)
|
|
0.5653 (0.4992–0.6314)
|
FCER1G + TNM stage
|
0.5693 (0.5050–0.6335)
|
|
0.5676 (0.5005–0.6348)
|
FCER1G + T category
|
0.5686 (0.5084–0.6288)
|
|
0.5688 (0.5059–0.6318)
|
CD68 + FCER1G
|
0.5753 (0.5145–0.6361)
|
|
0.5668 (0.5032–0.6305)
|
FCER1G, Fc fragment of IgE receptor Ig |