Patients and Gastric Tissue Sample
The study was primarily based on 4 independent patient cohorts. Cohort 1 included 201 gastric cancer tissue specimens collected from January 2010 to April 2014 at Fujian Medical University Union Hospital; thirteen samples were excluded due to data missing. Gastric tissue specimens included tumour tissue of the stomach and adjacent non-tumour tissue. To screen for valuable indicators, tissue microarray specimens (TMAs) consisting of the remaining 188 patients who underwent radical gastrectomy and standard D2 lymphadenectomy were used. Cohort 2 included 315 gastric cancer tissue specimens collected from January 2010 to April 2014 at Fujian Medical University Union Hospital; five samples were excluded due to data missing. The remaining 310 patients were employed for the internal validation set. Cohort 3 included 95 gastric cancer samples collected from January 2010 to April 2014 at Qinghai University Hospital; ten samples were excluded due to missing data. The remaining 85 patients were employed for external validation set 1. Samples of Cohort 2 and Cohort 3 were embedded in paraffin for immunohistochemistry. The inclusion criteria were as follows: (a) histological identification of gastric cancer; (b) availability of follow-up data and clinicopathological characteristics; and (c) TNM staging of gastric cancer tumours according to the 2010 International Union Against Cancer (UICC) guidelines. The exclusion criteria were as follows: (1) patients with no formalin-fixed paraffin-embedded (FFPE) tumour sample, including the centre of the tumour (CT) and the invasive margin (IM), from initial diagnosis; and (2) patients who received chemotherapy or radiotherapy before surgery. All participating patients with advanced GC routinely received fluorine-based chemotherapy. Cohort 4 was derived from the Cancer Genome Atlas (TCGA) with 343 gastric cancer patients in all; forty-four patients were excluded due to data missing. The remaining 299 patients were employed for external validation set 2. Comprehensive information of the 4 recruited cohorts is listed (Fig. 1A and Table 1). The study was approved by the Ethics Committees of Fujian Medical University Union Hospital and the Affiliated Hospital of Qinghai University. This study has been approved by the Ethics Committee of Union Hospital Affiliated to Fujian Medical University (Ethics approval number of scientific research project: 2019KTCX012). Informed consent was obtained from all participants.
Table1. Relationship Between CDK5RAP3 Expression and Baseline Characteristics of Patients
variables
|
Discovery Set
|
Internal Validation Set
|
External Validation Set 1
|
External Validation Set 2
2
|
CDK5RAP3low
|
CDK5RAP3high
|
P
|
CDK5RAP3low
|
CDK5RAP3high
|
P
|
CDK5RAP3low
|
CDK5RAP3high
|
P
|
CDK5RAP3low
|
CDK5RAP3high
|
P
|
All patients
|
87
|
101
|
|
187
|
123
|
|
34
|
51
|
|
153
|
146
|
|
Gender
|
|
|
0.323
|
|
|
0.324
|
|
|
0.136
|
|
|
1.000
|
Female
|
17
|
27
|
|
50
|
26
|
|
14
|
12
|
|
58
|
55
|
|
Male
|
70
|
74
|
|
137
|
97
|
|
20
|
39
|
|
95
|
91
|
|
Age at surgery(yr)
|
|
|
0.301
|
|
|
0.522
|
|
|
0.794
|
|
|
0.083
|
<65
|
48
|
47
|
|
113
|
69
|
|
25
|
40
|
|
76
|
57
|
|
≥65
|
39
|
54
|
|
74
|
54
|
|
9
|
11
|
|
77
|
89
|
|
BMI
|
|
|
0.649
|
|
|
0.850
|
|
|
1.000
|
|
|
|
<25
|
71
|
86
|
|
160
|
107
|
|
30
|
44
|
|
None
|
None
|
|
≥25
|
16
|
15
|
|
27
|
16
|
|
4
|
7
|
|
None
|
None
|
|
T classification
|
|
|
0.186
|
|
|
0.007
|
|
|
0.321
|
|
|
0.077
|
T1
|
5
|
5
|
|
12
|
18
|
|
7
|
5
|
|
5
|
10
|
|
T2
|
5
|
16
|
|
17
|
19
|
|
6
|
15
|
|
37
|
29
|
|
T3
|
38
|
40
|
|
63
|
44
|
|
4
|
9
|
|
65
|
77
|
|
T4
|
39
|
40
|
|
94
|
42
|
|
17
|
22
|
|
46
|
30
|
|
N classification
|
|
|
0.204
|
|
|
0.027
|
|
|
0.468
|
|
|
0.426
|
N0
|
7
|
11
|
|
0
|
2
|
|
16
|
22
|
|
48
|
44
|
|
N1
|
17
|
29
|
|
72
|
61
|
|
4
|
12
|
|
39
|
49
|
|
N2
|
21
|
27
|
|
33
|
24
|
|
6
|
5
|
|
32
|
28
|
|
N3
|
42
|
34
|
|
81
|
36
|
|
8
|
12
|
|
34
|
25
|
|
M classification
|
|
|
1.000
|
|
|
NA
|
|
|
NA
|
|
|
0.662
|
M0
|
85
|
98
|
|
187
|
123
|
|
34
|
51
|
|
144
|
140
|
|
M1
|
2
|
3
|
|
0
|
0
|
|
0
|
0
|
|
9
|
6
|
|
TNM stage
|
|
|
0.243
|
|
|
0.005
|
|
|
0.797
|
|
|
0.848
|
I
|
4
|
12
|
|
15
|
23
|
|
8
|
10
|
|
20
|
23
|
|
II
|
22
|
29
|
|
60
|
45
|
|
11
|
20
|
|
54
|
53
|
|
III
|
59
|
57
|
|
112
|
55
|
|
15
|
21
|
|
73
|
66
|
|
IV
|
2
|
3
|
|
0
|
0
|
|
0
|
0
|
|
6
|
4
|
|
Chemotherapy*
|
|
|
0.737
|
|
|
0.430
|
|
|
0.001
|
|
|
0.075
|
No
|
40
|
50
|
|
89
|
65
|
|
26
|
13
|
|
87
|
67
|
|
Yes
|
47
|
51
|
|
98
|
58
|
|
8
|
38
|
|
66
|
79
|
|
Surgery type
|
|
|
NA
|
|
|
0.572
|
|
|
NA
|
|
|
NA
|
Laparoscopic
|
None
|
None
|
|
184
|
119
|
|
0
|
0
|
|
None
|
None
|
|
open
|
None
|
None
|
|
3
|
4
|
|
34
|
51
|
|
None
|
None
|
|
Tumor size (mm)
|
|
|
0.054
|
|
|
0.048
|
|
|
0.098
|
|
|
NA
|
≤40
|
28
|
48
|
|
63
|
56
|
|
15
|
33
|
|
None
|
None
|
|
>40
|
58
|
53
|
|
124
|
67
|
|
19
|
18
|
|
None
|
None
|
|
Resection type
|
|
|
0.646
|
|
|
0.807
|
|
|
0.257
|
|
|
NA
|
Part gastrectomy
|
29
|
38
|
|
86
|
54
|
|
26
|
45
|
|
None
|
None
|
|
Total gastrectomy
|
58
|
63
|
|
101
|
69
|
|
8
|
6
|
|
None
|
None
|
|
Pathological type
|
|
|
0.835
|
|
|
0.949
|
|
|
0.163
|
|
|
NA
|
Adenocarcinoma
|
63
|
76
|
|
155
|
103
|
|
25
|
46
|
|
153
|
146
|
|
mix
|
16
|
18
|
|
25
|
26
|
|
5
|
4
|
|
0
|
0
|
|
non Adenocarcinoma
|
8
|
7
|
|
7
|
5
|
|
4
|
1
|
|
0
|
0
|
|
MSI status
|
|
|
0.001
|
|
|
0.570
|
|
|
0.468
|
|
|
0.019
|
MSI-L/MSS
|
62
|
93
|
|
141
|
97
|
|
28
|
46
|
|
135
|
113
|
|
MSI-H
|
25
|
8
|
|
46
|
26
|
|
6
|
5
|
|
18
|
33
|
|
TIL
|
|
|
0.500
|
|
|
0.239
|
|
|
1.000
|
|
|
NA
|
TIL low
|
50
|
52
|
|
126
|
74
|
|
19
|
29
|
|
None
|
None
|
|
TIL high
|
37
|
49
|
|
61
|
49
|
|
15
|
22
|
|
None
|
None
|
|
P < 0.05 marked in bold font shows statistical significance.
*Adjuvant chemotherapy after surgery, no radiotherapy was administered to anyone of the patients enrolled.
CDK5RAP3 indicates cyclin dependent kinase 5 regulatory subunit-associated protein 3; MSI, Microsatellite instability; TIL, Tumor-infiltrating lymphocytes; TNM, tumor-node- metastasis; NA, not available.
Construction of tissue microarray (TMA)
From January 2010 to April 2014, a total of 201 gastric cancer tissue samples were selected. Briefly, the pathologist examined all gastric cancer tissues and marked the paraffin blocks based on the tumour position of the HE stained section and immunohistochemical slides; the pathologist selected more areas of the tumour tissue without representation of necrosis and haemorrhagic material areas to prepare tissue chips for experiments. Paraffin wax was mixed with an equal amount of beeswax to make 2 blank wax blocks. A puncture hole with a diameter of 1 mm was made in the blank paraffin to separate the two holes, and 80 tissue punches were made. For each sample, a 1.5 mm core was punched from the donor block using a tissue microarray instrument. A tissue analyser was used to sample the tumour-marked wax block, the sampled tissue was placed into the corresponding channel of the blank wax block, and the determined array position was transferred to the recipient paraffin block. Several serial sections (4 μm thick) were cut from all TMAs, and one section of each TMA was stained with haematoxylin-eosin to ensure that the TMA was constructed correctly. The intratumour dot was harvested from the centre of the tumour, while the peritumoural dot was punched out from the area ≥2 cm from the tumour margin. The prepared TMA slides were used for immunohistochemistry (IHC).
IHC and evaluation
The serial sections of the FFPE sample were 4 μm and were mounted on a glass slide for IHC analysis. The sections were deparaffinized with xylene and rehydrated with alcohol. We blocked the endogenous peroxidase by immersing the slices in a 3% H2O2 aqueous solution for 10 minutes and microwaved them in 0.01 mol/L sodium citrate buffer (pH 6.0) for 10 minutes for antigen retrieval. The slides were then washed in phosphate-buffered saline (PBS) and incubated with 10% normal goat serum (Zhongshan Biotechnology Co., Ltd., China) to eliminate non-specific reactions. Subsequently, the primary antibody was incubated with the sections overnight at 4°C. The negative control was processed with the same methods, but the primary antibody was omitted. After rinsing 3 times with PBS, the secondary antibody was diluted and incubated on the slide for 30 minutes at room temperature, and the staining was developed with diaminobenzidine (DAB) solution. Finally, the slides were counter-stained with haematoxylin, dehydrated, and fixed with a cover glass and neutral resin.
We performed CDK5RAP3 (ab157203, Abcam, 1:100) immunohistochemical staining on tumour tissue of gastric cancer patients. The staining intensity and average percentage of positive cells in 5 randomly selected regions were evaluated to represent the protein expression level. The scoring criteria (Supplementary Figure 2) were as follows: staining intensity was divided into 0 (negative staining), 1 (weak staining, light yellow), 2 (medium staining, yellow-brown), or 3 (strong staining, brown). The proportion of positive staining tumour cells was categorized as the following thresholds: 0 (≤5% positive cells), 1 (6%-25% positive cells), 2 (26%-50% positive cells), and 3 (≥51% positive cells). The final expression was calculated by multiplying the staining intensity score by the proportional staining score (total 0 to 9). Patients with final scores of 0, 1, 2, and 3 were classified as the low expression group, and patients with scores 4, 6, and 9 were classified as the high expression group.
We performed CD3 (ab16669, Abcam, 1:150) and CD8 (ab4055, Abcam, 1:200) immunohistochemical staining on tumour tissue of gastric cancer patients to assess TIL. To date, there has been no consensus on evaluating TIL infiltration in gastric cancer through IHC, so we adopted and modified the GALON score which uses CD3 and CD8 as markers to reflect the status of TIL.18 The tumour area was divided into the invasive margin (IM) and the centre of the tumour (CT). We assessed the infiltration of CD3 and CD8 cells and analysed the positively stained cells in each area (CT or IM) at x400 magnification (determining the average percentage of positive cells in 5 fields). As shown in Supplementary Figure 4, the scoring standard was 0 points for <5% CD3CT-positive cells, 1 point for 5%-25% CD3CT-positive cells, 2 points for 26%-50% CD3CT-positive cells, and 3 points for >50% CD3CT-positive cells. A score of 0 or 1 was defined as "low", and a score of 2 or 3 was defined as "high". The same scoring standards were used for CD3IM, CD8CT, and CD8IM. The total TIL score (Immunoscore) was equal to CD3CT + CD3IM + CD8CT + CD8IM. All gastric cancer patients were divided into the TILhigh group and TILlow group according to the median of the total TIL score.
We performed MLH1 (ab92312, Abcam, 1:800), MSH2 (ab52266, Abcam, 1:800), MSH6 (ab92471, Abcam, 1:250), and PMS2 (ab110638, Abcam, 1:250) immunohistochemical staining on tumour tissue of gastric cancer patients to evaluate microsatellite instability (MSI) status.19 The scoring criteria (Supplementary Figure 3) are as follows: deficiency in at least one protein related to mismatch repair genes was interpreted as deficient Mismatch Repair (dMMR) which manifests as MSI‑H; no deficiencies in mismatch repair gene-related proteins was interpreted as proficient MMR (pMMR) which manifests as MSI-L/MSS.
The IHC results were evaluated by two independent gastroenterology pathologists who were blinded to the clinical prognosis of the patients. Approximately 90% of the scoring results were the same. When the scores of the two independent pathologists diverged, another pathologist checked the results again and selected one of the scores proposed by the first two doctors, or the three pathologists discussed the decision together.
Processing and analysis of genomic data
We used publicly available level 3 data from TCGA which was downloaded from the Genomic Data Commons (https://portal.gdc.cancer.gov) on June 15, 2020, and this download included clinical information and mRNA expression data. MSI status was downloaded from the UCSC Cancer Browser (http://xena.ucsc.edu) on June 15, 2020 for stomach adenocarcinoma (STAD) samples. The mRNA expression data were presented as counts and were normalized with R software (version 4.0.0) and the “limma” package.
Gene set enrichment analysis
Gene set enrichment analysis (GSEA) performed by the Molecular Signature Database (MSigDB) was used to identify the pathways that were significantly enriched in CDK5RAP3low tumour samples. If a gene set had a positive enrichment score, the majority of its members had higher expression accompanied with higher risk score, and the set was considered “enriched”.
Statistical analysis
All data were processed using SPSS 25.0 (SPSS Inc. Chicago, IL) and R software (version 4.0.0). The cut-off value for CDK5RAP3 expression was the median value. The results are expressed as the mean±SEM. Student’s t-test or Wilcoxon rank-sum test was used for continuous variables. We used the χ² test or Fisher exact test to compare categorical variables of clinical characteristics. The Kaplan-Meier method was used to estimate median survival. The log-rank test was used to compare survival between two groups. The association of relevant clinicopathological variables with OS was assessed using the Cox proportional hazard model. Interactions between the clinicopathological parameters and responsiveness to chemotherapy were tested with the Cox model. Clustering charts based on the Z‑score normalization method were used to describe the level of protein expression in each case. We defined the survival time of patients who were lost to follow‑up as the time from surgery to the last follow-up time, and the survival time of patients who were still alive at the end of the study was defined as the time from surgery to the database deadline. Two-tailed P values < 0.05 were indicated significant differences.