Page 101 - Read Online
P. 101
Li et al. Cancer Drug Resist. 2025;8:31 Page 7 of 26
saline with 20% Tween-20 containing 5% skim milk for 30 min. Subsequently, the membrane was incubated
with the primary antibody at room temperature for 10 min, and then overnight at 4 °C. After undergoing 5
washes, the membrane was incubated with goat anti-mouse/rabbit secondary antibody (Beijing TDY Biotech,
1:10000 dilution) for 40 min, followed by exposure using western ECL Substrate (Millipore). The relative
expression levels of each protein were assessed using ImageJ software. The primary antibodies used in this
study are listed in Supplementary Table 2. The raw data of blot images for Western blot (WB) analysis are
shown in the Supplementary Materials.
Statistical analysis
All statistical analyses were performed using the GraphPad Prism (version 6.0) or R software (https://www.r-
project.org/, version 4.1.2), and presented as mean ± SD. To compare non-normally distributed continuous
variables between two groups, we used the Wilcoxon rank-sum test. In cases where not otherwise stated,
correlation analyses were conducted using the Spearman method within the “cor” function of the R base
package. All P-values were two-sided, and the significance threshold, unless otherwise stated, was set at
P-values of < 0.05.
RESULTS
Clustering of tissue samples based on immune features
Evaluation using the fviz_nbclust function with the average silhouette width metric determined k = 2 as the
optimal cluster number [Figure 1A]. Subsequent unsupervised k-means clustering partitioned samples into
two distinct groups, revealing well-separated clusters in the dimensionality-reduced space [Figure 1B].
In the reduced dimensional space, a clear and distinct boundary was observed between the two sample
groups, indicating a robust clustering effect. We examined the differences in the infiltration levels of the 22
immune cell types between the two sample groups using a boxplot [Figure 1C]. Twelve of the 22 immune cell
types exhibited significant differences in infiltration levels between the two groups (P < 0.05), highlighting
variations in the immune characteristics between the groups.
Furthermore, by examining the proportions of different immune cells in the two groups [Figure 1D], we
observed variations in immune cell composition. In cluster 1, the proportion of infiltrating naïve B cells was
higher than that in cluster 2, which is consistent with the results depicted in the boxplot. A heatmap
depicting the infiltration levels of the 22 immune cell types and immune/stromal scores [Figure 1E] revealed
significant differences in both the immune and stromal scores between the two clusters. Cluster 1 exhibited
higher scores, suggesting a potentially more complex immune microenvironment and immune cell activity
within this cluster. These findings collectively reflect the substantial heterogeneity of the immune
microenvironment in PCa.
Biological feature analysis of inter-cluster tissue samples
To further analyze the biological differences between clusters 1 and 2, we first examined the correlations
between the 22 immune cell types within cluster 1 [Figure 2A] and cluster 2 [Figure 2B]. The results showed
differences between immune cell types in the two clusters, with variations primarily observed in the strength
of the correlations rather than their directions. Notably, cluster 1 exhibited significantly enhanced positive
correlations between plasma cells and naïve B cells, and between eosinophils and monocytes. This intensified
co-regulation suggests heightened cooperative activity within the immunologically active microenvironment,
including accelerated B cell differentiation from naïve to plasma cell states, and strengthened coordination
among innate immune components. Such cellular synergy likely potentiates antitumor immune responses,
ultimately influencing disease progression and clinical outcomes. These findings underscore how immune
cell interactions drive PCa microenvironmental heterogeneity and modulate immunological activity.
94
