mRNA expression level of PPP4C in DLBCL

First, we utilized data from the TCGA database to analyze the gene expression levels of PPP4C in various human cancer tissues and compared them with normal tissues. Our analysis revealed that the mRNA expression of PPP4C was notably elevated in multiple types of cancer tissues, including DLBCL, in comparison to their corresponding normal tissues (Fig. 1a). Further validation of the PPP4C expression level was conducted using two separate GEO datasets: GSE56315 and GSE32018. In both datasets, PPP4C expression was greater in DLBCL tumor tissues rather than normal tissues (Fig. 1b and c). PPP4C has good sensitivity and specificity for predicting patient outcomes, according to ROC analysis (AUC 0.896; Fig. 1d).

Fig. 1

The expression profile of PPP4C in diffuse large B-cell lymphoma. A PPP4C expression in 34 kinds of normal and cancerous tissues (TCGA and GTEx normal data in comparison with TCGA cancer data). B–C In the GSE56315 and GSE32018 datasets, the expression level of PPP4C was greater in DLBCL tissue compared to the nearby normal tissue. D PPP4C demonstrated good accuracy in predicting both normal and malignant outcomes, according to the ROC curve.(*p ≤ 0.05, **p ≤ 0.01,***p ≤ 0.001)

PPP4C expression levels in tissue samples of DLBCL patients

Next, we conducted an immunohistochemical assay to assess the expression of PPP4C in the DLBCL tissue microarray. The findings revealed that 76.84% (146/190) of DLBCL tissues exhibited positive staining of PPP4C, primarily localized in the nucleus of tumor cells. Of these, 128 patient samples (or 67.4%) were classified as having PPP4C low expression, and the remaining 62 samples (or 32.6%) were classified as having high PPP4C expression. Figure 2 displays various PPP4C IHC staining intensity.

Fig. 2

Typical IHC pictures of the expression of PPP4C in tissues from DLBCL patients

Correlation between PPP4C expression and clinical characteristics of DLBCL patients

We conducted an additional analysis to examine the relationship between PPP4C expression levels and clinicopathological parameters in 190 DLBCL patients. Using a median IHC score of 4 as the cut-off value for PPP4C expression, we found high expression in 62 (32.63%) tissue samples from DLBCL patients. Next, the association of PPP4C expression in tumor tissues was evaluated with various clinicopathologic parameters including sex, age of diagnosis, Ann Arbor stage, ECOG, extranodal site, LDH, IPI score, B symptoms, Hans typing, and Ki-67. Table 2 summarizes the correlation between PPP4C expression and clinicopathological parameters in 190 DLBCL patients. High PPP4C expression was found to be significantly associated with higher ECOG (P = 0.003) scores compared to low PPP4C expression. However, no significant correlation was observed between PPP4C expression level and other clinicopathological parameters.

Table 2 Relationship between PPP4C expression and clinicopathologic parametersPrognostic value of PPP4C in DLBCL

To assess whether PPP4C expression and clinicopathological characteristics are independent risk factors for DLBCL patients, univariate and multivariate Cox regression analyses were conducted. The univariate analysis revealed significant associations between decreased overall survival (OS) and several factors, including stage III/IV (P < 0.001), ECOG ≥ 2 score (P = 0.038), above normal LDH levels (P = 0.013), IPI > 2 score (P < 0.001), absence of rituximab use (P = 0.021), and high PPP4C expression levels (P = 0.001) (OS; Table 3).Similarly, stage III/IV (P < 0.001), above normal LDH levels (P = 0.018), Extranodal site ≥ 2 (P = 0.035), IPI > 2 score (P < 0.001), absence of rituximab use (P = 0.039) and high PPP4C expression levels (P = 0.002) were also found to be significantly associated with decreased progression-free survival (PFS; Table 4) in the univariate analysis. Furthermore, the multivariate Cox model analysis identified IPI (P = 0.043), use of rituximab (P = 0.004), and PPP4C levels (P = 0.011) as independent predictors for OS, while also indicating that IPI (P = 0.018), use of rituximab (P = 0.029), and PPP4C levels (P = 0.040) were independent predictors of PFS.

Table 3 Univariate and multivariate Cox regression analysis of OSTable 4 Univariate and multivariate Cox regression analysis of PFS

Based on the two independent risk factors mentioned above, in order to refine the risk stratification of DLBCL, all patients were divided into four groups, and a log-rank (Mantel-Cox) test was performed to consider the interaction between PPP4C expression levels and IPI. The level of risk for poor prognosis in the remaining three groups was analyzed, using patients in the 0 risk factor group as a reference. It was found that for OS, patients in the High expression + IPI 3–5 group had a 6.2-fold higher risk of poor prognosis compared to the 0 risk factor group (HR = 6.246, 95%CI 2.425–16.09, P < 0.001). Similarly, for PFS, Low expression + IPI 3–5 and High expression + IPI 3–5 were 2.0 times higher (HR = 1.952, 95%CI 0.9704–3.925, P = 0.0217) and 4.8 times (HR = 4.753, 95%CI 2.111–10.70, P < 0.001). Although there was no significant correlation between the Low expression + IPI 3–5 and High expression + IPI 0–2 groups, it is still informative for refining risk stratification (Fig. 3).

Fig. 3

OS and PFS of DLBCL patients in different risk groups

Association between PPP4C expression and survival outcome in DLBCL

We evaluated how well PPP4C predicts OS and PFS in all cases of DLBCL. The Kaplan–Meier survival analysis showed that individuals with DLBCL who had high PPP4C expression (n = 62) survived considerably less than those who had low PPP4C expression (n = 128) (P = 0.001) (Fig. 4a). In the group with high PPP4C expression, 45 patients (72.6%) experienced disease progression or death, compared to 58 patients (45.3%) in the group with low PPP4C expression. PFS was statistically significant (P = 0.002) (Fig. 4b). We used the GEO database (n = 414, Fig. 5) to confirm the association between survival and PPP4C expression. Consistent with the results of this study, the results of the GEO database showed that DLBCL patients with high PPP4C expression had a worse survival prognosis. Next, we commenced our investigation by comparing the OS between patient groups with low and high PPP4C expression, diversified by different disease phenotypes (Fig. 6). Notably, our analysis revealed that in high-risk patients, PPP4C expression was substantially related to OS. Moreover, upon further analysis depicted in Fig. 7, it became evident that high PPP4C expression corresponded significantly with poorer PFS in patients of stage III/IV (P = 0.001), IPI 3–5 (P = 0.011), age ≤ 60 (P = 0.010), No B symptoms (P = 0.009) ECOG 0–1 (P = 0.034), ECOG ≥ 2 (P = 0.044), normal LDH (P = 0.037), LDH > normal (P = 0.007) and Non-GCB (P = 0.004) subtypes. Thus, it is observed that high PPP4C expression is linked to poorer survival outcomes in several patient subgroups.

Fig. 4

Kaplan–Meier survival curves grouped by high and low PPP4C expression in DLBCL patients

Fig. 5

Association between PPP4C levels and survival in GSE10846

Fig. 6

Kaplan–Meier survival curves display OS in DLBCL patients with high and low PPP4C expression, stratified by various clinical factors. A Stage I/II, B stage III/IV, C IPI 0–2, D IPI 3–5, E GCB, F non-GCB, G age ≤ 60, H age > 60, I No B symptoms, J B symptoms, K ECOG 0–1, L ECOG ≥ 2, M extra node 0–1, N extra node ≥ 2, O Normal LDH. P LDH > Normal

Fig. 7

Kaplan–Meier survival curves display PFS in DLBCL patients with high and low PPP4C expression, stratified by various clinical factors. A Stage I/II, B stage III/IV, C IPI 0–2, D IPI 3–5, E GCB, F non-GCB, G age ≤ 60, H age > 60, I no B symptoms, J B symptoms, K ECOG 0–1, L ECOG ≥ 2, M extra node 0–1. N extra node ≥ 2, O Normal LDH, P LDH > normal