Semiquantitative immunohistochemical (IHC) pixelwise H-score of mitochondrial transcription factor A (TFAM) in gastric adenocarcinoma (GAC): clinicopathological significance and association with p53 and HER2

Abstract

Background: The roles of p53 and HER2 in gastric adenocarcinoma (GAC) have been extensively studied; nevertheless, the contribution of mitochondrial transcription factor A (TFAM) remains unclear. Concerning TFAM's pivotal role in mitochondrial biogenesis, this study aimed to evaluate the TFAM expression in GAC and to assess its associations with p53 and HER2 expressions and clinicopathological outcomes.

Methods: We retrospectively analyzed 77 GAC patients who underwent upfront gastrectomy at Taipei Hospital between 2012 and 2021. Their clinicopathological profiles were recorded in detail. Immunohistochemical (IHC) staining for TFAM, p53, and HER2 protein expressions was semiquantified using IHC pixelwise H-score analyzed by ImageJ plugins IHC profiler. Associations between two continuous variables were assessed by Spearman's correlation coefficient (CC), and trendlines were fitted using SPSS's curve estimation function. The optimal cutoff for survival discrimination was derived from receiver operating characteristic (ROC) curve analysis by selecting the threshold with the highest Youden index and area under the curve (AUC). Prognostic variables with a Log-rank test p-value ≤ 0.1 were entered into a multi-variate Cox proportional hazards regression (Cox regression) model to identify independent ones and their relative hazards ratio (HR).

Results: TFAM IHC pixelwise H-score was significantly associated with advanced T and N status, lymphovascular invasion, perineural invasion and poor differentiation (all's p < 0.05), and was inversely correlated with tumor size (Spearman's rho CC = -0.402, p < 0.001) in a logarithmic distribution (p < 0.001). A positive correlation (Spearman's rho CC = 0.312, p = 0.006) in cubic distribution (p < 0.001) was observed between p53and TFAM IHC pixelwise H-scores. ROC analysis yielded a TFAM IHC pixelwise H-score cutoff of 43.0 (AUC = 0.650, 95%CI = 0.515-0.785, p = 0.047; sensitivity = 0.490, specificity = 0.810) to dichotomize high and low groups. In multi-variate Cox regression, low TFAM IHC pixelwise H-score (HR = 2.332, 95%CI = 1.136-4.787, p = 0.021), M1 status (HR = 3.582, 95%CI = 1.608-7.979, p = 0.002), and perineural invasion (HR = 4.506, 95%CI = 1.541-13.177, p = 0.006) were identified as independent variables to poor prognosis with elevated HRs. Among patients with high TFAM expression, higher HER2 IHC pixelwise H-score was associated with elevated hazard (HR = 1.010, 95%CI = 1.002-1.019, p = 0.020, Cox regression, uni-variate). Among M1 patients, higher p53 IHC pixelwise H-score was related to elevated hazard (HR = 1.029, 95%CI = 1.004-1.056, p = 0.025, Cox regression, uni-variate).

Conclusions: ROC and multi-variate Cox regression identified low TFAM expression as an independent poor prognostic variable for operable GAC patients, implying its potential as a quantitative prognostic biomarker. The observed associations with p53 and HER2 are hypothesis-generating and they require further validation to clarify TFAM's role in Warburg effect and GAC progression.

Keywords: Gastric adenocarcinoma (GAC); HER2; Mitochondrial transcription factor A (TFAM); Warburg effect; p53.

Fig. 1

Flowchart illustrates the inclusion and exclusion criteria for recruiting GAC patients. A total of 77 cases were included in Cohort A for IHC analysis and pathological correlation, and 72 cases were included in Cohort B for prognostic evaluation. GAC, gastric adenocarcinoma; IHC, immunohistochemical

Fig. 2

Workflow for semiquantitative IHC pixelwise H-score analysis. Example of TFAM IHC pixelwise H-score in GAC case 68 is demonstrated. (a) ROI selection on WSI. A low-power overview (1× objective) of a TFAM-stained GAC WSI is shown at top panel. Five ROIs were sampled using a central + four-quadrant scheme. The central ROI is displayed at 40× objective on the bottom panel. The navigator thumbnail (bottom left of the 40× view) indicates the ROI’s location within the full WSI. (b) Color deconvolution in ImageJ. The 40× central-ROI image (middle panel) was opened in ImageJ (top panel) and processed via IHC Profiler plugins, yielding separate hematoxylin (left) and DAB (right) channel images (bottom panel) for quantification. (c) Pixelwise H-score calculation. The DAB-channel ROI was analyzed with the IHC Profiler plugin, which classifies pixels as high positive, positive, low positive, or negative. A representative pixel-intensity histogram (0–255 gray levels) shows the distribution within the ROI (top panel), and the plugin’s log output reports the percentage of pixels in each category (middle panel). The overall pixelwise H-score is calculated as (low panel) : Pixelwise H-score = (3 × %high positive) + (2 × %positive) + (1 × %low positive) + (0 × %negative); 0 ≤ Pixelwise H-score ≤ 300. The TFAM IHC pixelwise H-score of the central ROI was then calculated as follows: (27.3888 × 3) + (29.3678 × 2) + (13.7202 × 1) + (29.5232 × 0) = 154.6. GAC, gastric adenocarcinoma; IHC, immunohistochemical; ROI, region of interest; TFAM, mitochondrial transcription factor A; WSI, whole slide image; DAB, 3,3′-diaminobenzidine

Fig. 3

(a) In Cohort A of 77 GAC patients, a significant negative correlation was observed between TFAM IHC pixelwise H-scores (Y-axis) and maximal tumor diameters (X-axis, cm), based on non-parametric Spearman’s rank correlation (CC, Spearman’s Rho=−0.402, p < 0.001). A regression analysis evaluating multiple curve fitting models identified the logarithmic model as the best fit one (R²=0.178, p < 0.001). The logarithmic trendline is displayed as a solid black line. (b) In the same cohort, a significant positive correlation was observed between p53 IHC pixelwise H-scores (Y-axis) and TFAM IHC pixelwise H-scores (X-axis), based on non-parametric Spearman’s rank correlation (CC, Spearman’s Rho = 0.312, p = 0.006). A regression analysis evaluating multiple curve fitting models identified the cubic model as the best fit one (R²=0.262, p < 0.001). The cubic trendline is displayed as a solid black line. GAC, gastric adenocarcinoma; CC, correlation coefficient; IHC, immunohistochemical; TFAM, mitochondrial transcription factor A

Fig. 4

Using the Log-rank test and a multi-variate Cox regression model with the enter method, three independent predictors to distinguish survival differences and mortality risks were identified in Cohort B GAC patients (n = 72). (a) Kaplan–Meier survival curves demonstrate that M0 GAC patients (n = 59; solid black line) had a significantly better overall survival (mean = 58.1, 95%CI = 43.0–73.3, months) than did M1 patients (n = 13; dashed black line; mean = 11.0, 95%CI = 6.7–15.4, months, p < 0.001, Log-rank test). Using the M0 group as the reference (HR = 1.000), the HR for M1 patients was obviously elevated to 3.582 (95%CI = 1.608–7.979, p =  0.002, multi-variate Cox regression). The number of patients at risk is shown below at 12-month intervals. (b) Kaplan–Meier survival curves demonstrate that GAC patients without pleural invasion (n =  25 , solid black line) had a significantly better overall survival (mean = 86.3, 95%CI = 62.2–110.3, months) than did GAC patients with pleural invasion (n = 47, dashed black line, mean = 29.7, 95%CI = 17.8–41.7, months, p < 0.001, Log-rank test). Using the GAC patients without pleural invasion as the reference (HR = 1.000), the HR for GAC patients with pleural invasion was obviously elevated to 4.506 (95%CI = 1.541–13.177, p = 0.006, multi-variate Cox regression). The number of patients at risk is shown below at 12-month intervals. (c) Kaplan–Meier survival curves demonstrate that GAC patients with high TFAM IHC pixelwise H-score (> 43.0, n = 43, solid black line) had a significantly better overall survival (mean = 63.2, 95%CI = 45.3–81.2, months) than did GAC patients with low TFAM IHC pixelwise H-score (≤ 43.0, n = 29, dashed black line, mean = 30.5, 95%CI = 14.2–46.8, months, p = 0.004, Log-rank test). Using the GAC patients with high TFAM IHC pixelwise H-score (≤ 43.0) as the reference (HR = 1.000), the HR for GAC patients with low TFAM IHC pixelwise H-score (> 43.0) was obviously increased to 2.332 (95%CI = 1.136–4.787, p = 0.021, multi-variate Cox regression). The number of patients at risk is shown below at 12-month intervals. GAC, gastric adenocarcinoma; TFAM, mitochondrial transcription factor A; IHC, immunohistochemical; HR, hazard ratio

Fig. 5

Clinicopathological observation in GAC patients, biomolecular phenomenon in the literatures and the cautious hypothesis linking to Warburg effect in GAC. Clinical observations from our cohort: (a) lower TFAM IHC pixelwise H-scores were associated with disease progression, including advanced T/N stage, lymphovascular invasion, perineural invasion, larger tumor diameter, along with poor cell differentiation and worse survival; (b) TFAM and p53 IHC pixelwise H-scores showed a positive correlation (best fit: cubic); (c) lower HER2 and p53 IHC pixelwise H-scores were also associated with poor cell differentiation; (d) among M1 patients, higher p53 IHC pixelwise H-scores were associated with increased HR; (e) within the high-TFAM subgroup, higher HER2 IHC pixelwise H-scores were associated with increased HR. Literature-derived mechanisms: (f) TFAM regulates mtDNA replication and transcription; (g) p53 can promote mitochondrial biogenesis via the PGC-1α → NRF_1/2 → TFAM axis; (h) HER2 activates PI3K/Akt signaling to enhance glycolysis. Warburg effect and cautious hypothesis: (i) the Warburg phenotype features increased glycolysis with relative mitochondrial dysfunction; (j) we hypothesize that reduced TFAM may mark or contribute to Warburg-related progression in GAC, potentially modulated by p53 and HER2; this requires prospective validation. GAC, gastric adenocarcinoma; IHC, immunohistochemical; TFAM, mitochondrial transcription factor A; mtDNA, mitochondrial DNA; PGC-1α, peroxisome proliferator-activated receptor-γ coactivator-1α; NRF_1/2, nuclear respiratory factors 1/2; PI3K, phosphoinositide 3-kinase; HR, hazard ratio