Predictive value of intratumoral-metabolic heterogeneity derived from 18F-FDG PET/CT in distinguishing microsatellite instability status of colorectal carcinoma

Purpose/background Microsatellite instability (MSI) status is a significant biomarker for the response to immune checkpoint inhibitors, response to 5-fluorouracil-based adjuvant chemotherapy, and prognosis in colorectal carcinoma (CRC). This study investigated the predictive value of intratumoral-metabolic heterogeneity (IMH) and conventional metabolic parameters derived from 18F-FDG PET/CT for MSI in patients with stage I–III CRC. Methods This study was a retrospective analysis of 152 CRC patients with pathologically proven MSI who underwent 18F-FDG PET/CT examination from January 2016 to May 2022. Intratumoral-metabolic heterogeneity (including heterogeneity index [HI] and heterogeneity factor [HF]) and conventional metabolic parameters (standardized uptake value [SUV], metabolic tumor volume [MTV], and total lesion glycolysis [TLG]) of the primary lesions were determined. MTV and SUVmean were calculated on the basis of the percentage threshold of SUVs at 30%–70%. TLG, HI, and HF were obtained on the basis of the above corresponding thresholds. MSI was determined by immunohistochemical evaluation. Differences in clinicopathologic and various metabolic parameters between MSI-High (MSI-H) and microsatellite stability (MSS) groups were assessed. Potential risk factors for MSI were assessed by logistic regression analyses and used for construction of the mathematical model. Area under the curve (AUC) were used to evaluate the predictive ability of factors for MSI. Results This study included 88 patients with CRC in stages I–III, including 19 (21.6%) patients with MSI-H and 69 (78.4%) patients with MSS. Poor differentiation, mucinous component, and various metabolic parameters including MTV30%, MTV40%, MTV50%, and MTV60%, as well as HI50%, HI60%, HI70%, and HF in the MSI-H group were significantly higher than those in the MSS group (all P < 0.05). In multivariate logistic regression analyses, post-standardized HI60% by Z-score (P = 0.037, OR: 2.107) and mucinous component (P < 0.001, OR:11.394) were independently correlated with MSI. AUC of HI60% and our model of the HI60% + mucinous component was 0.685 and 0.850, respectively (P = 0.019), and the AUC of HI30% in predicting the mucinous component was 0.663. Conclusions Intratumoral-metabolic heterogeneity derived from 18F-FDG PET/CT was higher in MSI-H CRC and predicted MSI in stage I–III CRC patients preoperatively. HI60% and mucinous component were independent risk factors for MSI. These findings provide new methods to predict the MSI and mucinous component for patients with CRC.


Introduction
The world's second leading cause of cancer death is colorectal cancer (CRC), which is the third most common cancer (1).While surgery and adjuvant chemotherapy are currently the main treatments for CRC, immunotherapy has emerged as a new therapeutic strategy for cancer treatment.The programmed cell death 1 (PD-1)/programmed death-ligand 1 (PD-L1) axis is a critical immune response.Anti-PD-1/ anti-PD-L1 blockade therapy enhances the antitumor activity of T cells by blocking the binding of PD-L1 on tumor cells and PD-1 on immune T cells, representing a key anticancer strategy.Blocking these pathways effectively reduces tumor growth and improves survival in most solid tumors (2), especially in microsatellite instability-high (MSI-H) CRC (3).Nivolumab and pembrolizumab, PD-1 inhibitors, were officially approved by the FDA in 2017 for the treatment of MSI-H solid tumors (4), ushering in a new era of immunotherapy.
Under proficient mismatch repair (pMMR) conditions, errors in the replication of microsatellites are repaired by MMR proteins.Mismatch repair deficiency (MMR-D) can lead to errors beyond repair normally and has been detected in endometrial cancer, CRC, and gastric cancer (5).MSI results from errors in the DNA replication of microsatellites and is observed in approximately 12%-15% of localized CRC cases and 4% of stage IV CRC cases (2,6).The accumulation of MSI eventually leads to development of MSI-H.MSI status has been regarded as an important prognostic biomarker for patients with CRC and affects the selection of adjuvant 5-fluorouracilbased chemotherapy or immune checkpoint inhibitors.Studies showed that 5-fluorouracil-based adjuvant chemotherapy had limited therapeutic effect for patients with MSI-H CRC (7,8).MSI tumors have large proportions of mutant neoantigens and tumor-infiltrating lymphocytes (TILs) (2,6,9) and a high tumor mutational burden (TMB) that promotes the infiltration of immune cells; therefore, MSI plays a significant role in long-term durable immunotherapy (10,11).
CRC, especially CRC with MSI-H, exhibits significant heterogeneity, which refers to the different characteristics within a tumor in regard to genomics, histopathologic features, tumor microenvironment, T-cell infiltrate, and response to therapies (9,12,13).MSI is generally determined by evaluation of MMR proteins including MLH1, PMS2, MSH2, and MSH6 using immunohistochemical (IHC) or by PCR testing and next-generation sequencing to test MMR genes.However, the above methods are usually invasive and may be affected by intratumoral heterogeneity.Therefore, it is meaningful to screen tumors for MSI status to avoid unnecessary inspections and promote the immunotherapy of early-stage CRC patients.
PET/CT is a systemic examination that can detect the distant metastasis of tumors.In recent years, 18 F-FDG PET/CT has demonstrated efficiency in staging, predicting prognosis, assessing treatment response, and determining gene mutational status in patients with CRC (14)(15)(16)(17).MSI-H CRCs have higher intratumoral heterogeneity compared with microsatellite stability (MSS) CRCs (9,12,13,18). 18F-FDG PET/CT can reflect the intratumoral-metabolic heterogeneity (IMH) by the heterogeneity index (HI) and the heterogeneity factor (HF).These two quantitative indicators reflect IMH and have already been shown to predict survival outcomes in epithelial ovarian cancer, gastric cancer, and oral cavity squamous cell carcinoma (19)(20)(21).However, the role of IMH derived from 18 F-FDG PET/CT in predicting MSI in CRC patients is unknown (22,23).Therefore, we aimed to explore the ability of IMH and conventional metabolic parameters (standardized uptake value [SUV], metabolic tumor volume [MTV], and total lesion glycolysis [TLG]) derived from 18 F-FDG PET/CT for the prediction of MSI status in patients with stage I-III CRC.

Patient inclusion and exclusion
This study included patients who underwent 18 F-FDG-PET/CT in Wuhan Union Hospital from January 2016 to May 2022.The inclusion criteria were as follows: patients with (a) colorectal lesions diagnosed as colorectal adenocarcinoma with or without a mucinous component (part of the tumor volume with mucin associated likely produced by malignant glands or forming mucin pools, which account for 5%-100% of the tumor volume) by pathology; (b) complete 18 F-FDG PET/CT images before surgery within 1 month of surgery; and (c) IHC staining evaluated MMR protein expression status.
The exclusion criteria were as follows: (a) patients who received neoadjuvant chemoradiotherapy or surgery before 18 F-FDG PET/CT examination (n = 41); (b) the primary lesion did not show metabolism on PET/CT (n = 3); (c) the presence of other tumors that interfered with the measurement of the CRC primary tumor (n = 2); and (d) CRC patients in stage IV (n = 18).The flowchart is shown in Figure 1.

PET/CT imaging acquisition protocol
All patients got whole-body imaging (from the cranium to the upper third of the thighs) by 18 F-FDG PET/CT scans with a dedicated PET/CT system (GE Healthcare, Milwaukee, WI, USA; Discovery VCT and Discovery LS).An intravenous injection of 3.7-4.4MBq/kg 18 F-FDG was administered, after at least 6 h of fasting and a blood glucose level drops below 200 mg/dl.The examination was performed approximately 60 min after the injection of 18 F-FDG.A slice spiral CT scan was first performed.The PET imaging was collected in threedimensional mode at 2 min per frame for a total of seven to eight frames.PET data were reconstructed on the basis of ordered-subset expectation maximization.The reconstructed images were processed on a workstation (Xeleris Workstation, GE Healthcare).

Measurement of PET metabolic parameters
The PET/CT image was evaluated in a blind manner by two nuclear medicine doctors, each with more than 6 years of clinical experience.The ROI of the primary lesions of CRC patients was delineated in the workstation.The quadrate working frame was placed on the primary lesion, and if necessary, areas of physiologic uptake or nearby lymph node metastasis were manually excluded.The SUV max of the primary lesion was measured automatically by the workstation (AW4.6;GE Healthcare).The SUV mean and MTV were obtained by the percent threshold of SUV max in the working frame.MTV 30%-70% were calculated with 30%-70% of SUV max as the percent threshold, respectively (23).SUV mean was obtained from the corresponding MTV.TLG = MTV × SUV mean .IMH contains the HI and HF.HI = SUV max / SUV mean .HI 30%-70% was determined from the primary tumors.In addition, the HF was the absolute value of the linear regression slope calculated by the least square method from the different threshold of the ROI (Figure 2) (21).

Immunohistochemical evaluation
Clinicopathological data, including differentiation grade, mucinous component, TN stages, tumor deposit, lymphovascular invasion, perineural invasion, and IHC results for four MMR proteins (MLH1, MSH2, MSH6, PMS2), were evaluated by pathologists.The absence of any MMR protein was defined as MMR-D, also known as MSI-H.The expression of all four MMR proteins was defined as pMMR, characterized as MSS.All patients in stage I-III CRC had undergone surgery within 1 month after the PET/ CT scan.

Statistical analysis
The quantitative variable distribution was analyzed between groups.Following the distribution of data, data were expressed by mean ± standard deviation or median (interquartile range) and analyzed by T-test or Mann-Whitney U test, respectively.The categorical data were examined using the Fisher's exact test.Various metabolic parameters were standardized by Z-score normalization before performing logistic regression analysis, as the significant differences in various metabolic parameter values.The mathematical model was built, as described in the study by Wang et al. (24).The ROC curve was used to estimate the predictive value of different metabolic parameters for MSI.The DeLong test was used to assess the variation in AUC values.All data were analyzed by SPSS software version 23 (IBM Corp, Armonk, NY, USA) and MedCalc software (version 19.7.2).Statistical significance was defined as a twosided P value less than 0.05.

Patient characteristics
This study included 88 patients, including 59 male patients (70.2%) and 29 female patients (29.8%).The average age was 65 years (range 28-89 years).Among the total 88 patients, 19 (21.6%) patients were categorized into the MSI-H group and 69 (78.4%) patients were categorized into the MSS group.Patients and tumor characteristics are summarized in Table 1.The MSI-H CRC group tended to have a mucinous component, poor differentiation, and a larger maximum tumor diameter, with a higher frequency of male patients compared with the MSS CRC group, with statistical significance (all P < 0.05).Age, primary tumor location, TN stage, AJCC-TNM stage, tumor deposit, lymphovascular invasion, and perineural invasion were not significantly different between the MSI-H group and the MSS group (all P > 0.05).

Differences in PET/CT metabolic parameters between the MSI-H group and the MSS group
In terms of conventional metabolic parameters, the MSI-H group tended to have higher MTV 30% , MTV 40% , MTV 50% , and MTV 60% compared with the MSS group, with significant differences (all P < 0.05).Regarding IMH, HI 60% , HI 70% , and HF were significantly higher (all P < 0.05) in the MSI-H group compared with the MSS group.No Workflow of 18 F-FDG PET/CT metabolic parameter measurements.When using the percent threshold of 40%, SUV max was 19.04, the volume of ROI was 26.91 cm 3 , SUV mean was 10.98, and TLG was 293.05.HI = SUV max /SUV mean = 1.75 (A).The HF is the absolute value of the linear regression slope calculated by the least square method from the different threshold of the ROI; the threshold of tumor volumes ranged from 30% to 70% of SUV max (B).S1.

Independent predictive indicators of PET/CT parameters and clinical parameters for MSI
The aforementioned significant characteristics were included into logistic analysis.Univariate logistic analysis showed that the expression of MSI in CRC was associated with mucinous component, differentiation, and standardized parameters Z-HI 50% and Z-HI 60% (all P < 0.05) (Table 2).Multivariate logistic regression analysis showed that mucinous component and Z-HI 60% were the independent predictive indicators for MSI of the CRC (Table 3).

Model establishment and mucinous composition exploration
MSI-H CRC had higher IMH and MTV values in preoperative PET/CT examination and were more prone to exhibit a mucinous component than MSS CRC (Figures 3, 4).A model of the HI 60% + mucinous component was established, and the formula was as follows: y = exp(x)/[1+exp(x)], x = -2.083+ 0.749 × Z-HI60%+ 2.589 × mucinous component (the "-" or "+" of mucinous component was defined as "0" or "l") (Table S3).The cutoff value determined by ROC analysis was y = 0.1432, when y > 0.1432 was considered as MSI-H CRC, while y < 0.1432 was considered as MSS CRC.The AUC of HI 60% and the mathematical model were 0.685 and 0.85, with a sensitivity of 0.579 and 0.842 and specificity of 0.826 and 0.768, respectively.There was a significant difference in AUC between HI 60% and our model (Z = 2.339, P = 0.019).The predictive ability of ROC curves is illustrated in Figure 5.
The Mann-Whitney U test was used to further evaluate the connection between mucinous components and various metabolic parameters.HI 30% was significantly higher in CRC with mucinous component than in CRC without mucinous component.The AUC of HI 30% was 0.663, with sensitivity of 0.875 and specificity 0.5, respectively.The cutoff value was 1.974.The performance of PET/ CT parameters is illustrated in Table S2.

Discussion
Our study examined the value of 18 F-FDG PET/CT imaging parameters in identifying MSI in stage I-III CRC and explored the intratumoral heterogeneity.As we know, this is the first study examining the utility of IMH determined by PET/CT to predict MSI of CRC.Our results showed that HI 60% and the mucinous component were independent prediction factors for MSI status in stage I-III CRC.CRC is tumor with high glucose metabolism that rapidly converts glucose to lactate (known as the "Warburg effect") (25,26).Other effects include m6A modifications (27), glycolysis-associated lncRNA of colorectal cancer (GLCC1) (28), and insulin-like growth factor 1 (IGF-1) signaling (29), which could also enhance glycolysis in different paths.SUV max , as a semiquantitative parameter in the 18 F-FDG uptake of lesions, is considered to be an auxiliary index for judging benign and malignant tumors and a general prognostic factor in malignancies (30).The limitation of SUV max is that it only reflects a single voxel value and cannot reflect the characteristics of the entire ROI.MTV is an indicator usually based on a fixed SUV threshold of 2.5 or percentage SUV max threshold of 40%-42% (30), reflecting the tumor burden of the tumor; it plays an important role in predicting prognosis (30) and even target volume delineation before radiotherapy (31).TLG can reflect both volume and metabolic activity of lesions, which can reflect the advantage of PET/CT better.While SUV max , MTV, and TLG are established as parameters to assess for diagnosis, efficacy, prognosis, tumor burden, and TMB in cancer (32, 33), they have limited value in judging IMH.
IMH is one of the characteristics of solid tumors.The measurement of tumor heterogeneity is of great significance to the drug resistance, tumor invasion, and prognosis of patients (33,34).The difference in proliferation, receptor expression, necrosis, hypoxia, energy metabolites, micro-vessel density, protein expression, and gene expression in tumors may lead to uncertainties in therapeutic effects (9,18,35,36). 18F-FDG PET/CT can reflect tumor metabolic heterogeneity through different parameters, including HI (21,33), HF (19), coefficient of variation (19,20,33), and texture parameters (34).In our research, HI was calculated by dividing SUV max by SUV mean for the tumor, which reflects not only the 18 F-FDG metabolism in the ROI of tumors but also the average of metabolic activity.Research showed that higher HI 30% levels were related to poorer overall survival in patients with oral cavity squamous cell carcinoma (21).In the study by Kim et al. (37), a lower HI based on a threshold of 2.5 of the primary tumors was associated with better prognosis in locally advanced nasopharyngeal carcinoma.Compared with calculation of the HI, calculation of the HF is more complicated and determined by linear regressions of MTVs under different SUV thresholds, determined by two methods (percentage threshold method and fixed threshold method).In general, the percentage threshold method is suitable for lesions with high SUV max , which avoids interference with the surrounding normal tissue metabolism (21).The fixed threshold method is more suitable for tumors with a low 18 F-FDG uptake (38).For this reason, we selected the percentage threshold method in our study.In a study of 55 patients with advanced gastric cancer (19), HF (SUV threshold based on 2.5-3.5) was significantly associated with TNM stage and overall survival.
In our study, several conventional metabolic parameters were significant predictors of MSI, which was similar to previous findings (23,39).Song et al. (39) retrospectively analyzed the relationship between conventional metabolic parameters based on the threshold of 40% SUV max in 420 CRC patients.The results proved that MTV ≥32.19 cm 3 of CRC was linked to the presence of MSI-H and the increased density of TILs in MSI-H CRC may lead to higher TMB.The result was consistent with the previous study of Liu et al. (23) and our results (Table 4).Moreover, our analysis further proved the existence of IMH of CRC.The HI is the ratio of the highest 18 F-FDG uptake and the average 18 F-FDG uptake of the lesions.MSI-H CRCs showed a higher HI compared with MSS CRC even though SUV max and SUV mean were not statistically different in the two groups, which might be caused by the larger proportion of mutant neoantigens increasing glucose metabolism and upregulated immune checkpoints of MSI-H CRC that preserve MSI-H CRC from the pernicious immune microenvironment (2).Therefore, HI 60% could make up for the limitation of conventional metabolic parameters in predicting MSI.HF reflects the metabolic changes in tumors under different thresholds.In our research, MSI-H CRC had a higher HF than MSS CRC, which indicated a higher heterogeneity in MSI-H CRC.
MSI-H CRC is a tumor with higher heterogeneity and increased TMB and tumor burden compared with MSS CRC.MSI-H CRC has more TILs in both tumor and tumor-adjacent tissues.The higher   Mucins are highly O-glycosylated glycoproteins that are essential for a variety of biological processes.Tumor growth results in an unfavorable microenvironment; however, mucins help to evade acidity, hypoxia, and other inhospitable conditions that block drug delivery and promote cancer progression.Mucin is classified into more than 20 subtypes, and the functions are not fully understood (41).MUC2 and MUC5AC mucins produced by colonic goblet cells tend to overexpress in mucinous CRC and MSI-H CRC (42).The EGFR pathway substantially leads to the formation of CRC with the mucinous component, regardless of the percentage (43).Additionally, the mutation rate of RAS/RAF/MAPK and PI3K/AKT pathways in mucinous adenocarcinoma (with more than 50% of mucinous component) is higher than that in non-mucinous adenocarcinoma (42).Our study revealed a higher HI 30% in CRC with a mucinous component than in CRC without a mucinous component.Therefore, 18 F-FDG PET/CT may contribute to detect the existence of a mucinous component in CRC and help to predict MSI of CRC.
Multivariate logistic regression analysis demonstrated that Z-HI 60% and mucinous component were independent risk factors of MSI in stage I-III CRC.These results indicate that more attention should be given to heterogeneity of tumors in the early stage.MSI-H in CRC is associated with pathological features such as mucinous carcinoma, proximal colon, poorly differentiation, lymphatic invasion, tumor staging, tumor size, KRAS mutation, and BRAFV600E mutation (2,6); some results of our study are consistent with these findings.CRC with MSI-H more frequently appears in stage II cases (approximately  20%) compared with stage III (approximately 12%) and is relatively uncommon among stage IV CRC (around 4%) (6).Compared with pMMR/MSI tumors, stage I-III CRC with MSI-H generally have a better prognosis.Stage IV CRC has significant heterogeneity compared with stage I-III CRC, especially between the metastasis and primary tumors (35).The assessment of M staging is one of the most obvious advantages of PET/CT examination.Early detection of intratumoral heterogeneity may be helpful for the formulation of clinical treatment decisions; to avoid the influence of the differences of intertumoral heterogeneity, we selected stage I-III CRC into our research.
The proportion of MSI-H in our study (21.6%) was higher than that in epidemiological reports (12%-15%), which may be because MSI-H CRC is more common in hypermetabolic CRC and makes it easier to be found by PET/CT.In addition, our research object was stage I-III CRC, while MSI-H CRC accounts for only 4% of stage IV CRC (6), which implied that many MSS CRC patients were not included in this study.Although we had built a mathematical model, the limitations of our study include the limited number of cases and the absence of a validation cohort to support, and thus further data validation is required.As a retrospective study, selective bias was inevitable.In our research, imaging data were obtained from two different equipment.While monthly quality control was routinely performed and the same workstation was used, there may still be some uncontrollable differences in the measurement of metabolic parameters by the two devices.Finally, our research was based on a single center, and therefore data from more centers need to be included to validate the utility of IMH and conventional metabolic parameters for predicting MSI in patients with CRC.

Conclusion
This study not only demonstrated that MSI-H CRCs have a higher tumor metabolic burden than MSS CRCs but also revealed a higher IMH in MSI-H CRCs compared with MSS CRCs.The HI 60% derived from 18 F-FDG PET/CT and mucinous component were the independent risk factors for MSI in CRC.The mathematical model from HI 60% + mucinous component demonstrated the highest predictive performance.PET/CT is a non-invasive approach to evaluate MSI and the mucinous component in CRC and will help in guiding immunotherapy in CRC patients.

FIGURE 1 Flowchart
FIGURE 1Flowchart of patient selection.

FIGURE 2
FIGURE 2 CD3+ and CD8 T lymphocyte densities are associated with higher tumor size and MTV value(9, 12,39).The consumption of glucose by tumor limits T-cell metabolism, leading to the dampened glycolytic capacity, mTOR activity, and IFN-g production, representing the adaptation of resistance to the MSI-induced immunoreactive microenvironment and thereby allowing tumor progression.Due to the inhibition of TIL metabolism, the insignificant change of SUV mean in HI 60% can be explanted.Checkpoint blockade antibodies against PD-1, PD-L1, and CTLA-4 can restore glycolysis, IFN-g production, and immune responses in T cells (36, 40) and lead to an insignificant decrease of SUV mean in MSI-H solid tumors in the early stage of immunotherapy.

FIGURE 5 ROC
FIGURE 5ROC curve for HI 60% and the mathematical model in predicting MSI.
difference was found in SUV max and other metabolic parameters between the groups.Details of IMH and conventional metabolic parameters in the MSS group and MSI-H group are listed in Table significant

TABLE 1
Clinicopathologic characteristics of patients.
MSI, microsatellite instability; MSS, microsatellite stability; AJCC, American Joint Committee on Cancer; TNM, tumor-node-metastasis.P values were calculated using the *Fisher's exact test and # T-test.

TABLE 2
Univariate regression analyses for predicting MSI in CRC patients.

TABLE 3
Multivariate regression analyses for predicting MSI in CRC.

TABLE 4 PET
/CT for predicting MSI.