Assessing tumor heterogeneity and FLOT/FOLFOX-induced alterations in cancer stem cells and cancer immune checkpoint molecules using patient-derived tumor organoids

Abstract

Background

Gastric cancer is a deadly disease with poor prognosis and unfavorable therapeutic outcomes. It is known that a major cause of treatment failure for gastric cancer is related to tumor heterogeneity. However, it is difficult to assess heterogeneity, as the composition of tumors is dynamic and cancer treatments reshape cell diversity and gene/protein expression profiles.

Objective

We utilized gastric cancer patient-derived tumor organoids to monitor tumor heterogeneity and the impacts of chemotherapy (FLOT and FOLFOX) on cancer stem cells (LGR5 and CD133) and cancer immune checkpoint molecules (MHC-I and PD-L1).

Methods

Six tumor organoids were established from different patients or different regions of the same tumor. The tumor organoids were used to investigate the real-time profiles of cancer stem cell marker expression and cancer immune checkpoint molecule expression after chemotherapy treatment. Tumor suppression was monitored by the propidium iodide assay and measurements of tumor organoid diameter. Expression levels of cancer stem cell markers and immune checkpoint molecules were analyzed by flow cytometry.

Results

Growth of all tumor organoids could be suppressed by chemotherapy. However, diverse impacts of FLOT and FOLFOX were observed in terms of the expression profiles of cancer stem cell markers and cancer immune checkpoint molecules. In particular, tumor organoids from different patients exhibited distinct expression profiles for cancer stem cell markers and immune checkpoint molecules after treatment with the same chemotherapeutic regimen. Similar observations were made with tumor organoids derived the same patient but different tumor regions.

Conclusion

Our findings demonstrated that spatial and temporal tumoral heterogeneity can be observed after treatment with chemotherapy, suggesting that repeated monitoring of tumor characteristics may be essential for clinical decision making. In addition, sampling of multiple sites at multiple time-points might be important when evaluating the impacts of chemotherapy on cancer stem cells markers and immune checkpoint molecules.

Introduction

Gastric cancer is one of the most frequently diagnosed cancers and the third-leading cause of cancer-related death globally. ¹ Currently, multiple treatment options are available for gastric cancer, including surgery, chemotherapy, targeted therapy and immunotherapy. ¹ Except for surgery, the responses to all of these therapeutic approaches are highly affected by tumor heterogeneity.^2,3 Tumor heterogeneity refers to the diversity of cellular compositions, cellular phenotypes and cellular morphologies within and among tumors.^4,5 Different subsets of tumor cells may display unique gene or protein expression patterns which cause tumor cells respond differently to the same treatments, even if the cells are present in the same tumor.^6–8 For instance, different tumor cells may have different levels of expression for human epidermal growth factor receptor 2 (HER2) and programmed death-ligand 1 (PD-L1), which respectively predict therapeutic responses to trastuzumab and anti-PD-1/PD-L1 inhibitors. Variations of PD-L1 and HER2 expression were previously reported in different regions of gastric cancer tumors.^7,8 Furthermore, a clinical study revealed that gastric cancer patients with homogeneous HER2 expression exhibit better therapeutic responses to trastuzumab-based chemotherapy (capecitabine or 5-fluorouracil plus cisplatin) than patients with heterogeneous HER2 expression. ⁷ Moreover, chemotherapy treatments can lead to acquired resistance, which limits the efficacy of the treatment, ⁹ as well as reshaped tumor heterogeneity.^10,11

Cancer stem cells (also as known as tumor-initiating cells) are among the most important contributors to therapeutic resistance. ¹² In these cells, genetic and epigenetic alterations are crucially involved in plasticity of differentiation. ¹² Notably, chemotherapy induces cellular stress in cancer stem cells, which cause genetic and epigenetic alterations that render cancer stem cells capable of differentiating into lineages with therapy-resistant or aggressive phenotypes.^13–15 Thus, monitoring plasticity-associated alterations in cancer stem cells during course of chemotherapy could reveal the effects of chemotherapy on cancer stem cells and aid in the development of new approaches to prevent therapeutic resistance. However, monitoring the cancer stem cells during the course of treatment is difficult in clinical practice. In addition, there are no gold standard biomarkers that can be used to identify and isolate cancer stem cells. Instead, researchers typically isolate cancer stem cells via a few specific biomarkers, as not every subset of cancer stem cells expresses the same biomarkers.^16,17 As a result, subsets of cancer stem cells might be overlooked in the isolation procedure. Since the purified cancer stem cells might not fully reflect the entire population in a tumor, new monitoring strategies are needed.

In general, tumor resection surgery is performed soon after gastric cancer diagnosis. After surgery, the patients are given chemotherapy if lymph node metastasis is noted. ¹⁸ Upon cancer recurrence, anti-PD-1/PD-L1 inhibitors with chemotherapy are among the most common treatments. ¹⁹ This combination therapy is only prescribed if a patient’s PD-L1 combined positive score (CPS) meets a criteria of CPS ≥ 5. ²⁰ Although PD-L1 expression is measured from the surgically resected tumor specimen, ²¹ it is unclear whether PD-L1 expression in the newly diagnosed tumor can accurately reflect the level of PD-L1 expression in the recurrent tumor. In fact, one clinical study revealed that PD-L1 expression levels change dynamically in non-small cell lung cancer (NSCLC) patients during the course of disease. ²² In addition, recent evidence suggests that downregulation of the neoantigen presenter major histocompatibility complex class I (MHC-I) is correlated with poor efficacy of PD-1/PD-L1 therapy.^23,24

Tumor organoids are generated from cancer stem cells. ²⁵ These in vitro organoids retain the heterogeneity of parental tumors ²⁵ and can therefore serve as a platform to predict or monitor the efficacy of treatments for different tumors. In this study, we utilized tumor organoids to investigate tumor heterogeneity and to perform real-time monitoring of chemotherapy-induced alterations in cancer stem cell markers (LGR5 and CD133) and immune checkpoint molecules (MHC-I and PD-L1).

Materials and methods

Chemicals

5-Fluorouracil (5-FU), oxaliplatin (OXA), leucovorin (Leu), and docetaxel (DTX) were purchased from MedChemExpress. Matrigel was purchased from Corning. Advanced DMEM and propidium iodide (PI) was purchased from Thermo Fisher Scientific. Dispase II, collagenase and Penicillin-Streptomycin-Amphotericin B solution were purchased from Merck.

Tumor organoids and patient characteristics

Organoids were established from the resected tumor tissues of patients diagnosed with gastric cancer, with minor modifications to a previously described protocol. ²⁶ Briefly, tumor tissues were minced with a scalpel blade and digested with collagenase XI (Merck) and dispase II (Merck) at 37°C for 1 h. Then, the tissue debris was resuspend in advanced DMEM/F12 and filtered through a 100-μm cell strainer. Afterward, the filtrate was centrifuged. The cell pellets were collected and resuspended in advanced DMEM/F12 containing ice-cold 70% Matrigel (Corning). The cell suspension (cells with Matrigel) were seeded onto ultralow attachment 6-well plates (Corning), at 20 μl per drop. After the droplets polymerized, organoid culture medium was added, and the cultures were incubated at 37°C cell incubator. We noticed that the growth condition of each organoid was different. Patient 11’s organoids stopped growing within 8 months. The patient characteristics, including age, gender, tumor histology, tumor/node/metastasis (TNM) stage, and mismatch repair (MMR) status were recorded. The pathological characteristics of the patient’s tumors were determined by pathologists at Wan Fang Hospital, Taipei Medical University. The ethics committee of the Taipei Medical University (TMU)-Joint Institutional Review Board granted approval for this study (N202405105). Written informed consent was performed prior to starting of the study.

Cell size and cell viability measurement

After gastric tumor tissues were digested, the suspension was filtered through a 100-μm cell strainer and centrifuged. The cell pellets were resuspended in PBS and stained with trypan blue. Cell size and cell viability were measured using a Cellometer Auto T4 (Nexcelom).

Hematoxylin and eosin staining

Tumor organoids were harvested and fixed with 10% formaldehyde at room temperature. Then, the fixed tumor organoids were sectioned and stained with hematoxylin and eosin (H&E).

Drug treatment, chemotherapy response and flow cytometry

A total of 5,000 dissociated tumor cells were resuspended in 100 μl tumor organoid culture medium containing 0.5% Matrigel (Corning). Then, the suspended cells were seeded in an ultralow attachment 96-well plate (Corning). For testing chemotherapy response, tumor organoids were incubated with FLOT (5-FU: 200 μM, OXA: 6 μM; Leu: 16 μM; DTX: 4 μM) and FOLFOX (5-FU: 200 μM, OXA: 6 μM; Leu: 14 μM) for the indicated times. ²⁷ Tumor organoid morphology was observed and captured at day 1, 3, 5 and 7 by light microscopy. Images were used to measure the diameter of tumor organoids with ImageJ. To measure cytotoxicity, a PI uptake assay was performed as previously described. ²⁸ For measuring how chemotherapy affects the surface expression levels of CD133, LGR5, MHC-I and PD-L1, tumor organoids were incubated with FLOT and FOLFOX for 48 h. After treatment, the tumor organoids were dissociated using TrypLE Express enzyme (ThermoFisher Scientific) and incubated with the following antibodies: FITC-conjugated CD133 antibody (BioLegend), PE-conjugated LGR5 antibody (BioLegend), Alexa Fluor 488 conjugated PD-L1 (extracellular domain specific) antibody (Cell Signaling), and Alexa Fluor 488-conjugated MHC-I antibody (Santa Cruz) respectively for 1 h at 4°C. Then, the dissociated cells were washed with PBS. The intensities of CD133, LGR5, PD-L1 and MHC-I signals were measured by flow cytometry (Beckman). To measure the levels of cadherin-17, dissociated tumor cells were incubated with FITC-conjugated cadherin-17 antibody (Novus Biologicals) for 1 h at 4°C. Then, the dissociated cells were washed with PBS. The intensity of cadherin-17 was measured by flow cytometry (Beckman).

Statistical analysis

All experiments were independently repeated at least three times, and comparable results were obtained in each experiment. Data are expressed as mean ± standard deviation. P-values were calculated with Student’s t-test (paired two-tailed). P < 0.05 was considered significant.

Results

Establishment of patient-derived tumor organoids

To assess spatial tumoral heterogeneity, samples were taken from different regions of the same tumor (patient 11). Two samples were taken from the peritumor region (region 1 and region 3). One sample was taken from center region (region 2) (Figure 1(a)). Tumor samples were minced with a scalpel and enzymatically digested. Then, the viability of dissociated cells obtained from peritumor region and central region of the tumor using an image-based cell counter and trypan blue exclusion assay. Intriguingly, tumor cells obtained from peritumor region (region 1 and region 3) exhibited higher viability and larger sizes than those obtained from central region of tumor (Figure 1(b)). In addition, tumor organoids were readily cultured from regions 1 and 3, whereas no tumor organoids grew from region 2 (Figure 1(c) and (d)). The tumor core often displays distinct features, such as hypoxia, nutrient deprivation and extremely high cell density. ²⁹ Thus, we suspected that tumor cells within the central region (region 2) might be mostly dead due to high cell density and poor nutrient availability. To test this idea, we compared tumor tissues from adjacent peripheral and central tumor regions (region 1 and region 2) by histopathological analysis (Figure 1(e)). As expected, a high cell density was observed in the central region of tumor, yet there were few dead cells present. In contrast, the cell density in the peritumor region was relatively low. These results demonstrate that tumor heterogeneity will affect the ability to establish organoid cultures. In addition to comparing samples from the same tumor, we also established tumor organoids from tissues taken from different patients. Patient characteristics are shown in Table 1. The morphological features of the tumor organoids were observed by light microscopy and H&E staining (Figure 1(f)). We found that tumor organoids exhibited cystic and balloon-like morphology. In addition, all of the tumor organoids were positive for cadherin-17 (gastric cancer marker) expression (Figure 1(f)), in line with the gastric tumor origin.

Figure 1.

Establishment of patient-derived tumor organoids. (a) Samples were taken from different tumor regions for organoid preparation. (b) Tumor cell viability from different regions. (c-d) Tumor organoid growth from different sampled regions. Region 1: Day 1 vs Day 7: P = 0.057, Day 1 vs Day 13: P = 0.037 (one sample t test). Region 3: Day 1 vs Day 7: P = 0.020, Day 1 vs Day 13: P = 0.005 (one sample t test). Scale bar: 60 μm. (e) Histopathology of peritumoral (region 1) and central (region 2) tumor samples (n = 1). Scale bar: 50 μm. (f) Morphology and gastric cancer marker (Cadh-17) expression level for different tumor organoids. Data are expressed as mean ± standard deviation. *: P < 0.05.

Table 1.

Patient characteristics.

​	Age	Gender	Histology	Types	TNM stage	MMR
Patient 8	46	Female	Adenocarcinoma	Diffuse	Stage IA, pT1aN0	Intact
Patient 9	75	Female	Adenocarcinoma	Intestinal	Stage IA, pT1bN0	PMS2 loss
Patient 11	71	Male	Adenocarcinoma	Diffuse	Stage IVA, pT2N3	Intact
Patient 12	83	Male	Adenocarcinoma	Intestinal	Stage IA, pT1aN0	PMS, MLH1 loss
Patient 13	73	Male	Adenocarcinoma	Diffuse	Stage IIIA, pT4aN2	Intact

Chemotherapy regimens are efficacious in tumor organoids

Tumor organoids derived from different patients (patients 8, 9, 11, 12, and 13) and from different regions of the same tumor (patient 11-1 and 11-3) were treated with FLOT (5-FU, OXA, Leu, and DTX) and FOLFOX (5-FU, OXA, and Leu). The tumor organoids were then observed at days 1, 3, 5 and 7 of treatment. We found that tumor organoids receiving control treatments continuously grew, appearing larger and opaquer over time. However, the organoids treated with FLOT and FOLFOX did not grow (Figure 2(a)). We also quantified the FLOT- and FOLFOX-mediated inhibitory effects on tumor organoids by measuring the organoid diameters (Figure 2(b) and (c)). In line with the qualitative observations, the diameters of control-treated tumor organoids gradually increased. However, the organoids treated with FLOT and FOLFOX showed no increase in diameter over time. We also noted that FLOT and FOLFOX exhibited similar inhibitory effects on tumor organoids, regardless of the organoid origin. This result suggested that intertumoral and spatial tumor heterogeneity did not greatly impact chemotherapy response. To validate that the chemotherapy treatments caused cytotoxicity, the PI uptake assay was performed on the tumor organoids (Figure 2(d)). The results showed that PI was readily taken up by cells in the FLOT- and FOLFOX-treated small tumor organoids, confirming that the cells were killed by chemotherapy.

Figure 2.

Chemotherapy response in tumor organoids. (a) Tumor organoids were treated with FLOT and FOLFOX for 7 days. Tumor organoids were observed and recorded at Days 1, 3, 5 and 7. Scale bar: 60 μm. (b–c) Measurement of diameters of tumor organoids. To measure the diameters of tumor organoids, we used ImageJ to measure the lengths of both ends of the organoids (c) and defined this measurement as the organoid diameter (n = 10 for each group). Scale bar: 60 μm. Ctrl: control. Patient 8-Day 7: Ctrl vs FLOT: P = 0.00000001, Ctrl vs FOLFOX: P = 0.00000006. Patient 9-Day 7: Ctrl vs FLOT: P = 0.000006, Ctrl vs FOLFOX: P = 0.000008. Patient 11-1-Day 7: Ctrl vs FLOT: P = 0.000000004, Ctrl vs FOLFOX: P = 0.000000005. Patient 11-3-Day 7: Ctrl vs FLOT: P = 0.000000005, Ctrl vs FOLFOX: P = 0.000001. Patient 12-Day 7: Ctrl vs FLOT: P = 0.000002, Ctrl vs FOLFOX: P = 0.000006. Patient 13-Day 7: Ctrl vs FLOT: P = 0.0000005, Ctrl vs FOLFOX: P = 0.0000004. Student’s t-test (paired two-tailed). *: P < 0.05. Data are expressed as mean ± standard deviation. (d) Tumor organoids were treated with FLOT and FOLFOX for 7 days. Tumor organoid viability was examined with a PI uptake assay. Scale bar: 60 μm. All experiments were performed in triplicate, each being repeated at least three times.

Chemotherapy differentially affects cancer stem cell markers in tumor organoids

Previous work has shown that chemotherapy can enrich cancer stem cells.^30–33 Thus, we wanted to test whether chemotherapy affects cancer stem cells in the tumor organoids by monitoring gastric cancer stem cell markers (LGR5 and CD133) with flow cytometry. We found that FLOT and FOLFOX had little to no effects on induced LGR5 expression in most tumor organoids, except for patient 8 organoids (Figure 3(a) and (b)). In the organoids derived from patient 8, FOLFOX elevated LGR5 expression. In contrast to the minimal effects on LGR5, expression of CD133 was increased in most organoids by FLOT treatment; the only exception was for patient 11-3 and patient 13 organoid (Figure 3(c) and (d)). FOLFOX only increased CD133 expression in patient 9 organoids. Our findings therefore suggest that different chemotherapy regimens might have different effects on cancer stem cells from different tumors or regions of the same tumor. Importantly, these differential effects on cancer stem cells suggest that chemotherapy regimens may lead to different outcomes in different patients. Furthermore, a chemotherapy regimen can have different effects on cancer stem cells derived from different regions of the same tumor (patient 11-1 and patient 11-3).

Figure 3.

Effects of chemotherapy on cancer stem cell markers. Tumor organoids were treated with FLOT and FOLFOX for 2 days. The levels of LGR5 (a–b) and CD133 (c–d) were measured by flow cytometry. Ctrl: Control. Ctrl vs FLOT: P = 0.137, Ctrl vs FOLFOX: P = 0.0006. LGR5-Patient 9: Ctrl vs FLOT: P = 0.010, Ctrl vs FOLFOX: P = 0.021. LGR5-Patient 11-1: Ctrl vs FLOT: P = 0.005, Ctrl vs FOLFOX: P = 0.016. LGR5-Patient 11-3: Ctrl vs FLOT: P = 0.039, Ctrl vs FOLFOX: P = 0.853. LGR5-Patient 12: Ctrl vs FLOT: P = 0.048, Ctrl vs FOLFOX: P = 0.0007. LGR5-Patient 13: Ctrl vs FLOT: P = 0.006, Ctrl vs FOLFOX: P = 0.040. CD133-Patient 8: Ctrl vs FLOT: P = 0.005, Ctrl vs FOLFOX: P = 0.399. CD133-Patient 9: Ctrl vs FLOT: P = 0.039, Ctrl vs FOLFOX: P = 0.048. CD133-Patient 11-1: Ctrl vs FLOT: P = 0.000002, Ctrl vs FOLFOX: P = 0.623. CD133-Patient 11-3: Ctrl vs FLOT: P = 0.540, Ctrl vs FOLFOX: P = 0.230. CD133-Patient 12: Ctrl vs FLOT: P = 0.003 Ctrl vs FOLFOX: P = 0.012. CD133-Patient 13: Ctrl vs FLOT: P = 0.187, Ctrl vs FOLFOX: P = 0.026. Student’s t-test (paired two-tailed). Data are expressed as mean ± standard deviation. *: P < 0.05. Heatmap showing the expression (mean) of LGR5 (b) and CD133 (d). All experiments were performed in triplicate, each being repeated at least three times.

Chemotherapy differentially affects immune checkpoint molecule expression in tumor organoids

Expression of cancer immune checkpoint molecules (PD-L1 and MHC-I) is essential for therapeutic efficacy of immune checkpoint inhibitors,^{19,20,22–24} and it is known that expression of PD-L1 and MHC-I can be altered by chemotherapy.^34,35 Thus, it is essential to determine the expression levels of cancer immune checkpoint molecules after treatment in order to understand the immunogenicity status in recurrent tumors. We found that FLOT had little to no effects on surface MHC-I expression in any tested organoid. However, FOLFOX increased the levels of surface MHC-I in most tumor organoids, with the exception of patient 11-3, 12 and 13 organoids (Figure 4(a) and (b)). In contrast, FLOT had diverse effects on surface PD-L1, causing either an increase, decrease or no change in PD-L1 surface expression for different tumor organoids. FOLFOX had no measurable effects on surface PD-L1 in tumor organoids derived from patients 8, 11-1, 12, and 13 while the treatment increased surface PD-L1 in patient 9 and 11-3 organoids (Figure 4(c) and (d)). Intriguingly, these observations were somewhat inconsistent with the conclusions of several previous studies,^{19,20,22–24} as we found that chemotherapy does not consistently increase MHC-I and PD-L1 surface expression levels in tumor organoids. Furthermore, we found that FLOT and FOLFOX can sometimes reduce surface MHC-I and PD-L1 in patient-derived tumor organoids.

Figure 4.

Effects of chemotherapy on cancer immune checkpoint molecules. Tumor organoids were treated with FLOT and FOLFOX for 2 days. The levels of surface MHC-I (a–b) and PD-L1 (c–d) were measured by flow cytometry. Ctrl: Control. MHC-I-Patient 8: Ctrl vs FLOT: P = 0.149, Ctrl vs FOLFOX: P = 0.010. MHC-I-Patient 9: Ctrl vs FLOT: P = 0.035, Ctrl vs FOLFOX: P = 0.001. MHC-1-Patient 11-1: Ctrl vs FLOT: P = 0.061, Ctrl vs FOLFOX: P = 0.010. MHC-I-Patient 11-3: Ctrl vs FLOT: P = 0.012, Ctrl vs FOLFOX: P = 0.001. MHC-I-Patient 12: Ctrl vs FLOT: P = 0.346, Ctrl vs FOLFOX: P = 0.604. MHC-I-Patient 13: Ctrl vs FLOT: P = 0.003, Ctrl vs FOLFOX: P = 0.006. PD-L1-Patient 8: Ctrl vs FLOT: P = 0.966, Ctrl vs FOLFOX: P = 0.932. PD-L1-Patient 9: Ctrl vs FLOT: P = 0.000006, Ctrl vs FOLFOX: P = 0.003. PD-L1-Patient 11-1: Ctrl vs FLOT: P = 0.330, Ctrl vs FOLFOX: P = 0.076. PD-L1-Patient 11-3: Ctrl vs FLOT: P = 0.023, Ctrl vs FOLFOX: P = 0.0003. PD-L1-Patient 12: Ctrl vs FLOT: P = 0.026, Ctrl vs FOLFOX: P = 0.894. PD-L1-Patient 13: Ctrl vs FLOT: P = 0.014, Ctrl vs FOLFOX: P = 0.161. Data are expressed as mean ± standard deviation. *: P < 0.05. Heatmap showing the expression (mean) of surface MHC-I (b) and PD-L1 (d). All experiments were performed in triplicate, each being repeated at least three times.

Discussion

In this study, we established a panel of gastric cancer patient-derived tumor organoids to study chemotherapy-induced changes in cancer stem cell and cancer immune checkpoint molecules in the context of tumor heterogeneity.

When culturing tumor organoids from different gastric tumors, we found that tumor organoids could not be successfully cultured from the central region of the tumor in patient 11 (Figure 1). Notably, histopathological analysis suggested that most of cells in this region were still alive. However, our cell viability analysis revealed that roughly half of the cells in the central region of tumor were dead (Figure 1(b)). We therefore speculate that the tumor cells in the central region might be more vulnerable to tissue dissociation. This vulnerability could explain why tumors organoids could not be successfully cultured from these cells (Figure 1(c) and (d)). Another possible explanation is related to the distinct spatial distribution of cancer stem cells.^36,37 It is well-established that tumor organoids originate from cancer stem cells, ²⁵ and cancer stem cells are not evenly distributed in a tumor.^36,37 Therefore, it is possible that cancer stem cells were not present or rare in the region we sampled, which could explain why the tumor organoids were not successfully cultured from the central region of tumor.

Several previous studies have shown that tumor heterogeneity plays a crucial role in therapeutic failure and drug resistance.^4,38,39 However, we found that both FLOT and FOLFOX effectively suppressed tumor organoid growth, regardless of the origin (Figure 2). This consistent organoid suppression suggests that intertumoral and intratumoral heterogeneity had no major impact on chemotherapy efficacy. While these findings may appear to conflict with current understanding of cancer stem cells and drug resistance, they actually are consistent. Most cancer patients exhibit a good initial response to first-line chemotherapy. ⁴⁰ However, the problem is patient relapse that occurs in the months after treatment. ⁴⁰ In our experimental design, the tumor organoids received a first exposure to FLOT and FOLFOX, which is likely to reflect the “initial response” of the patient. Thus, tumor heterogeneity may have relatively little impact on initial chemotherapy response, similar to clinical experience.

Although the initial response to chemotherapy is often effective shrinkage of tumors, cancer stem cells may be enriched by the chemotherapy, leading to tumor recurrence and acquired chemotherapy resistance.^30–33 Previous studies have suggested that chemotherapy may enrich cancer stem cell populations through several molecular mechanisms, including activation of nuclear factor kappa B (NFκB) signaling, Wnt signaling, hypoxia inducible factor 1 α (HIF-1α) pathways, and the DNA damage checkpoint response.^41–43 Among these mechanisms, NFκB signaling has been the most extensively studied. Activation of NFκB has been shown to enhance the expression of stemness-related factors such as Nanog and Sox2, thereby promoting cancer stem cell maintenance and self-renewal. ⁴³ In addition, NFκB upregulates the expression of pro-tumorigenic cytokines, including interleukin-6, which contribute to cancer cell transformation and stem-like phenotypes. ⁴³ Moreover, NFκB has been reported to cooperate with Wnt and HIF-1α signaling pathways in driving cancer cell transformation and cancer stem cell enrichment following chemotherapy.^44,45 This phenomenon has been observed in the context of gastric cancer. For instance, it was shown that certain subsets of gastric cancer cells exhibiting cancer stem-like cell properties are increased after 5-FU treatment. ³² Moreover, upregulation of gastric cancer stem cell markers (CD133 and LGR5) is correlated with chemoresistance, metastasis and tumor relapse.^46–48 Notably, different subsets of cancer stem cells will exhibit different phenotypes and tumorigenic potentials.^49,50 In line with this heterogeneity of cancer stem cells, we found that chemotherapy had diverse effects on two cancer stem cell markers in tumor organoids (Figure 3). For instance, FOLFOX increased LGR5 and decreased CD133 in patient 8 organoids. Meanwhile, both FLOT and FOLFOX reduced LGR5 but increased CD133 in patient 9 organoids. Other patient organoids exhibited similarly unique responses to chemotherapy. Taken together, these results demonstrated that different chemotherapy regimens have different effects on enrichment of different subsets of cancer stem cells. In addition to the differences between tumors, we also found spatial heterogeneity in chemotherapy-mediated cancer stem cell enrichment. Tumor organoids established from different regions of the same tumor had different responses to the same chemotherapy regimen. In particular, exposure of patient 11-1 and 11-3 organoids to chemotherapy caused different effects on cancer stem cells, with FLOT increasing CD133 in patient 11-1 organoids but having no effect on CD133 in patient 11-3 organoids.

The expression levels of MHC-I and PD-L1 are indicators of anti-PD-1/anti-PD-L1 therapy efficacy.^19–24 MHC-I downregulation is associated with failure of anti-PD-1/anti-PD-L1 therapy,^23,24 and higher PD-L1 expression is correlated with better response to PD-1/PD-L1 inhibitors.^19–21 Furthermore, several studies have demonstrated that chemotherapy upregulates PD-L1 and MHC-I on cancer cells, which increases sensitivity to PD-1/PD-L1 inhibitors.^23,24,34,35 NFκB is regarded as one of the major mediators of this process.^51,52 Upon exposure to chemotherapeutic agents, activation of the NFκB pathway enhances the transcription of PD-L1 and MHC-I, resulting in their upregulation.^51–53 Furthermore, chemotherapy may increase PD-L1 expression through post-translational mechanisms that enhance the stability of PD-L1 on the cell surface. Notably, chemotherapy-induced upregulation of CKLF-like MARVEL transmembrane domain-containing 6 (CMTM6), ⁵⁴ a key stabilizer of membrane-bound PD-L1, may further contribute to the sustained surface expression of PD-L1 by protecting it from lysosome-mediated degradation. ⁵⁵ In contrast to these studies, our results show diverse effects of chemotherapy on MHC-I and PD-L1 expression levels (Figure 4). The same chemotherapy regimen could either upregulate or downregulate MHC-I and PD-L1 in different tumor organoids, suggesting that intertumoral heterogeneity influences chemotherapy-mediated PD-L1 and MHC-I modulation. In addition, different chemotherapy regimens had diverse effects on MHC-1 and PD-L1. For instance, in patient 8 organoids, FLOT had no effect on MHC-I and PD-L1 levels, while FOLFOX increased MHC-I but not PD-L1 (Figure 4(a) and (c)). Intriguingly, the same chemotherapy regimens applied to organoids derived from different regions of the same tumor had different effects on PD-L1. FOLOFX had no effect on PD-L1 in patient 11-1 organoids, but it increased PD-L1 in patient 11-3 organoids. Thus, our results suggest that chemotherapy-mediated modulation of PD-L1 and MHC-I is heavily influenced by spatial heterogeneity within the same tumor and heterogeneity between tumors.

Increasing evidence suggests that spatial heterogeneity is a key feature of cancer stem cells,^36,37 as well as PD-L1^22,56 and MHC-I ⁵⁶ expression. These factors are each known to influence cancer treatment responses. In addition to spatial heterogeneity, temporal heterogeneity in cancer stem cell populations and cancer immune checkpoint molecules may also occur. An especially remarkable example of this concept is that chemotherapy induces or enriches cancer stem cells in many contexts.^12,32,33 In NSCLC, PD-L1 expression levels in first biopsy samples, surgical specimens, and re-biopsy (relapsed tumor sample) differ, indicating that PD-L1 expression is dynamic over the course of disease. ²² It was also shown that in gastroesophageal adenocarcinoma, PD-L1 expression in primary tumors and metastatic tumors are different. ⁵⁶ There was only a 63% concordance of PD-L1 expression levels when comparing pretreatment primary and post-treatment tumor samples, indicating major temporal heterogeneity. ⁵⁶ Thus, it is likely that spatial and temporal heterogeneity in tumors may account for differential patient responses and outcomes.

In this study, we analyzed the effects of spatial and temporal tumoral heterogeneity on cancer stem cell populations and cancer immune checkpoint molecules using patient-derived organoids. Evaluating the impact of chemotherapy on these parameters may help predict treatment responses and potential changes in tumor characteristics following therapy. However, obtaining all of these parameters in a clinical setting is often challenging. For example, there are limited direct clinical data available to accurately assess the effects of chemotherapy on the expression of cancer immune checkpoint molecules in individual patients. Nevertheless, numerous studies have demonstrated that chemotherapy can modulate the expression of these molecules.^57,58 This suggests that the baseline expression of cancer immune checkpoint molecules measured before treatment may not accurately reflect their post-treatment status. In this context, patient-derived organoids may provide a valuable platform for simulating patient-specific therapeutic exposures and evaluating treatment-induced changes in cancer immune checkpoint molecule expression. Such information could help clinicians determine whether a patient is likely to benefit from PD-1/PD-L1 inhibitor therapy following chemotherapy. In addition, if chemotherapy preferentially enriches specific cancer stem cell populations, organoid-based testing may identify these changes and facilitate the selection of combination therapeutic strategies targeting both bulk tumor cells and the enriched cancer stem cell populations. Such an approach may reduce the likelihood of tumor recurrence and improve long-term treatment outcomes. Therefore, patient-derived organoids may serve as a clinically relevant functional platform for predicting therapy-induced molecular alterations, optimizing treatment selection, and guiding personalized therapeutic strategies for individual patients. However, limitations in the success rate of tumor organoid culture and the availability of suitable patient samples meant that our findings were primarily based on a small number of patients. Therefore, future research with larger sample sizes is necessary to validate these results.

Conclusions

Our findings demonstrate that both spatial and temporal tumoral heterogeneity exist in cancer stem cell populations and cancer immune checkpoint molecules in gastric cancer. Thus, multi-site and multi-time-point sampling of tumors may be necessary to better predict and evaluate the impacts of therapy. Since real-time monitoring of cancer stem cells and cancer immune checkpoint molecules may be difficult in clinical practice, tumor organoids can serve as an ex vivo tool to simulate the treatment responses and monitor alterations of tumor characteristics induced by therapy. Furthermore, our conclusions are based on observations from a relatively small sample size. Larger-scale studies are still needed.

ORCID iD

Bor-Chyuan Su https://orcid.org/0000-0002-0454-0624

Ethical considerations

This study complies with the ethical guidelines of the 1975 Helsinki Declaration (modified in 2008). The study protocol was approved by the Taipei Medical University-Joint Institutional Review Board (Approval No: N202405105).

Consent to participate

Informed consent was performed prior to starting of the study. Confidentiality of patient information was strictly maintained in accordance with the Declaration of Helsinki. This study does not include identifiable patient data or images requiring publication consent. No conflicts of interest/funding are reported.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author, Y.H.H., upon reasonable request.*