Distinct but interchangeable subpopulations of colorectal cancer cells with different growth fates and drug sensitivity

Summary Dynamic changes in cell properties lead to intratumor heterogeneity; however, the mechanisms of nongenetic cellular plasticity remain elusive. When the fate of each cell from colorectal cancer organoids was tracked through a clonogenic growth assay, the cells showed a wide range of growth ability even within the clonal organoids, consisting of distinct subpopulations; the cells generating large spheroids and the cells generating small spheroids. The cells from the small spheroids generated only small spheroids (S-pattern), while the cells from the large spheroids generated both small and large spheroids (D-pattern), both of which were tumorigenic. Transition from the S-pattern to the D-pattern occurred by various extrinsic triggers, in which Notch signaling and Musashi-1 played a key role. The S-pattern spheroids were resistant to chemotherapy and transited to the D-pattern upon drug treatment through Notch signaling. As the transition is linked to the drug resistance, it can be a therapeutic target.


INTRODUCTION
Cancer is characterized by extensive intratumor heterogeneity (ITH). 1,2 It has been increasingly recognized that ITH contributes to drug resistance and cancer recurrence following therapy. 3 ITH is known to be caused by a variety of genetic mutations, microenvironmental conditions, and cell-intrinsic plasticity. 2,4,5 Genetic mutations-driven ITH has been systematically studied, 6 whereas the ITH-related nongenetic processes remain largely elusive and are currently receiving intense research attention. [7][8][9] Cancer stem-like cells (CSCs) are one of the theories that attempt to explain the nongenetic heterogeneity of cancer. 5,10 Recently, CSC models have been revisited with the evidence that cancer cells can dynamically fluctuate from a non-stem cell-like state to a stem cell-like state. 10 For this reason, the importance of the heterogeneity and plasticity within cancer cell subpopulation is highlighted. 5,8 Colorectal cancer (CRC) is one of the leading causes of cancer-related death worldwide. 11 Accumulating evidence suggests that CRC cells represent phenotypically dynamic (rather than static), heterogeneous cell populations that display cell plasticity characteristics. [12][13][14][15][16] Recently, 3D cell culture systems utilizing patient-derived tumors have been developed for various cancer types, including CRC. 17 Herein, we used the cancer tissue-originated spheroid (CTOS) method developed by us, in which the cell-cell contact is maintained throughout the organoid preparation, culture, and passaging. 18 The growth of each CRC organoid within the same line is quite heterogeneous, 19 suggesting that CRC organoids prepared by CTOS method retain heterogeneous populations of cells. Furthermore, we demonstrated that a small subset of cells within the CRC organoids can initiate regrowth after exposure to high-dose radiation, an observation that was nongenetically and reversibly determined. 20 As current cancer therapies have been designed and developed mainly against the fast-growing cancer cells, the importance of the quiescent or slow-growing cancer cells has been overlooked. Consequently, these slow-growing cancer cells may survive the anticancer treatment, revert to the fast-growing cancer cells, and serve as a reservoir for tumor regrowth. 13,[21][22][23] To characterize the dynamics of quiescent or slow-growing cancer cells, an analysis at the single-cell resolution is required because the nature of the slow-growing cells is usually masked by that of the fast-growing cells. Recently, single-cell transcriptome ll OPEN ACCESS analyses were recruited and served the study of the characteristics of the slow-growing CSCs. 24 However, such a ''snapshot'' analysis involving a procedure for isolating CSCs with the use of markers 24 or dye retention 25 exerts limitations when applied to a continuously changing process. Therefore, the employment of a phenotypically trackable cell culture system allowing for a single-cell resolution is necessary.
In this study, we modified the conventional spheroid-forming assay, with which we tracked the cell fate of forming spheroids as well as the growth at a single-cell resolution and revealed the existence of heterogeneous subpopulations in CRC organoids with distinct growth patterns. We characterized the distinct but interchangeable subpopulations, and revealed the molecular mechanisms regulating the transition.

Heterogeneous growth ability of the cells in CRC organoids at a single-cell level
To precisely track the cell fate of the growth in CRC organoids at a single-cell resolution, we modified the conventional spheroid-forming assay and developed a single-cell-derived spheroid-forming and growth (SSFG) assay, which includes the undertaking of (i) an initial confirmation of the strict single-cell status within a well, (ii) culture under growth-permissive conditions, and (iii) a time-course growth assessment of each well ( Figure 1A). We applied the SSFG assay to a CRC organoid line, C45, which has wide range of spheroid growth abilities. 19 We excluded non-single cells and large cells which can be doublets or the cells just before cell division at the very beginning of the assay ( Figure 1B), as well as the non-growing cells ( Figures 1C and 1D) that presented with several patterns: early death (Figure S1A-a), late death ( Figures S1A-b ), growth arrest ( Figures S1A-c), and a decline in spheroid size ( Figures S1A-d). The maximum area of the non-growing spheroids was below 2.5 3 10 3 mm 2 . The spheroid-forming capacity was, on average, 59%. The growth range of the single-cell-derived growing spheroids in the C45 line was 228-fold ( Figures 1C-1F). We measured the growth variation of the spheroids derived from single cells through the SSFG assay in 13 additional lines of CRC organoids from different patient tumors (Table S1, Figures 1G-1I, and S1B-S1K). The spheroid-forming capacities were between 19% and 59% (Table S2). In all 14 lines (including C45), the sizes of the spheroids within each line varied substantially, and two of the lines (C120 and C132) demonstrated a statistically significant bimodal distribution. The mutational profile (Table S3) showed no clear correlation with the spheroid-forming capacity, the growth, or the growth range. These results indicate that each spheroid-forming cell in the CRC organoids has diversity of the growth ability as a single cell.
Each clone from the CRC organoids generated distinct subpopulations with different growth capacity To characterize the heterogeneous capacity of spheroid formation and growth of the single cells in more detail, we selected several individual spheroids at the end of the first round of SSFG assay using C45 line ( Figure 2A). The growth curves and the images of each clone in the first round SSFG assay are shown in Figures S2A and S2B. As each spheroid was strictly derived from a single cell, it can be called a clone. We named the clones as C45-1 and C45-2 from the small spheroids and two clones as C45-3 and C45-4 from the large spheroids. To characterize each clone, we first evaluated its proliferation capacity and viability. To evaluate proliferation, we immunostained PCNA, a proliferating cell nuclear antigen. The PCNA-positive rates were higher in the large spheroids (C45-3 and C45-4) compared to the small spheroids (C45-1 and C45-2) ( Figures S3A and S3B), suggesting a higher proliferative state of the cells in the large spheroid-forming clones during in vitro culture. To evaluate cell death, we performed propidium iodide (PI) staining, which showed no differences in viability between the C45 clones during culture ( Figures S3C and S3D). Consistent with these results, all the C45 clones had similar spheroid-forming capacity ( Figure 2B), suggesting that the differing growth capacity was due to the differences in the cell proliferation rather than survival. Next, we expanded each clone in vitro and performed the second round of SSFG assays. The spheroid-forming capacity was preserved in all clones ( Figure 2B). The cells derived from the small spheroids gave rise to only small spheroids (S (small)-pattern), while the cells derived from the large spheroids gave rise to both small and large spheroids (D (dual)-pattern) ( Figure 2C). We then set the putative threshold between the two phenotypes at 1.0 3 10 5 mm 2 , based on the rounded value of the maximum size of the C45-1 spheroids at day 13. To further investigate the stability of the growth features, we performed additional rounds of the SSFG assay. In all four C45 clones, the growth pattern of single cells was preserved during all three rounds of the SSFG assay ( Figures 2D-2G iScience Article types of cells; a cell which has the capacity of generating a small spheroid, an S-cell, and generating a large spheroid, an L-cell. During spheroid formation, an S-cell only gave rise to a small spheroid consisting of pure S-cells, while an L-cell gave rise to a large spheroid consisting of both the S-and the L-cells ( Figure 2J).
Establishment of an S-cell clone from the small spheroids derived from an L-cell clone Next, we investigated the fate of the cells in the small spheroids of the L-cell clone, C45-4. From the first round of SSFG assay, we collected small and large spheroids, respectively ( Figure 3A). We then repeated the SSFG assay followed by the collection three times ( Figure 3A). The D-pattern was stably preserved in the large-spheroids pool. However, the small-spheroids pool exhibited the D-pattern in the first round, and then the ratio of the large spheroids decreased at the second round and showed the S-pattern in the third round, the phenotype of which was stable in the subsequent experiments. We then named the subclones of C45-4 after three rounds of the SSFG assay derived from the small spheroids and the large spheroids as C45-4S and C45-4L, respectively. The spheroid-forming capacity of both subgroups did not differ from that of the parent C45-4 (see Figure 2C and the control groups in the following results of the SSFG assay). The disappearance of the L-cells in the small-spheroids pool during the multiple rounds of SSFG assay might be attributable to an epigenetic event, while it can also be caused by the elimination of the contaminated L-cells in the small spheroids. In any way, stable phenotype of the S-cells could be generated from the L-cells. The ratio of PCNA-positive cells was higher in C45-4L spheroids compared to the C45-4S spheroids, suggesting more proliferation of the cells in the large spheroid subclone (C45-4L) ( Figures S4A and  S4B). In contrast, there were no differences in PI staining between the C45-4L and C45-4S subclones ( Figures S4C and S4D). Since the C45-4 spheroids were derived from a single clone, the growth pattern of single cells is likely to be regulated by nongenetic mechanisms.

The L-cells can be generated from the S-cells in vivo
We next examined the tumorigenicity of the clones. We subcutaneously injected the spheroids of the S-cell clones, C45-1 and C45-4S, and the L-cell clones, C45-4 and C45-4L, into immunodeficient mice. All spheroids were tumorigenic ( Figures 3B and 3C). Although the S-clones demonstrated a latency period of tumor growth, which was not observed in the L-clones, S-clones' growth rate eventually caught up. To confirm initially that S-cells were capable of accelerating the cell proliferation under in vivo conditions, we examined the tumor xenografts at two different time points: a ''midpoint'', when the tumor volume was detectable for the first time, and an ''endpoint'', when the tumor volume reached $200 mm 3 . PCNA-positive rate increased from 30% in vitro to an average of 50%-60% in the xenograft tumors at the midpoint and endpoint, reaching the levels of D-pattern spheroids ( Figure S5). This suggests that S-cells are capable of restarting the cell cycle in vivo. The change in the proliferation rate was already observed at the beginning of the tumor growth, indicating that the transition of S-cells to a D-pattern phenotype occurred during the latency period.
In contrast, the PCNA-positive rates of D-pattern spheroid-derived xenograft tumors were the same in the in vitro and in vivo conditions ( Figure S5). These results suggested that S-cells and L-cells have different dependencies on the tumor microenvironment for regulating proliferation. We then prepared organoids from the xenografts and subjected them to the SSFG assay. The spheroid-forming capacity was preserved after the xenograft formation ( Figure 3D). The D-pattern was preserved in the cells of C45-4L spheroids ( Figure 3E). Interestingly, the S-cell clones, C45-1 and C45-4S, acquired the D-pattern phenotype following the xenograft formation ( Figure 3E). These results indicate that the L-cells can be generated from the S-cells after interacting with the tumor microenvironment, although we could not exclude the possibility that the phenotype change was attributable to the contamination of the small number of the L-cells in the S-cell clones, even if they were undetectable through the multiple rounds of the SSFG assay in vitro. (I) Violin plots of the SSFG assay for the indicated C132 clones.
(J) Schematic overview of the growth characteristics of a single cell. A circle represents a cell; green, S-cells; orange, L-cells. Note that the component of the spheroids is assessable only after the next round SSFG assay. In Figure 2, the data of the SSFG assay were tested by the Mann-Whitney U test. ***, p < 0.001; and ****, p < 0.0001. We examined the differences in the molecular features between the spheroids with the different growth patterns. Direct comparison of the S-cells and the L-cells is impossible because the growth capacity is assessed only after the SSFG assay. Taking advantage of the finding that the spheroids derived from the S-cell clones consisted of the pure S-cells, we compared the S-pattern spheroids derived from the S-cell clones with the D-pattern spheroids derived from the L-cell clones, which consisted of the mixture of the S-cells and the L-cells. As expected from the slow growing feature, the S-pattern spheroids showed less ERK activity than those of the D-pattern spheroids ( Figure 4A).
To shed more light on the molecular characteristics, we analyzed the differentially expressed genes between the spheroids with different growth patterns derived from a clone; C45-1, C45-4S, and C45-4L. The single cells from each subclone were cultured for 7 days and were subsequently subjected to microarray analyses. Of the 29,596 genes examined, 408 genes were found to be upregulated more than 1.5-fold, and 620 genes were downregulated less than 0.67-fold when the three aforementioned subclones were compared ( Figure 4B and Table S4). The volcano plot analyses revealed the similarity of their gene expression profiles and the existence of some differentially expressed genes ( Figures 4C-4E). Surprisingly, among these differentially expressed genes, many have been previously reported as stem cell markers of CRC. 10,15,26 The levels of MSI1, MEX3A, SOX4, EPHA4, and LRIG1 were higher in the C45-4L spheroids; the level of PTPPRO was higher in the C45-1 spheroids; and the levels of LGR5, PROM1 (CD133), and RGMB were higher in the C45-4S spheroids. We could confirm the expression patterns of LGR5, PROM1, and MSI1 by a semi-quantitative RT-PCR ( Figure 4F). The results suggested that the reported CSC genes were differentially expressed among these clones. Gene set enrichment analysis (GSEA) revealed that the MYC signature 27 was significantly enriched in the C45-4L spheroids compared with the C45-1 spheroids ( Figure 4G), thereby supporting the growth difference as MYC lies at the crossroads of many growth-promoting signal transduction pathways. 28 Additionally, GSEA revealed that the Notch pathway activation (PID_NOTCH_PATHWAY) 29 was enriched in C45-4L spheroids when compared to the C45-4S spheroids ( Figure 4H), suggesting the role of Notch signaling in regulating the D-pattern phenotype. Indeed, the protein levels of the Notch intracellular domain (NICD) were higher in the D-pattern spheroids (C45-4L) when compared with the S-pattern spheroids (C45-4S) ( Figure 4I). Moreover, we have analyzed the expression of HES1, a downstream target of Notch signaling by in situ hybridization (RNA-Scope) ( Figures 4J and 4K). The expression of HES1 mRNA in the D-pattern spheroids was higher than that in the S-pattern spheroids. Interestingly, the HES1-positive cells were not homogeneously distributed but were scattered in some small areas. Taken together, the activation of the Notch pathway was higher in the D-pattern spheroids than the S-pattern spheroids.

Transition between the growth patterns was regulated by cell-cell contact through notch signaling
To further reveal the transition, we conducted well-controlled experiments in vitro. Since the L-cells naturally generate the S-cells during spheroid formation while the S-cells did not generate the L-cells when isolated, we speculated that the transition mechanisms from the S-to the L-cells might depend on a cell-cell interaction, especially Notch signaling. Thus, we generated chimeric spheroids by aggregating the EGFPlabeled C45-4S cells (C45-4S-EGFP) and the mCherry-labeled C45-4L cells (C45-4L-mCherry) (Figures 5A and 5B) and subjecting them to the SSFG assay. The C45-4S-EGFP cells acquired the D-pattern phenotype within the chimeric spheroids, similar to the C45-4L-mCherry cells ( Figures 5C and 5D). In contrast, the S-pattern was maintained in the C45-4S-EGFP cells when both the types of cells were co-cultured in different gel droplets within the same wells ( Figures 5E and 5F), indicating that a close cell-cell interaction was necessary for this transition. To further investigate the molecular mechanisms of this cell-cell interaction, we examined Notch signaling, which was enriched in C45-4L spheroids ( Figures 4H and 4I) and reportedly played an important role in determining the cell fate in the context of cell-cell contact. 30, 31 We (B and C) Growth curves of xenograft tumors originating from C45-1 and -4 (3B) and C45-4L and -4S (3C) spheroids. The mean G SD is shown; n, the number of animals in each group; statistical analyses were performed by two-way ANOVA, followed by Bonferroni's test. N, number of mice.
(D and E) Spheroid-forming capacity (3D) and violin plots of the SSFG assay (3E) for the C45-1, C45-4S, and C45-4L subclones comparing control organoids maintained in vitro (Ctrl) with those prepared from the xenografts in (3B, 3C) (tumor-derived). In Figure 3, the data of the spheroid-forming capacity were tested by one-way ANOVA, followed by Tukey's test, and those of the SSFG assay were tested by the Mann-Whitney U test. Ns, not statistically significant. *, p < 0.05; **, p < 0.01; ***, p < 0.001; and ****, p < 0.0001. iScience Article inhibited Notch signaling with DAPT, a notch inhibitor, which was confirmed through Western blotting of NICD in the C45-4L-mCherry/C45-4S-EGFP chimeric spheroids ( Figure 5G). The D-pattern of both parent C45 and C45-4L did not change through DAPT treatment ( Figure S6A), indicating that Notch signaling does not affect the ability of L-cells to generate the D-pattern of spheroids or affect the growth of the spheroids. We, then, treated the C45-4L-EGFP and C45-4S-mCherry spheroids with DAPT and generated chimeric spheroids ( Figure S6B). The treatment of DAPT did not affect the spheroid-forming capacity in either clone ( Figure 5H); however, it inhibited the transition of the C45-4S-EGFP cells to the D-pattern within the chimeric spheroids ( Figure 5I). Thus, the Notch signaling is involved in the transition from the S-cells to the L-cells. Meanwhile, delayed treatment of DAPT ( Figure S6C) did not affect the transition ( Figure S6D), indicating that Notch signaling was critical for the transition of S-cells to L-cells at early time point of the cell-cell contact with the L-cells.

ll
To elucidate the role of Notch signaling on cell cycle reactivation in the S-cells-derived xenograft tumors, we evaluated the expression of HES1 in both mid-and end-point tumor xenografts as described above ( Figure S7). Consistent with the PCNA staining results ( Figure S5), the ratio of HES1-positive cells in the S-cells-derived xenograft tumors increased from approximately 2%-4% in the in vitro cultured S-pattern spheroids to an average of approximately 7%-10%, reaching the values of D-pattern spheroid-derived tumors ( Figure S7). These results indicated that S-cells can also activate Notch signaling in vivo. Interestingly, the ratio of HES1-positive cells was higher at the midpoint than the endpoint, decreasing to approximately 5% at the endpoint of the S-cells-derived tumors, suggesting that the activation of Notch signaling is involved in the cell fate transition in S-cells in vivo.

S-cells were the drug-resistant fraction
Next, to examine the drug sensitivity of the subpopulations with different growth pattern, we subjected them to the conventional sensitivity assay of organoids. 19 We tested 5-FU (a drug that is currently used in clinical practice) and PD0325901 (a MEK inhibitor, and a candidate drug for CRC treatment. 32 The S-pattern spheroids (C45-4S, C45-1, C45-2, C132-1, and KUC16-1) were significantly more resistant to both the drugs than the D-pattern spheroids (C45-4L, C45-3, C45-4, C132-2, and KUC16-2) ( Figures 6A  and S8), except for the C132 clones, which responded equally to PD0325901 ( Figure S8). When the clones were exposed to higher doses of each drug, a few small and intact spheroids remained, even in the sensitive C45-4L spheroids ( Figure 6B). We then performed the SSFG assays for the remaining small spheroids immediately after exposure to each drug. The MEKi treatment did not decrease the spheroid-forming capacities, and 5-FU made it approximately half, indicating that the remaining small spheroids retained substantial spheroid-forming capacity after drug exposure ( Figure 6C). The SSFG assay revealed that the C45-4L cells had the S-pattern of spheroid growth ( Figure 6D). After the removal of the drugs, the C45-4S clone grew in a similar manner to the non-treated spheroids ( Figure 6E). When Notch signaling was inhibited by DAPT at the drug withdrawal time, the regrowth was suppressed ( Figure 6E). The SSFG assay for the spheroids after drug withdrawal and regrowth revealed that the regrown spheroids of C45-4S switched to the D-pattern ( Figure 6F). DAPT treatment diminished the switching to the D-pattern after drug withdrawal and regrowth in C45-4S. ( Figure 6F). However, the spheroid-forming capacity did not change with the treatment of the drug or DAPT ( Figure 6G). Taken together, the S-cells were the drug-resistant fraction, and the transition of the isolated S-cells to the L-cells occurred during the regrowth after the drug withdrawal depending on the Notch signaling. To further explore the genes that characterize the different phenotypes of spheroid growth, we selected 148 differentially expressed genes with relatively high intensity by a clustering and a heatmap analyses of the microarray data comparing the C45-clones and subclones ( Figure S9A) (Table S5). Among the candidate genes, RNA-binding proteins, MSI1 and MEX3A, were included. We focused on MSI1 since it has been reported that the Notch signaling regulates MSI1 expression in metastatic CRC cells (Pastò et al., 2014). In addition, MSI1 is reportedly a CSC marker in the normal intestine and the CRC, 33,34 and is involved in the self-renewal of stem cells in both the normal intestine 35 and CRC. 36,37 The expression levels of MSI1 were confirmed to be higher in the D-pattern spheroids (C45-4L) than in the S-pattern spheroids (C45-1 and C45-4S) in parallel with those of MYC and NICD ( Figure 7A). This was also true in the subclones of other lines, such as C132 and the KUC16 ( Figure 7B). Moreover, DAPT treatment suppressed the expression of not only NICD but also MSI1 and MYC ( Figure 7C). To further investigate the functional role of MSI1 in the transition between the different growth patterns, we knocked out the MSI1 gene in the C45-4L spheroids by using the CRISPR/Cas9 system, generating the C45-4L_sgMSI1 spheroids ( Figure 7D). No difference was observed regarding the spheroid-forming capacity ( Figure 7E), while the ability of generating large spheroids was impaired ( Figure 7F), thereby suggesting that MSI1 is not involved in the growth of the S-cells but in the transition of the growth pattern. We examined whether the cell-cell contact with the cells from D-pattern spheroids can rescue the impaired transition of the C45-4L_sgMSI1 cells to the D-pattern, as shown in Figure 5C. For this reason, we generated chimeric spheroids by mixing RFP-labeled C45-4L_sgMSI1 cells and wild-type C45-4L cells and subjected them to the SSFG assay. The C45-4L_sgMSI1 cells could not demonstrate their D-pattern phenotype, even in the chimeric spheroids with wild-type cells ( Figure 7G), thereby suggesting that MSI1 is involved in the cell-cell contact-induced transition from the S-to the D-pattern of growth. We subsequently generated chimeric spheroids mixing EGFP-labeled C45-4S cells and RFP-labeled C45-4L_sgMSI1 cells, to check whether C45-4L_sgMSI1 cells retain the ability of parent C45-4L cells to switch the S-pattern spheroid (C45-4S) to the D-pattern in the chimera. Unlike the parent C45-4L, C45-4L_sgMSI1 cells failed to switch the S-pattern spheroids to the D-pattern ( Figure 7H). We then overexpressed MSI1 in C45-4S and generated the C45-4S_MSI1OE spheroids ( Figure 7I) and subjected to the SSFG assay. No difference was observed in their spheroid-forming capacity ( Figure 7J), whereas growth of the C45-4S_MSI1OE showed the D-pattern ( Figure 7K). Notably, not all the cells generated large spheroids. Similar results were obtained using another S-cell clone, KUC16-1 ( Figures S9B-S9D). Taken together, MSI1 is not simply involved in the growth of the spheroids but functionally in the transition of the growth pattern.
The expression levels of MSI1 in the spheroids were accompanied with the increased protein levels of NICD and MYC in the clones with different growth patterns ( Figures 7A and 7B), suggesting three molecules formed a network. Notch inhibition by DAPT in D-pattern spheroids (C45-4L) leads to a decrease of the MSI1 protein levels compared to the untreated control ( Figure 7C), suggesting Notch signaling regulates the MSI1 expression. On the other hand, MSI1 reportedly blocks the translation of NUMB, which promotes the degradation of NICD, 38 whereas neither the MSI1 knockout in D-pattern spheroids (C45-4L) nor the overexpression of MSI1 in S-pattern spheroids (C45-4S) affected the NICD protein levels ( Figures 7D  and 7I), suggesting MSI1 does not regulate Notch signaling in this experimental context. For the MYC  Figure 4, the data of the spheroid-forming capacity were tested by one-way ANOVA, followed by Tukey's test, and those of the SSFG assay were tested by the Mann-Whitney U test. *, p < 0.05. (G) Spheroid-forming capacity of C45-4S treated as indicated. In Figure 5, the data of the spheroid-forming capacity were tested by one-way ANOVA, followed by Tukey's test, and those of the SSFG assay were tested by the Mann-Whitney U test. Ns, not statistically significant. **, p < 0.01; ***, p < 0.001; and ****, p < 0.0001. iScience Article expression, the MSI1 knockout in C45-4L resulted in a decrease in the MYC protein levels ( Figure 7D), meanwhile the MSI1 overexpression in the C45-4S spheroids resulted in an increase of the MYC protein levels ( Figure 7I). Taken together, MSI may be located downstream to Notch and upstream to MYC.

DISCUSSION
Our experiments revealed that the patient-derived CRC organoids consist of phenotypically heterogeneous and interchangeable spheroid-forming cells with different cell fate of growth and drug sensitivity. At the second-round spheroid-forming assay, the cells in a small spheroid gave rise to only S-pattern, whereas the cells in a large spheroid gave rise to D-pattern. The transition from the S-to the D-pattern is molecularly regulated, and Notch signaling and MSI1 play a significant role in it. Our data provide new insights into the molecular mechanisms regulating cancer cell heterogeneity and plasticity.
In this study, we applied a method that allowed us to precisely track the capacity of not only spheroid formation but also the growth of each spheroid-forming cell in CRC organoids with a single-cell resolution: the SSFG assay. The spheroid-forming capacity was varied, but it was generally high in this study, ranging from 19% to 59% (Table S2). Interestingly, the spheroid-forming capacity was similar among the studied clones and subclones, a finding that supports the idea that the variation in spheroid growth is an event that occurs within cell populations that demonstrate spheroid-forming ability.
Herein, we demonstrated that the CRC cells consist of two distinct but interchangeable subpopulations. The juxtacrine interaction (or cell-cell interaction) with the D-pattern cells as well as the tumor microenvironment was found to be critical in order for the S-pattern cells to transit to the D-pattern phenotype. Our results have also suggested the existence of a gate 39,40 that regulates the transition from the S-to the D-pattern in CRC cells through a nongenetic process. In fact, Notch signaling and MSI1 seem to be the regulators of this transition.
Since all of the clones studied here were able to form spheroids at a high frequency and be tumorigenic, the existence of the subclones with different type of growth patterns and their interchangeable nature might be relevant to CSC, although the growth-permissive culturing conditions 18 in the SSFG assay in this study were different from those of earlier studies on CSCs from CRC. 41,42 Indeed, recent studies have revealed that stem-like properties of cancer cells can dynamically fluctuate. 5,10,15,40 In addition, many of the reported CSC markers in CRC were in the list of the differentially expressed genes between the clones with different type of growth patterns. Notably, the increase or decrease depended on each marker.
An overexpression of MSI1 has been reported in different tumor types, 43 including CRCs, 36 and MSI1 has been described as a CSC marker in CRC. 33,34 We, herein, demonstrated that the expression levels of MSI1 were lower in the S-pattern spheroids, and could functionally modulate the transition of the growth status in the studied spheroids ( Figure 7). Notably, the spheroid-forming capacity was affected by neither the gene knockout nor the overexpression of MSI1, thus indicating that MSI1 is not essential for the cells to be stemlike. In previous reports, the knockdown of MSI1 in CRC cells suppressed their capacity of spheroid formation and their tumorigenicity. 36,37,44 However, in these cases, the number of spheroids could have been underestimated because of the increased slow-growing populations, and the fact that the MSI1-downregulated cells in these studies did not actually lose their ability to form tumors, but they rather exhibited growth retardation. The molecular mechanisms by which the phenotypically different subpopulations (J and K) Spheroid-forming capacity (7J) and violin plot of the SSFG assay (7K) for the C45-4S subclones, comparing the control and the MSI1 overexpressing cells. In Figure 7, the data of the spheroid-forming capacity were tested by one-way ANOVA, followed by Tukey's test, and those of the SSFG assay were tested by the Mann-Whitney U test. Ns, not statistically significant. *, p < 0.05. **, p < 0.01. iScience Article are generated in the D-pattern cells remain to be elucidated. One possible mechanism is that of asymmetric division, 30,45,46 as MSI1 was originally reported to play a role in asymmetric division. 47 The majority of current cancer therapies have been designed and developed against fast-growing cancer cells, even when CSCs are targeted. 9 This is despite the fact that resistance to these anticancer therapies has been repeatedly linked to the presence of quiescent or slow-growing CSCs. 8,48 Given that CSCs fluctuate between different growth states, therapeutic anticancer strategies targeting CSCs should be seriously revisited. 5,10,13,[21][22][23]40 In this study, we revealed that the slow-growing cells play an important role in drug resistance, and that the two identified subpopulations are interchangeable.
The concept of the ''drug-tolerant persister'' (DTP) has recently emerged as an important driver of therapy failure and tumor relapse. A DTP is a cancer cell that is characterized as being quiescent or slow-cycling. 23 Cancer cells may enter a nongenetic and reversible DTP state to evade cellular death from conventional chemotherapies or molecular-targeted therapies. [49][50][51] The S-cell showed phenotypically DTP-like status in CRC organoids when treated with chemotherapeutic drugs. The isolated S-cells could be a novel platform for investigating DTPs and developing the DTP targeting treatment. Study of the transition to other fates, especially to the non-growing spheroids in future would provide more attractive targets for cancer therapy.

Limitation of the study
While our spheroid formation and growth assay with the single-cell resolution revealed the existence of two distinct but interchangeable subpopulations in CRC tumors and organoids, there are some limitations in this study. First, the assays in vitro generally run the risk of making the cells adapted to the culture conditions. Therefore, characteristics of these subpopulations in both patients' tumors and patient-derived tumor xenografts remain to be elucidated. In addition, CRC organoids showed S-pattern of growth when treated with chemotherapeutic drugs, while it should be further clarified how similar or different these spheroids are compared with the DTP cells in vivo. Investigating these issues will contribute to the clinical application of the findings in this study.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

Materials availability
This study did not generate new unique reagents.
There are restrictions to the availability of CRC CTOS lines through Material Transfer Agreement requirements at Kyoto University specific to the CRC CTOS and the subclones.
Data and code availability d The gene expression microarray data have been deposited at GEO and are publicly available as of the date of publication. Accession number is listed in the key resources table.
d Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.  52 Briefly, organoids or spheroids were disrupted into smaller fragments by passing them through a 1 mL syringe with a 27 G needle (Terumo, SS-10M2713) at a high flow rate ($30 mL/min). Organoids were spontaneously re-formed from these fragments. All experiments were performed at least one day after passaging to avoid the influence of the disruption and remodeling of the spheroids. Within one month of culture, organoids were freeze-stocked with StemCell Keep (BioVerde, BVD-VPL-AI-20). 'CRC organoid lines' were defined by the following criteria: i) growing continuously in culture, ii) generating xenograft tumor (at least 2 passages in vivo), and iii) being sufficiently freeze stocked in order to reproduce the experiments. Table S1 presents clinical details regarding the 14 organoid lines in the CRC panel. Single-cell-derived sphere-forming and growth (SSFG) assay A hundred to a thousand organoids were collected and dissociated into single cells by treating with 0.25% Trypsin-EDTA (Thermo Fisher Scientific, 25200072) and DNase I (10 mg/mL) for 10 min at 37 C and 1 min at room temperature, respectively. Then, the cell suspension was gently pipetted a hundred times to promote cell dissociation and filtered through a 35 mm cell strainer (Corning, 352235) to remove cell clusters. The dissociated single cells were diluted in the SSFG medium: CTOS organoid medium containing 2% Matrigel GFR and 10 mM of a ROCK inhibitor, Y-27632 (Selleckchem, S1049), and seeded in a non-treated 384-well plate (Sumitomo Bakelite, MS-9384U) with a ratio of 1 cell per well (50 mL/well) using an E1-ClipTip electronic pipette (Thermo Fisher Scientific, 4672060BT). Within 2 h after cell seeding (day 0), the presence of one cell per well was confirmed by image acquisition using the LEICA DMI4000B microscope (Leica Microsystems) equipped with Lumina Vision (Mitani Corporation). Single cells with an area greater than 300 mm 2 (small cluster) on day 0 and wells containing multiple cells were excluded from the subsequent analyses. The growth of single cells into spheroids was monitored by image acquisition. Fresh SSFG medium without Y-27632 (30 mL/well) was added on day 7. The ability to form spheroids as well as the growth were evaluated on day 13 for C45 clones and on day 20 for other lines unless otherwise noticed, by measuring the area of each single-cell-derived spheroid using the acquired pictures and ImageJ Fiji software (https:// imagej.net/software/fiji). Spheroid-forming capacity was calculated and expressed as the percentage of single cells able to grow and form spheroids. Mean G SD from at least three independent experiments is shown. For the experiments with the organoids (C45-1, C45-4S, and C45-4L) derived from freshly harvested xenograft tumors, we prepared organoids from the xenografts and subjected them to the SSFG assay within 6 h of preparation. For the experiments with chimeric spheroids, the single cell-derived spheroids with different fluorescence were assessed by image acquisition using the LEICA DMi8 microscope and LAS X software (Leica Microsystems). For the DAPT treatment experiments, tumor organoids (C45 and C45-4L) were treated with 50 mM of DAPT (Abcam, ab120633) or 0.1% of DMSO (Millipore Sigma, D5879) as control, cultured for 7 days, and subjected to the SSFG assay.

METHOD DETAILS
Isolation and culture of the slow-and dual-growing spheroids iScience Article round of SSFG assay was performed for each selected tumor spheroid clone to evaluate its growing pattern: slow-or dual-growing phenotype. For the additional rounds of the SSFG assay, the assay was sequentially performed for each clone and indicated as round one (x1), two (x2), and three (x3). Between the rounds, a pool of spheroids with similar size was collected and subjected to the next round of the SSFG assay.
Chimeric spheroids and co-culture system To generate chimeric spheroids, fluorescent-labeled spheroids were dissociated into single cells and mixed at a 2:1 ratio (1 3 10 4 cells of EGFP-labeled C45-4S: 5 3 10 3 cells of mCherry-labeled C45-4L; or 1 3 10 4 cells of tRFP-labeled sgRNA MSI1-C45-4L: 5 3 10 3 cells of wild-type C45-4L; or 1 3 10 4 cells of EGFP-labeled C45-4S: 5 3 10 3 cells of tRFP-labeled sgRNA MSI1-C45-4L). The cell mixture was suspended in the SSFG medium and dispensed into the round-bottom non-treated 96-well plate (Greiner bio-one, 650185). The plate was centrifuged at 400x g for 3 min to facilitate aggregation and incubated for 7 days. The chimeric spheroids were collected and subjected to the SSFG assay. For the DAPT treatment experiments, the fluorescent-labeled spheroids were pretreated overnight with 50 mM of DAPT or 0.1% of DMSO as control. Then, the chimeric spheroids were generated in the presence of DAPT (50 mM) or DMSO (0.1%), cultured for 7 days, and subjected to the SSFG assay. For the delayed treatment with DAPT, the chimeric spheroids were treated with 50 mM of DAPT or DMSO (0.1%) at day 2 of culture, and then cultured and subjected to the SSFG assay as described above. For the co-culture system without physical cell-cell interaction, 1 3 10 3 cells for each clone were separately embedded in 7 mL of Matrigel GFR, solidified as a droplet in a non-treated 24-well (IWAKI, 1820-024), overlaid with CTOS organoid medium containing 10 mM Y-27632, and cultured for 7 days. The EGFP-labeled-C45-4S spheroids were collected and subjected to the SSFG assay. For control experiments, the same number of EGFP-labeled C45-4S cells were cultured alone in the same way.

Organoid drug sensitivity assay
The organoid drug sensitivity assay, 19 spheroids with similar size and shape (diameter: 40-100 mm) were collected and seeded in non-treated 24-well plates at a density of 1 3 10 2 per well. Spheroids were cultured for one week in the CTOS organoid medium containing the indicated dose of the drugs or 0.1% of DMSO as control (n = 3 wells for each condition). Pictures of the entire well were captured on day 0 and day 7. The viabilities of the spheroids were evaluated using the CellTiter-Glo assay (Promega, G7570). ATP content was measured at day 7 and adjusted to the control group. The replicas of the spheroids at day 7 were subjected to the SSFG assay. The 5-FU (Kyowa Kirin Co., Ltd) compound was provided by the Department of Pharmacy, Osaka International Cancer Institute. The PD0325901 MEK1/2 inhibitor (Selleckchem, S1036) was used in this study.

Spheroid-forming cells treatment and regrowth assay
Tumor spheroids were dissociated into single cells as described before. Subsequently, 5 3 10 2 cells were embedded in 7 mL of Matrigel GFR and solidified as a droplet in non-treated 96-well plates (Thermo Fisher Scientific, 260860) with a ratio of one drop per well. Upon solidification, dispersed single cells were cultured for 7 days in 100 mL of CTOS organoid medium containing 10 mM of Y-27632 and 1 mM of the the indicated drugs (5-FU, PD0325901) or DMSO (0.1%) as control (n = 3 G 4 wells for each condition). Reconstituted spheroid viability was evaluated using the CellTiter-Glo assay. ATP content was measured at day 1, day 7, and adjusted to the control group. For the regrowth assay, the medium containing the indicated compound was removed at day 7 of culture. Then, each well was washed twice with HBSS and fresh CTOS organoid medium was added (100 mL). Spheroids were cultured for an additional week (day 14). The spheroid regrowth was assessed by measuring ATP content (CellTiter-Glo assay) at day 14 of culture and adjusted to the control group. Pictures of the entire well were captured on day 1, day 7 and day 14. For the DAPT treatment at the timing of the drug withdrawal, spheroids were treated with 50 mM of DAPT or DMSO (0.1%) at day 7, and the spheroid regrowth was evaluated at day 14 of culture as described above. The replicas of the spheroids at day 14 were subjected to the SSFG assay.

Cell death assay
To detect the cell death, tumor spheroids were cultured in non-treated 24-well plates, at a density of 0. 5