Differential regulation of the proliferative and autophagic responses in murine and human pancreatic cancer cells

ABSTRACT

Pancreatic ductal adenocarcinoma (PDAC) cells rely on autophagy to adapt to microenvironmental stress. In our recent work, we showed that baseline autophagy flux levels shape the proliferative capacity of human PDAC cells. Here, we extend these findings by demonstrating that low baseline autophagy is consistently associated with enhanced proliferation in vitro and in vivo across multiple human and murine PDAC models. While this inverse relationship between autophagy flux and proliferation is conserved across species, the underlying regulatory mechanisms diverge. Namely, changes in baseline autophagy levels relied predominantly on canonical nutrient-sensing pathways in murine but not human PDAC cells. Additionally, we found that autophagy flux was regulated independently of the cell cycle or the genetic status of p53 (i.e., allelic deletion, point mutation). Collectively, our results reveal conserved phenotypic outcomes but divergent mechanisms, underscoring the need for parallel comparative approaches before extrapolating findings from mouse PDAC models to human disease.

Introduction

Macroautophagy (hereafter referred to as autophagy) is a key cellular process that captures and degrades various intracellular substrates such as organelles and proteins to support cellular fitness under stressful conditions [1]. Given its important role as a gatekeeper of cellular homeostasis, autophagy is dysregulated in severe diseases, including cancer [2]. Among various cancer types, the progression of pancreatic ductal adenocarcinoma (PDAC) is particularly dependent on autophagy to sustain cellular growth and metabolic demands [3,4]. The recent development of tools to quantitatively measure autophagy flux [5,6] has brought to light the dynamic nature of this process and the ability of tumor cells to transiently modulate their autophagy flux levels in response to specific microenvironmental stresses and drug treatments [7–9].

Using the GFP-LC3-RFP (GLR) reporter, developed by the Mizushima lab [5], we have recently shown that human PDAC cells display heterogenous levels of high, moderate, and low autophagy levels in orthotopic pancreatic tumors; these observations could be reproduced in 3D cultures of PDAC spheroids [9]. Through flow sorting and analyzing human PDAC cells with the 10% lowest and highest autophagy levels from spheroids and tumors, we found that they display different growth capabilities and responsiveness to drugs [9]. Namely, human PDAC cells with low autophagy levels were conducive to growth, while those with high autophagy levels were chemo-tolerant [9].

While autophagy promotes overall PDAC progression, decreasing its flux levels appears to be a prerequisite to allow for active proliferation. Our recent results draw a more complex picture of autophagy regulation, where tumor cells with low and high autophagy flux levels both contribute to PDAC homeostasis.

Previous studies have shown a reciprocal co-regulation between autophagy and cell growth [10,11]. It is thought that these two processes inhibit one another, as exemplified by mTOR signaling, which inhibits autophagy and promotes cell growth. Similarly, prooncogenic factors such as MYC and YAP1 [9,12,13] reduce autophagy levels to promote proliferation, emphasizing the dire need of PDAC cells to coordinate active proliferation and autophagic survival for long-term maintenance.

Studies investigating the interconnections between autophagy and cell growth have found contrasting results, suggesting that the mechanisms are context- and disease-dependent [14,15]. However, the exact relationship between autophagy and cell proliferation remains poorly investigated in PDAC, leaving open the possibility for further investigations.

Here, we aimed at expanding our recent observations to question whether the negative association between autophagy levels and cell proliferation is generalizable to various human and mouse PDAC lines. Additionally, we asked: Do PDAC cells with various autophagy levels associate with distinct cell cycle phases? Do similar mechanisms operate in mouse and human PDAC cells?

We found that both human and mouse PDAC cells with low baseline autophagy levels exhibited better proliferative capabilities, in vitro and in vivo, independently of cell cycle phases. Interestingly, although both human and mouse cells displayed a similar proliferative phenotype, the underlying autophagy-regulating mechanisms were different, indicating that data from PDAC mouse models should be interpreted with care before extrapolating to the human context.

Results and discussion

Baseline autophagy levels are inversely associated with growth rates in PDAC

To assess the relationship between autophagy levels and cell growth, we generated multiple human and mouse PDAC cell lines expressing the GLR reporter and cultured them as spheroids in 3D to closely recapitulate the spatial distribution of PDAC cells in the tumor as well as their various autophagy flux levels and biological responses [9,16–18].

Using flow cytometry, we sorted the 10% of PDAC cells with the lowest (Low-Auto) and highest (High-Auto) autophagy flux levels. RNA-Seq analysis revealed that a large network of replication- and proliferation-related genes were enriched in human and mouse Low-Auto cells compared to their High-Auto counterparts (Figure 1(A); Table S1); this is in line with the idea that the regulation of autophagy and cell growth is coordinated at the transcriptional level [11].

Figure 1.

Low baseline levels of autophagy flux promote PDAC proliferation in vitro and in vivo. (A) RNA-Seq GSEA analysis on Low-Auto and High-Auto populations flow-sorted from HY19636 and PATU8988T spheroids. (B) Illustration of the strategy used to assess the growth of Low-Auto and High-Auto PDAC cells. (C) 2D clonogenic growth of Low-Auto and High-Auto cells flow-sorted from HY15549, HY19636, PATU8988T and PATU8902 spheroids. OD, optical density. (D) 3D growth of Low-Auto and High-Auto cells flow-sorted from HY15549 spheroids. (E) Illustration of the strategy used to assess EdU incorporation in Low-Auto and High-Auto PDAC cells. (F) EdU incorporation in the Low-Auto and High-Auto populations from HY19636 and PATU8988T spheroids. (G) Quantification of data presented in panel F. (H) Illustration of the strategy used to assess the in vivo tumor growth of Low-Auto and High-Auto PDAC cells. (I) Weight and autophagy flux of orthotopic pancreatic tumors derived from the Low-Auto and High-Auto populations flow-sorted from HY19636 and MIAPACA2 spheroids. Data are mean ± SEM. Statistical significance was tested by two-tailed Student’s t-test. *p < 0.05, **p < 0.01, ***p < 0.001.

To test the functional relevance of our observations, we assessed the ability of Low-Auto and High-Auto cells to grow under stress conditions using 2D clonogenic and 3D spheroid-forming assays. In both experiments, Low-Auto cells formed a higher number and larger clones or spheroids than their High-Auto counterparts Figure 1(B-D); these results were reproducible across multiple human and mouse PDAC cell lines. Accordingly, DNA replication assessed by EdU incorporation was ~4-fold greater in Low-Auto compared to High-Auto cells Figure 1(E-G). In Vivo, orthotopic pancreatic tumors derived from Low-Auto cells grew significantly larger than those originating from High-Auto cells Figure 1(H-I).

Since the mouse PDAC cell lines, HY19636 and HY15549, have a mono-allelic deletion of Trp53, we expanded the genetic background to include the MT4 and 4662 mouse PDAC lines harboring a Trp53^R172H point mutation. Similar to HY19636 and HY15549, we found that Low-Auto cells from MT4 and 4662 cell lines grew faster in 3D culture than their High-Auto counterparts (Figure S1).

Altogether, our data show that human and mouse PDAC cells with the lowest baseline autophagy levels have a higher proliferative capacity. These results were reproducible across PDAC cell lines regardless of whether p53 was inactivated through allelic deletion or point mutation.

Cell cycle progression does not impact autophagy flux levels in PDAC

Building on the different proliferative capabilities of the Low-Auto and High-Auto cells, we questioned whether these populations are present in distinct cell cycle phases. Previous studies investigating the regulation of autophagy by the cell cycle have obtained contrasting data [15]. For instance, using colon cancer cells, Tasdemir et al. found that autophagy induction preferentially occurred in the G1 and S phases [19]. In line with this observation, Furuya and colleagues postulated that the maintenance of genome integrity during mitosis is achieved through CDK1-mediated autophagy inhibition in neuroglioma cells [20]. However, other studies using lung, breast, and cervical cancer cells found no difference in autophagic flux activation in G1, S, and G2/M phases and demonstrated the persistence of autophagic activity during mitosis [21–23]. These divergent results might reflect a context-specific regulation of autophagy that is dependent on the type of cancer.

Using our system, we found that most of the Low-Auto and High-Auto (70–80%) populations were present in the G1 phase Figure 2(A). In line with their higher proliferative capacity, Low-Auto cells tended to be more present in the S and G2/M phases, but this portion represented only a minor fraction of the Low-Auto population Figure 2(A). We concluded that there is no major change in the distribution of Low-Auto and High-Auto cells across different cell cycle phases. Our analysis suggests that the proliferative Low-Auto cells progress faster into the cell cycle, rather than being stably present in a specific cycling phase.

Figure 2.

Cell cycle status does not control the level of autophagy flux in PDAC spheroids. (A) Distribution of Low-Auto and High-Auto populations in different cell cycle phases. (B) Illustration of the strategy used to assess autophagy flux in synchronized and asynchronized PDAC cells. (C) Autophagy flux in both synchronized and asynchronized PATU8902 cells. (D) Distribution of synchronized and asynchronized PATU8902 cells across different cell cycle phases. Statistical significance was tested by two-tailed Student’s t-test. *p < 0.05, **p < 0.01, ***p < 0.001. DTB: double-thymidine block.

To test whether the progression into different cell cycle phases is sufficient to modulate autophagy flux levels, we synchronized human PDAC cells at the G1/S boundary and released them in 3D culture at different time points Figure 2(B). Over the first 4 h post-release, we observed a drastic difference in human PDAC cell distribution in the G1 and S phases, which did not translate into change in autophagy flux levels Figure 2(C-D). Similarly, human PDAC cells shifting into the G2/M phase, at 8 h post-release, did not modulate autophagy flux levels Figure 2(C-D). These findings were reproducible in mouse PDAC cells (Figure S2).

Collectively, our results indicate that the level of autophagy flux defines the proliferative rate of PDAC cells, whereas the progression into different cell cycle phases does not determine autophagy flux levels in PDAC.

Mouse and human PDAC cells display distinct molecular features and autophagy-regulating mechanisms

As cell cycle status did not impact autophagy flux in PDAC cells, we examined the potential environmental cues controlling autophagy levels and, therefore, the formation of Low-Auto and High-Auto populations in PDAC spheroids and tumors. To get more insights, we spatially mapped the Low-Auto and High-Auto populations on frozen spheroid sections through calculating the endogenous GFP/RFP ratio.

Previous studies have reported the presence of an outside-in nutrient gradient within epithelial spheroids, where cells at the center of the spheroid have limited access to nutrients [16]. Accordingly, if High-Auto cells are nutrient-deprived, we expect them to be present at the center of the spheroids, while the Low-Auto ones, having access to nutrients, would be at the periphery.

Interestingly, Low-Auto and High-Auto populations were, respectively, located at the periphery and the center of mouse PDAC spheroids Figure 3(A), whereas no clear spatial segregation between these two populations was observed in human PDAC spheroids Figure 3(A). Similarly, we recently described the absence of a drastic spatial regionalization between Low-Auto and High-Auto populations across various cell lines of human PDAC spheroids [9]. In this report, we complement our previous data and discover that, contrary to the human model, mouse PDAC cells display clear features of nutrient starvation.

Figure 3.

Murine PDAC spheroids, rather than human ones, rely on nutrient sensing to control autophagy flux levels. (A) Representative images showing immunolabeling for p-S6^S240/44 (red) and the spatial distribution of Low-Auto (green/yellowish cells; white arrows) and High-Auto (orange cells; asterisk) cells on frozen spheroid sections from the indicated mouse and human PDAC cell lines. GFP-LC3 and RFP reflect endogenous baseline fluorescence. (B) Quantification of mTOR activity, through p-S6^S240/44 immunolabeling, related to panel A. (C) Western blot analysis on Low-Auto and High-Auto populations flow-sorted from HY19636 and PATU8902 spheroids; ERK2 and vinculin are used as loading controls. (D) Illustration summarizing the main findings of the study. Data are mean ± SEM. Statistical significance was tested by two-tailed Student’s t-test. **p < 0.01.

mTOR is a master regulator of cell growth and autophagy [10]. In response to nutrient starvation, mTOR is inhibited leading to autophagy activation, while under nutrient-replete conditions, mTOR is activated, which in turn inhibits autophagy flux [24,25]. From this perspective, mTOR could be considered as a rheostat regulating these cellular processes in response to nutrient sensing. Spatial analysis of p-S6^S240/44 immunolabeling on spheroid sections revealed a restricted activation pattern of mTOR to the periphery of mouse PDAC spheroids; thus, colocalizing with the Low-Auto population Figure 3(A,B). However, mTOR activation did not exhibit a significant difference between Low-Auto and High-Auto populations in human PDAC spheroids Figure 3(A,B).

Using western blot, we confirmed the reduced activation pattern of mTOR (S6^S240/44 phosphorylation) in the murine High-Auto cells compared to their Low-Auto counterparts Figure 3(B-C). Similarly, we found that the upstream phosphorylation of p70S6K at Thr389 and Thr229 sites was reduced in the murine High-Auto group compared to the Low-Auto population (Figure S3A). These differences in S6 and p70S6K phosphorylation occurred in mouse PDAC cell lines regardless of whether Trp53 was deleted or inactivated through the R172H mutation (Figure S3B-C).

Interestingly, human PDAC spheroids did not display any major differences in the phosphorylation patterns of S6 and p70S6K (Figure 3(B,C) and S3A).

Therefore, compared to the human setting, murine PDAC cells display a higher intrinsic dependency on canonical nutrient sensing pathways to regulate their baseline autophagy levels, regardless of the genetic status of p53.

Conclusions

In this report, we show that autophagy induction in PDAC organotypic cultures is associated with a reduction in DNA replication. We believe that this might reflect a strategy employed by PDAC cells to slowdown energy-consuming processes, such as protein synthesis and DNA replication, whose maintenance under stress and limited resources would be lethal for the tumor [26].

Surprisingly, we found that while both mouse and human PDAC cells displayed similar growth phenotypes in response to changes in autophagy flux levels, the underpinning autophagy-regulating mechanisms were different Figure 3(D). We have recently demonstrated that human PDAC cells can rely on the Integrin α3–Hippo–YAP1 axis to regulate autophagy flux levels, in response to extracellular matrix sensing, independently of canonical pathways such as mTOR [9]. Here, we found that mouse PDAC cells appear to be more dependent on mTOR to regulate their autophagy levels.

Although both murine and human PDAC cells are able to induce autophagy by inhibiting mTOR activity, under nutrient deplete conditions, such responses do not necessarily reflect the predominant regulatory mechanisms operating within the tumor microenvironment, where PDAC cells are exposed to more nuanced spatial and temporal stresses. Tumor-like 3D cultures might provide a physiologically relevant platform to uncover autophagy-regulating pathways driven by intrinsic PDAC dependencies, rather than global stress responses; an important distinction for translational relevance and therapeutic targeting.

The divergence in the signaling responses that we describe in the tested mouse and human PDAC models might either be due to an intrinsic species-specific mechanism or more directly to a difference in the mutational profile; this might represent one limitation of our study.

In this report, we have partially investigated this possibility by showing that the genetic status of p53 (i.e., allelic loss, point mutation) did not seem to impact on the signaling pathways or the phenotypic outcomes described in mouse PDAC cell lines. Future studies including additional models such as human PDAC cell lines with TP53 allelic loss can complement our observations and help elucidate whether the difference we observe between the mouse and human PDAC models is genetic, environmental, or a combination of both factors.

In the meantime, and regardless of the underlying cause, our findings convincingly show that a similar phenotypic observation in the mouse and human PDAC models, widely used by the cancer research community, does not always reflect similar autophagy-regulating mechanisms. In the same way, the regulation of autophagy by a shared mechanism between mouse and human PDAC cells might translate into different biological phenotypes. These discrepancies may partly explain the limited clinical success of numerous promising discoveries made in PDAC mouse models. Our results emphasize the importance of parallel comparative approaches using both mouse and human PDAC models, encompassing diverse genetic backgrounds within each species, to enable more comprehensive phenotypic and mechanistic characterization of autophagy regulation.

Materials and methods

Orthotopic pancreas surgery

The description of our mouse work adheres to the ARRIVE guidelines. All animal experiments were approved by the NYU Institutional Animal Care and Use Committee (protocol number IA16-00507). Athymic nude female mice were used in the described experiments at 8-to-12-week-old. All mice were housed in ventilated cages (5 mice/cage) within a temperature controlled ABSL-2 facility with 12:12 h day–night light cycles. Chow and water were provided ad libitum.

Mice were anesthetized using a ketamine/xylazine solution. After making an incision in the upper left abdomen, the spleen and pancreas were gently pulled out of the peritoneal cavity. Then, PDAC cells in HBSS were mixed in Matrigel (1:1 ratio) and injected into the pancreas (20 µl total volume). The incision was then sutured, and mice placed on a temperature-controlled heat-pad and monitored until they recover from anesthesia. Mice received one subcutaneous injection of extended-release pain killer treatment (391096, Covetrus), lasting 48–72 h after surgery.

A total of 7 × 10³ Low-Auto and High-Auto cells from the mouse HY19636 and human MIAPACA2 spheroids were flow-sorted and separately implanted into the pancreas of B6 (mice number: n = 12 Low-Auto; n = 13 High-Auto) and nude (mice number: n = 7 Low-Auto; n = 8 High-Auto) mice, respectively. At the time of implantation, mice were randomly chosen to receive an injection of either Low-Auto or High-Auto cells. The operator performed the pancreatic implantation blindly, unaware of the specific identity of the cells. The Low-Auto group was used here as control. HY19636 and MIAPACA2 tumors, respectively, were collected and weighed at 3- and 9-weeks post-implantation. All animals and data points were included in this study.

Cell culture

Unless otherwise stated, all experiments on PDAC cells were done in 3D culture and within 3-to-4 days after flow sorting.

HY19636 (Gift from Dr. Haoqiang Ying, Kras^G12D/+, Trp53^−/+), HY15549 (Gift from Dr. Haoqiang Ying, Kras^G12D/+, Trp53^−/+), 4662 (Gift from Dr. Bong Jun Kim, Kras^G12D/+, Trp53^R172H/+), MT4 (Gift from Dr. Bong Jun Kim, Kras^G12D/+, Trp53^R172H/+), PATU8988T (DSMZ, ACC 162, KRAS^G12V, TP53^R282W), PATU8902 (DSMZ, ACC 179, KRAS^G12V/+, TP53^C176S), MIAPACA2 (ATCC, CRL-1420, KRAS^G12C, TP53^R248W), and HEK293T (ATCC, CRL-3216) were maintained at 37°C and 5% CO₂ in a humidified atmosphere and cultured in DMEM full media (MT10017CV, Fisher Scientific) containing 10% fetal bovine serum (FBS) (S11550H, Bio-Techne) and 1% penicillin–streptomycin (15140163, Life Tech).

To generate cell lines stably expressing the GLR autophagy reporter (117413, Addgene), we first produced retroviral particles in HEK293T cells. Briefly, retroviral packaging vectors (gag/pol and VSV-G, 2 µg each), diluted in Opti-MEM, were transfected into HEK293T cells using lipofectamine 2000 (11668019, Thermo Fisher Scientific). Transfection media was replaced with fresh media after 24 h. Virus was collected after 48–72 h for immediate use or storage at −80°C. PDAC cells were stably transduced with retroviral particles using polybrene (8 µg/mL). Since the used GLR reporter has no antibiotic resistance cassette, we used flow cytometry to purify the cells expressing both GFP and RFP. Flow gating was restricted to the center of the double-positive population to ensure the sorting of cells with similar integrated copy number of the reporter. The percentage of cells stably expressing the GLR reporter was around ~85–90% as determined by post-sort analyses.

For 3D spheroid culture, cells were mixed with 3–5% Matrigel (CB 40230C, Fisher Scientific) in full DMEM and cultured for 3–4 days in ultra-low attachment 6-well plates at 5.5 × 10⁵ cells/well (CLS3471-24EA, Sigma-Aldrich). Spheroid images were captured using an optical microscope to assess their number and diameter using ImageJ software. In select experiments, spheroids were dissociated into single cells for subsequent applications.

Dissociation of spheroids and flow cytometry

Human and mouse spheroids expressing the GLR reporter were dissociated into single-cell solution using TrypLE Select 2X (A12177-01, Thermo Fisher Scientific) diluted in PBS containing 1 mM of EDTA (PBS-EDTA) for 10–15 min at 37°C. The digestion was stopped by adding full DMEM media and pelleting the cells by centrifugation (3 min, 500 g, 4°C). Dissociated cells were incubated for 5 min on ice with DAPI viability dye (564907, BD Biosciences) diluted to 1:1000 in PBS-EDTA. After incubation with viability dyes and centrifugation, cells were resuspended in HBSS containing 1 mM EDTA and 3% FBS (flow buffer) and subjected to flow cytometry.

Autophagy flux was determined using the GFP/RFP ratio. The 10% of cells with the lowest (Low-Auto) and highest (High-Auto) GFP/RFP ratio were gated and sorted using the MoFlo XDP cell sorter (Beckman Coulter) for subsequent experiments. The purity of sorted cells was confirmed to be higher than 85% by post-sort analyses. In another set of experiments, autophagy flux was acquired using the ZE5 analyzer (Bio-Rad), and Low-Auto and High-Auto populations were gated and analyzed using the FlowJo v10.8.1 software.

Clonogenic growth assays

Human and mouse PDAC spheroids expressing the GLR reporter were dissociated into single cells and flow-sorted to isolate Low-Auto and High-Auto populations. To assess clonogenic growth, flow-sorted cells were counted and immediately plated into 2D 6-well plates (700 cells/well) in full DMEM media.

After 7 days, cells were washed with PBS and then fixed for 15 min at RT with 10% methanol solution. Then, cells were stained with crystal violet solution (0.5% crystal violet dye, 25% methanol) for 15 min at RT. Excess stain was removed by immersing the plates 3-to-5 times in a running water bath. After drying the plates, the colorant was eluted using 10% acetic acid and optical density, which developed proportionally to cell count, was determined at 560–590 nm using a multiplate reader [27].

EdU incorporation

All reagents were prepared according to manufacturer’s instructions, but the protocol was slightly adapted to our spheroid-derived cells. EdU incorporation was assessed using the Click-iT™ EdU Alexa Fluor™ 647 (C10419, Thermo Fisher Scientific). After 3 days of culture, spheroids expressing the GLR reporter were dissociated into single cells and incubated with full DMEM media containing EdU (10 µM) for 2 h under standard growth conditions, in round-bottom low-attachment plates (351177, Corning). After incubation, cells were washed once with PBS containing 1% BSA (PBS-BSA), prior to permeabilization and fixation with ice-cold 70% ethanol solution for 15 min at 4°C. Cells were then washed once with PBS-BSA supplemented with 10% saponin (PBS-BSA-saponin), prior to incubation with the Click-iT® reaction cocktail for 30 min at RT, protected from light. Next, cells were stained with FxCycle™ Violet Stain (F10347, Thermo Fisher Scientific) diluted at 1:1000 in PBS-BSA-saponin buffer for 30 min at RT. Finally, cells were washed once and resuspended in PBS-BSA prior to analysis by flow cytometry using the ZE5 analyzer (Bio-Rad). Centrifugations were performed at 500 g for 3 min at 4°C.

Cell cycle analysis

Spheroids expressing the GLR reporter were dissociated into single cells and fixed/permeabilized with ice-cold 70% ethanol solution for 15 min at 4°C. Cells were stained with FxCycle™ Violet Stain diluted at 1:1000 in PBS-BSA buffer for 30 min at RT. Data were acquired using the flow cytometry ZE5 analyzer (Bio-Rad). Using the FlowJo v10.8.1 software, we gated on the 10% of cells with the lowest (Low-Auto) and highest (High-Auto) autophagy levels to calculate the distribution of Low-Auto and High-Auto cells across G1, S, and G2 cell cycle phases.

In another set of experiments, PDAC cells expressing the GLR reporter were synchronized at the G1/S boundary using a double thymidine block. Cells cultured in 2D were treated with 2 mM thymidine (T1895-1 G, Sigma-Aldrich) in full DMEM media for 18 h. Then, cells were washed once with PBS and allowed to grow in fresh DMEM media for 9 h. Thymidine treatment was repeated for an additional 18 h. Cells were either collected immediately at t = 0 or released from the G1/S boundary in 3D culture and collected at different timepoints (2, 4, 8, 12, 24, and/or 96 h). Cells were harvested, fixed, and stained for cell cycle analysis as mentioned above. Next, cells were analyzed by flow cytometry (ZE5 analyzer) and FlowJo v10.8.1 software to determine autophagy flux and PDAC cells distribution at G1, S, and G2 phases.

RNA-Seq

Total RNA was extracted from flow-sorted Low-Auto and High-Auto PDAC cells using a column-based approach [28] (AM1912, Thermo Fisher Scientific). Sequencing libraries were performed on enriched mRNA pools after the depletion of ribosomal RNA. First-strand cDNA was synthesized using random hexamer primers and M-MuLV Reverse Transcriptase (RNase H minus). The second strand cDNA synthesis reaction was performed using DNA polymerase I and RNase H, in which dTTPs were replaced with dUTPs. Then, the adenylated 3’ ends of cDNA fragments were ligated to P5/P7 sequencing adapters. Fragments of preferentially ~150–200 bp in length were selected through purification using the AMPure XP system (Beckman Coulter, Beverly, USA). Library preparations were sequenced on the NovaSeq PE150 platform with a coverage of 6 Gb.

For data analysis, clean reads were obtained by removing reads containing adapters and poly-N and low-quality reads. The index of the reference genome was built using Bowtie v2.0.6, and paired-end clean reads were aligned to the reference genome using TopHat v2.0.9. GSEA enrichment analysis was performed using R package clusterProfiler v4.10.1.

Immunolabeling of spheroids

Spheroids were fixed with 4% PFA for 5 min at RT, washed twice with PBS, and allowed to settle down, without centrifugation. Spheroids were allowed to pass through a solution of 30% sucrose overnight at 4°C, and they were mounted as OCT blocks, cryotome-sectioned, and stored at −80°C.

For immunolabeling, frozen sections were removed from −80°C and normalized to RT for 5 min. Then, they were permeabilized with a PBS containing 0.3% of Triton-100X for 15 min at RT, followed by blocking with PBS containing 10% normal goat serum and 1% BSA for 45 min at RT. Finally, the sections were incubated with the following primary antibodies overnight at 4°C (1:100 diluted in PBS): p-S6^S240/44 (5364, Cell Signaling Technology). Next day, slides were washed (3 × 5 min) with PBS containing 0.1% of Triton-100X and incubated with DAPI (1:1000) and the following secondary antibodies for 1 h at RT (1:200 diluted in PBS containing 5% normal goat serum): Goat anti-Rabbit IgG (H+L), Alexa Fluor 647 (A-21245, Thermo Fisher Scientific). Slides were mounted using Prolong Diamond Antifade solution (P36970, Thermo Fisher Scientific).

Confocal images were taken using Cytation C10 confocal imaging reader (Agilent, USA), and image analysis was performed using the Gen5 v3.12 software. DAPI staining was used as a primary mask to spatially localize individual cells. For each cell, the endogenous GFP/RFP ratio was calculated, which allowed the spatial localization of cells with the 10% lowest (Low-Auto) and highest (High-Auto) autophagy levels within spheroids. Using this system, we also measured the fluorescence intensity of p-S6^S240/44 within the Low-Auto and High-Auto populations.

Western blot

Flow-sorted Low-Auto and High-Auto cells were lysed in RIPA buffer (20–188, Sigma-Aldrich), supplemented with protease and phosphatase inhibitors (5892970001 and 4,906,837,001, Sigma-Aldrich). After centrifugation at maximal speed for 5 min protein quantification was performed on the supernatant using the DC protein assay (5000112, Bio-Rad). Samples were diluted in Laemmli buffer (1610747, Bio-Rad) and heated for 5 min at 95°C. A total of 30 µg of proteins were resolved on 4–20% Mini-Protean SDS-PAGE gels (4568096, BioRad). Transfer was performed at constant voltage at 100 V, for 1 h on ice. PVDF membranes (1620177, Bio-Rad) were blocked with 5% nonfat dry milk in TBST (Tris-buffer saline+Tween-20 (0.1%)) for 45 min at RT. Membranes were incubated, under gentle agitation at 4°C overnight, with the following primary antibodies: p-S6^S240/44 (5364, Cell Signaling Technology), S6 (2217 or 2317, Cell Signaling Technology), p-p70S6K^Thr389 (9205S, Cell Signaling Technology), p-p70S6K^Thr229 (81592–1-RR, Proteintech), p70S6K (9202S, Cell Signaling Technology), Vinculin (13901, Cell Signaling Technology), and ERK2 (sc-1647, Santa Cruz Biotechnology). Primary antibodies were diluted at 1:1000 in TBST containing 5% BSA. The phosphorylated and corresponding total protein forms were assessed on different membranes or on the same membrane after stripping with Restore PLUS Buffer (46430, Thermo Fisher Scientific). Next day, the membranes were washed with TBST (3 × 5 min) followed by an incubation with secondary antibodies (7074 and 7076, Cell Signaling Technology) diluted at 1:2000 in TBST containing 5% nonfat dry milk for 1 h at RT. After a final set of washes, the signal was developed using Clarity Western ECL Substrate (1705061, Bio-Rad), and images were acquired using the ChemiDoc Touch imaging system (Bio-Rad).

Statistical analysis

Two-tailed student’s t-test was applied for statistical analysis, using GraphPad Prism 10. p values were calculated and reported as follows: *, p < 0.05; **, p < 0.01; ***, p < 0.001. The error bars in each figure represented the standard error of the mean (SEM).

Funding Statement

M.A. is supported by the Brains-for-Brussels grant from Innoviris [BFB 2025-BFB-2]. M.A. is a group leader at the de Duve institute and a permanent research associate at the Belgian National Fund for Scientific Research (FNRS). R.W. is a postdoctoral fellow at NYU Langone Health.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

All data related to the study are presented in the supplementary material.

Assi, M., Wang, R., Kawaler, E.A., Sohn, A.S.W., Zahidunnabi Dewan, M., Kalfakakou, D., Encarnacion-Rosado, J., Kapner, K.S., Ganguly, K., Paulo, J.A., et al. (2026). Extracellular matrix sensing regulates intratumoral heterogeneity of autophagic flux. Cell. 10.1016/j.cell.2025.12.053.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/27694127.2026.2715823