Senescent cells and macrophages cooperate through a multi-kinase signaling network to promote intestinal transformation in Drosophila

SUMMARY

Cellular senescence is a conserved biological process which plays a crucial and context-dependent role in cancer. The highly heterogeneous and dynamic nature of senescent cells and their small numbers in tissues make in vivo mechanistic studies of senescence challenging. As a result, how multiple senescence-inducing signals are integrated in vivo to drive senescence in only a small number of cells is unclear. Here, we identify cells that exhibit multiple features of senescence in a Drosophila model of intestinal transformation which emerge in response to concurrent activation of AKT, JNK, and DNA damage signaling within transformed tissue. Eliminating senescent cells, genetically or by treatment with senolytic compounds, reduces overgrowth and improves survival. We find that senescent cells promote tumorigenesis by recruiting Drosophila macrophages to the transformed tissue, which results in non-autonomous activation of JNK signaling. These findings identify senescent cell-macrophage interactions as an important driver of epithelial transformation.

Graphical Abstract

eTOC blurb

Datta and Bangi identify cells that exhibit multiple features of senescence in a Drosophila model of intestinal tumorigenesis. They show that senescent cells promote tumorigenesis by recruiting Drosophila macrophages to the transformed tissue which then drive non-autonomous activation of JNK signaling in transformed cells.

INTRODUCTION

Cellular senescence is a highly conserved, dynamic, and regulated process crucial for proper embryonic development, tissue remodeling, repair, and aging^1–5. It is also a critical stress response mechanism induced by telomere shortening, DNA damage, oxidative stress, tissue damage, and aberrant activity of oncogenes^5,6. Oncogene-induced senescence (OIS), triggered in response to hyperproliferation and transformation driven by cancer-driving genetic alterations, occurs independent of telomere erosion and is a crucial defense against cancer^7–13. This tumor suppressive role is primarily mediated by inducing a stable cell cycle arrest of transformed cells and promoting immune surveillance^14,15. However, senescent cells can also promote tumorigenesis nonautonomously by remodeling the tumor microenvironment, which can result in increased proliferation, migration, and stemness of the neighboring non-senescent tumor cells and modulation of the immune system to make it more favorable for tumorigenesis¹⁴.

Cellular senescence is broadly defined as a stable and generally irreversible growth arrest mediated by CDK inhibitors p21 and p16, a persistent DNA damage response, extensive heterochromatinization evident by the formation of Senescence Associated Heterochromatic Foci (SAHF), elevated senescence-associated β-gal activity (SA-β-gal) and a senescence-associated secretory phenotype (SASP) enriched in soluble factors with proliferative, immunomodulatory, proinflammatory and ECM remodeling activity^1,16. Notably, features associated with senescence appear to be very heterogeneous, depending on the tissue, disease state, nature of the trigger, and time^1,17,18. In addition, not all senescent cells exhibit all features of senescence, and markers associated with senescence are not exclusive to this process^19–21. The highly heterogeneous nature of senescent cells, combined with their relatively low numbers in tissues, makes it particularly challenging to study how various stress signals are integrated at the tissue and cellular level to induce senescence in vivo²². As therapeutic targeting of senescence will require the ability to clearly define and identify these cells in different disease contexts, the heterogeneity of the senescent cell phenotype also poses a significant barrier to biomarker discovery and drug development efforts^17,18.

In addition to their heterogeneous features, the roles of senescent cells can vary widely depending on the disease and tissue contexts. For instance, senescence represents a powerful defense mechanism against cancer by inducing growth arrest of cells with activated oncogenes; a process demonstrated in experimental models and premalignant specimens from patients^5,23–25. However, the persistence of senescent cells in malignant tissue can lead to strong pro-tumorigenic effects, including tumor relapse and increased malignancy^26,27. The tumor-promoting effects of senescent cells are particularly alarming as many chemotherapies are known to induce senescence²⁷. The heterogeneous features and pleiotropic functions of senescence in cancer are further complicated by the genetically diverse nature of human tumors and complex interactions among tumor cells and their neighbors during tumorigenesis¹⁸. Experimental models that can capture and explore this complexity are crucial for a comprehensive mechanistic understanding of context-dependent contributions of senescent cells to tumor progression.

We have previously leveraged similarities between the Drosophila and human intestine to study intestinal transformation by genetically manipulating Drosophila orthologs of genes recurrently mutated in colon tumors^28,29. More recently, we established a 4-hit cancer model, KRAS TP53 PTEN APC, which expresses the oncogenic, mutant version of dRas^G12V while simultaneously knocking down Drosophila orthologs of the recurrently mutated colorectal cancer tumor suppressors TP53, PTEN, and APC in the larval hindgut epithelium³⁰. APC, TP53, and KRAS are the three most frequently mutated genes in colon tumors³¹. Similarly, the PI3K pathway, represented by pten loss in our model, is recurrently activated in many tumor types, including colon cancer, and is one of the most heavily targeted pathways in oncology drug development^31,32, making this model representing a common colon cancer genome landscape a highly relevant experimental system to study senescence.

Multiple studies in Drosophila demonstrated the presence of senescent cells in various Drosophila tissues, indicating that cellular senescence is an evolutionarily conserved mechanism^28,33–36. In this study, we show that cells that exhibit multiple features of senescence emerge in KRAS TP53 PTEN APC transformed hindguts, including nuclear accumulation of the Drosophila ortholog of p21, Dacapo (Dap), activation of DNA Damage Response (DDR) and formation of Senescence Associated Heterochromatic Foci (SAHF). We show that the senescent cell fate is determined in response to a multi-kinase network that includes the concurrent activation of AKT, JNK, and DDR signaling. Eliminating senescent cells from transformed tissue either genetically or by treatment with senolytic compounds³⁷ reduces tissue overgrowth and improve survival, demonstrating a tumor-promoting role for senescence in this genetic context. Our analysis shows that senescent cells recruit phagocytic hemocytes, the Drosophila macrophages^38,39, to the transformed tissue by inducing Mmp1 expression and compromising basement membrane integrity. Rather than clearing senescent cells and restoring homeostasis, which is their normal function⁴⁰, macrophages recruited to the transformed tissue trigger a second phase of broader, non-autonomous JNK pathway activation in the transformed epithelium to further promote tumorigenesis.

RESULTS

Cells with multiple senescence markers emerge in KRAS TP53 PTEN APC hindguts.

We have previously shown that targeting KRAS TP53 PTEN APC combination to the developing hindgut epithelium results in the expansion of the anterior region of the hindgut, where progenitor cells reside (Figure 1A,B) and increased proliferation (Figure 1C,D). Our analysis of KRAS TP53 PTEN APC transformed hindguts revealed a small number of cells expressing the Drosophila p21 ortholog Dacapo (Dap), which is undetectable in GFP-only control hindguts (Figure 1E). Additionally, we noted that subcellular localization of the Dap/p21 protein was different: some positive cells had nuclear localization of Dap/p21 while others had cytoplasmic (Figure 1F). p21 is a well-established marker of cellular senescence, suggesting that these cells may be senescent. A large body of prior work in multiple experimental systems has demonstrated that senescence is a dynamic and highly heterogeneous cellular state, emphasizing the importance of not relying on a single read-out to define cellular senescence. Therefore, we investigated whether Dap/p21 positive cells in KRAS TP53 PTEN APC hindguts exhibited other well-established senescence markers.

Figure 1. Emergence of cells with multiple markers of senescence and a tumor-promoting function in KRAS TP53 PTEN APC transformed hindguts.

(A) Anterior hindgut area, marked with yellow dotted lines, in GFP-only control and KRAS TP53 PTEN APC larval hindguts. The right panels show magnified views of areas highlighted in red boxes on the left. (B) Quantification of anterior imaginal ring area of control versus KRAS TP53 PTEN APC transformed larval hindguts. (C,D) phospho-Histone 3 (pH3) staining (white) of the control and KRAS TP53 PTEN APC transformed hindgut epithelium and (green) (C) and quantification of the number of pH3+ cells/hindgut (D). (E) Dap/p21 (white) and H3K9me3 (magenta) co-staining in hindguts (green) with indicated genotypes. (F) Magnified images of KRAS TP53 PTEN APC panels in (E) focusing on areas with nuclear and cytoplasmic Dap/p21 positive cells. White arrows indicate Dap/p21 negative cells that show the wildtype punctate H3K9me3 pattern. (G) Quantification of the number of cells positive for nuclear Dap/p21 and elevated H3K9me3 in hindguts with indicated genotypes. (H) Co-staining of H3K9me3 (magenta) and γ-H2AV (white) in hindgut epithelium (green) (I) Magnified areas of the KRAS TP53 PTEN APC panels outlined in red dashed rectangles in (H). Yellow and red arrows indicate cells with high and low γ-H2AV signal intensity, respectively. (J) Quantification of the fraction of H3K9me3 positive cells with low, high, or no γ-H2AV signal in KRAS TP32 PTEN APC hindguts. (K–P) The quantification of the anterior hindgut area (K–M) and survival to the pupal stage (N–P) of experimental animals with indicated genotypes and treatment conditions. (M,P) Error bars: Standard Error of the Mean (SEM). ****: p≤0.0001, ***: p≤0.001, **: p≤0.01, *: p≤0.05 (B,D,G,K,L,N,O) t-tests; (M,P) ANOVA, PRISM Software. Scale bars: 50μM.

SAHF, another commonly observed feature of senescence, can be evaluated using the heterochromatin marker H3K9me3 (tri-methylated Histone 3 on Lysine 9)⁴¹. KRAS TP53 PTEN APC transformed cells with nuclear Dap/p21 also had elevated levels of H3K9me3 visible throughout their nuclei; GFP-only controls or Dap/p21 negative cells in KRAS TP53 PTEN APC hindguts only exhibited the wildtype punctate staining that marks the mostly heterochromatic 4th chromosome pair in Drosophila cells (Figure 1E,F). In contrast, differentiated, polyploid enterocytes in the posterior hindguts of control and KRAS TP53 PTEN APC animals showed elevated levels of H3K9me3 staining, consistent with their higher ploidy⁴², but no Dap/p21 staining (Figure S1A). Combined presence of nuclear Dap/p21 and elevated H3K9me3 in KRAS TP53 PTEN APC is a strong indicator of senescence. Furthermore, nuclear Dap/p21 also colocalizes with RFP-HP1, a marker for Heterochromatin Protein 1, another well-established marker of SAHF^43,44(Figure S1B).

Notably, while nuclear Dap/p21 and elevated H3K9me3 were fully concordant, none of the cells with cytoplasmic Dap/p21 exhibited elevated H3K9me3 (Figure 1F). Cytoplasmic localization of p21 has been observed in some tumors and may have an oncogenic function, but the mechanisms underlying this role remain unclear^45,46. Whether cytoplasmic Dap/p21 plays a similar role in KRAS TP53 PTEN APC hindguts remains to be determined. Consistent with observations of senescent cells in vivo, the number of cells with nuclear Dap/p21and elevated H3K9me3 is low and variable, but they are present in all KRAS TP53 PTEN APC transformed hindguts (Figure 1G). We focused the rest of our analysis on these cells as they are also positive for two other well-established senescence markers.

Persistent DNA damage is another key feature of cellular senescence, commonly evaluated by the presence of γ-H2AX, which is produced by the phosphorylation of the histone variant H2AX and is a well-established read-out of DNA damage, repair, and senescence in mammalian systems and Drosophila^35,47–49. We found a high number of cells positive for γ-H2AV, the Drosophila equivalent of γ-H2AX, in KRAS TP53 PTEN APC transformed hindguts (Fig 1H). Colocalization studies using the SAHF marker H3K9me3 demonstrated that all cells with elevated H3K9me3 were also positive for γ-H2AV, with the majority exhibiting a low level of γ-H2AV signal intensity and a small number having high levels (Figure 1I,J). As all H3K9me3-high cells also have nuclear Dap/p21 (Figure 1E,F), this analysis demonstrates the presence of three well-established features of cellular senescence in these cells: Nuclear Dap/p21, formation of SAHF, and activation of DNA Damage Response (DDR). Significantly, not all γ-H2AV positive cells in KRAS TP53 PTEN APC transformed hindguts were positive for other senescence markers, consistent with previous studies that γ-H2AX is not an exclusive marker of cellular senescence. Our findings agree with previous work in mammalian models emphasizing the importance of using multiple markers to define senescent cells.

Senescent cells promote KRAS TP53 PTEN APC-induced intestinal transformation.

To determine whether senescence plays a role in KRAS TP53 PTEN APC-induced intestinal transformation, we sought to genetically eliminate senescent cells from the transformed tissue by reducing Dap/p21 levels. We found that knocking down Dap/p21 eliminated both nuclear and cytoplasmic Dap/p21 in KRAS TP53 PTEN APC hindguts and prevented the formation of H3K9me3 positive foci (Figure S1C), demonstrating that Dap/p21 is required for the emergence of SAHF, another key marker of senescence. Dap/p21 knockdown in KRAS TP53 PTEN APC hindguts resulted in a significant reduction in the anterior hindgut expansion and rescue of lethality (Figure 1K,N, S1D). We found similar results upon feeding the p21 inhibitor UC2288⁵⁰ (Figure 1L,O, S1E). Furthermore, senolytic drugs Navitoclax and Fisetin, which have been shown to clear senescent cells in multiple experimental systems³⁷, also eliminated senescent cells in KRAS TP53 PTEN APC hindguts, reduced tissue expansion, and improved organismal survival (Figure 1M,P, S1F,G). Combined, these results demonstrate a tumor-promoting function for senescent cells in KRAS TP53 PTEN APC-induced intestinal transformation.

Senescence emerges in response to concurrent activation of AKT, JNK, and DDR signaling.

Consistent with observations in human and other mammalian tissues, the number of Dap/p21 and H3K9me3 positive senescent cells in KRAS TP53 PTEN APC hindguts is low, and their location within transformed tissue is variable (Figure 1G), suggesting the presence of spatially restricted signals as determinants of the senescent phenotype. To understand how senescent cells emerge within transformed tissue and why some transformed cells become senescent while others in the same local microenvironment do not, we investigated whether signaling pathways activated in KRAS TP53 PTEN APC hindguts play a role as inducers of senescence.

We have previously demonstrated strong activation of AKT signaling in KRAS TP53 PTEN APC and other RAS/PI3K co-activated tumor models^28,30 (Figure S2A). Reducing Akt levels in KRAS TP53 PTEN APC hindguts eliminated Dap/p21 expression (Figure 2A), demonstrating that AKT signaling has a conserved role in inducing Dap/p21 expression during transformation. However, AKT signaling is broadly activated throughout the transformed tissue (Figure S2A), making it unlikely to be the localized signal determining which transformed cells will become senescent.

Figure 2. Determinants of senescence in KRAS TP53 PTEN APC transformed hindguts.

(A) Dap/p21 staining (yellow) of KRAS TP53 PTEN APC transformed hindguts (green) with and without akt knockdown. (B) Dap/p21 (yellow) and H3K9me3 (magenta) co-staining of KRAS TP53 PTEN APC transformed hindguts (green) with and without bsk knockdown. No nuclear dap/p21 or elevated H3K9me3 observed upon bsk knockdown. (C,D) phospho-JNK (pJNK, white, C) and γ-H2AV (white, D) staining of KRAS TP53 PTEN APC transformed hindguts (green) with and without akt knockdown. (E) γ-H2AV staining (white) of KRAS TP53 PTEN APC transformed hindguts (green) upon bsk knockdown. (F–H) γ-H2AV (white, F), Dap/p21 (white, G), and pJNK (white, H) staining of KRAS TP53 PTEN APC transformed hindguts (green) upon p53 overexpression. (I) Working model: Senescence emerges in response to concurrent activation of AKT, JNK and DDR signaling (see main text for details). (J–M) Staining and quantification γ-H2AV (magenta, J), and Dap (white, L) in KRAS TP53 PTEN APC with and without tefu knockdown in transformed hindguts (green). (K,M) Fraction of DDR positive cells with low γ-H2AV intensity (K) and the number of cells with nuclear Dap/p21 per hindgut (M) upon tefu knockdown in KRAS TP53 PTEN APC hindguts. Error bars represent SEM. ***p<_0.001, **p<_0.01. (t-tests, PRISM Software). Scale bars: 50μM.

Next, we explored a role for JNK signaling, which is activated in a more localized and variable manner in transformed tissue (Figure S2B). Knockdown of the Drosophila JNK ortholog basket (bsk) resulted in the complete elimination of both nuclear Dap/p21 and H3K9me3 as well as broad induction of cytoplasmic Dap/p21 (Figure 2B) and significant increase in survival of experimental animals (Figure S2C), suggesting that JNK signaling is required for Dap/p21 nuclear localization and senescence.

To better understand the relationship between localized JNK signaling and senescence, we took advantage of the well-established JNK reporter TRE-dsRed.⁵¹ We found that senescent cells were either positive for TRE-dsRed or were in close proximity to TRE-dsRed positive cells (Figure S2D). The high background of the TRE-dsRed reporter in the rest of the Drosophila intestine prevented a higher resolution analysis; it is possible that Dap/p21 positive TRE-dsRed negative cells may have a low level of JNK pathway activity that was not detectable in our experiments. Regardless, these experiments indicate that senescent cells emerge in regions of JNK pathway activity within transformed tissue and that JNK signaling is required for senescence.

As AKT signaling is also required for both nuclear and cytoplasmic Dap/p21 (Figure 2A), we next tested whether Akt acts upstream of JNK signaling to induce Dap/p21 expression. Reducing Akt levels in KRAS TP53 PTEN APC hindguts resulted in a complete loss of JNK pathway activity (Figure 2C), demonstrating that JNK pathway activation is downstream of AKT signaling. Furthermore, co-activation of JNK and AKT pathways in otherwise wildtype tissue by co-expressing wildtype Akt and a constitutively active form of the Drosophila JNKK Hemipterous (Hep) was sufficient to induce nuclear Dap/p21 in wildtype hindguts, while ectopic activation of each pathway alone was not, indicating a synergistic requirement (Figure S2E)

These findings demonstrate that activation of both AKT and JNK pathways is necessary to induce senescence within transformed tissue. We hypothesize that JNK activity in pAKT-positive cells leads to Dap/p21 nuclear localization and SAHF formation. However, the pattern of AKT activation is much broader than that of JNK signaling in KRAS TP53 PTEN APC transformed tissue (Figure S2A,B), suggesting that not all pAKT-positive cells activate JNK signaling. Combined, these findings point to the involvement of another signal regulating Dap/p21 expression and nuclear localization.

Persistent DNA Damage Response (DDR) activation is a well-established feature of senescent cells, also conserved in our experimental system (Figure 1H–J). As DNA damage is a known activator of JNK signaling, we next investigated the relationship between AKT signaling, JNK signaling, and DDR. Reducing Akt levels in KRAS TP53 PTEN APC transformed hindguts resulted in a complete loss of γ-H2AV signal (Figure 2D), demonstrating that AKT signaling is required to activate DDR. On the other hand, reducing JNK pathway activity by knocking down bsk did not affect γ-H2AV (Figure 2E, S2F), indicating that DDR is activated downstream of AKT signaling but does not require JNK signaling.

We next sought to determine whether persistent DDR signaling activates the JNK pathway in transformed tissue and induces senescence. As persistent DNA damage typically results in p53-dependent apoptosis in wildtype tissues, we reasoned that p53 overexpression in transformed tissue would result in the elimination of DDR-positive cells. As expected, overexpressing p53 in KRAS TP53 PTEN APC hindguts eliminated all γ-H2AV-positive cells from transformed tissue (Figure 2F). The loss of DDR activity in transformed tissue also prevented both Dap/p21 expression and JNK pathway activation (Figure 2G,H). These findings demonstrate that the JNK pathway is activated downstream of DDR and that the senescent phenotype is an emergent feature of high pAKT, JNK, and DDR signaling (Figure 2I).

We hypothesize that AKT-dependent excessive proliferation in transformed hindgut tissue results in DNA damage in many transformed cells, broadly activating DDR. Persistent DDR results in the activation of JNK signaling in some of the cells, leading to nuclear Dap/p21 and other features of senescence. While all senescent cells exhibit DDR, not all DDR-positive cells become senescent. In addition, γ-H2AV signal intensity varies from cell to cell and the senescent phenotype largely correlates with a low level of γ-H2AV intensity (Figure 1H–J) suggesting that JNK pathway activation and subsequent senescence may require a moderate threshold of DDR. In line with this hypothesis, TRE-dsRed positive cells within transformed tissue either had low levels of or no γ-H2AV signal (Figure S2G). These results, combined with the dependence of JNK signaling on DDR (Figure 2G), indicate that JNK signaling activation in response to a low level of DDR drives senescence. The presence of TRE-dsRed positive but γ-H2AV negative cells on the other hand, point to another potential source of JNK pathway activity that does not require cell-autonomous DDR activation but still dependent on it.

To further explore whether senescence is triggered in response to a low level of DNA damage response, we next sought to genetically reduce the level of γ-H2AV signal by knocking down tefu, the Drosophila ortholog of ATM and a key component of the DDR pathway⁵². tefu knock-down significantly increased the fraction of DDR+ cells with a low level of γ-H2AV in transformed tissue (Figure 2J,K). Consistent with our predictions, having more cells with a low level of DDR signaling significantly increased the number of senescent cells within transformed tissue (Figure 2L,M). Combined, our findings demonstrate that senescence is a highly calibrated and integrated cellular response to excessive proliferation, DNA damage, oncogenic and stress signaling.

A positive feedback loop between senescent cells and macrophages promotes intestinal transformation.

Despite their low numbers within transformed tissue (Figure 1G), senescent cells appear to have a significant tumor-promoting effect (Figure 1K–P). To investigate how senescent cells promote intestinal transformation, we tested whether they contribute to the expression of Mmp1, which is secreted by senescent cells in both mammalian systems and in Drosophila. KRAS TP53 PTEN APC transformed hindguts also exhibit strong and broad Mmp1 expression, which is eliminated upon Dap/p21 knockdown (Figure 3A), demonstrating that senescent cells are necessary for inducing Mmp1 expression in KRAS TP53 PTEN APC transformed hindguts.

Figure 3. Mechanism of tumor-promoting role of senescent cells via recruitment of hemocytes.

(A,B) MMP1 (white, A) and NimC1 (white, B) staining in hindguts (green) with indicated genotypes. (C) Quantification of the number of NimC1+ hemocytes associated with hindguts with indicated genotypes ****: p≤0.0001, *: p≤0.05 (ANOVA, PRISM Software). (D,E) NimC1 staining (white, B) and the quantification of the number of NimC1+ hemocytes recruited to KRAS TP53 PTEN APC (green) transformed hindguts upon ectopic expression of timp (E). ****: p≤0.0001 (t-tests, PRISM Software). (F,G) Anterior hindgut area, marked with yellow dotted lines in hindguts with indicated genotypes (green, F) and quantification of the anterior hindgut area size per hindgut (G) p: **: p≤0.01 (t-tests, PRISM Software). Error bars: Standard Error of the Mean. Scale bars: 50μM.

A well-established function of senescent cells in mammalian tissues is to recruit macrophages to facilitate tissue remodeling and act as a surveillance mechanism to prevent the accumulation of senescent cells^53–55. As macrophages are often recruited to sites with compromised basement membrane integrity^56,57, we tested whether senescent cells facilitate the recruitment of phagocytic hemocytes, the functional equivalent of mammalian macrophages^38,39,58, to the KRAS TP53 PTEN APC transformed tissue. We found that hemocytes, identified by the expression of the phagocytosis receptor NimC1^59,60, were recruited to KRAS TP53 PTEN APC transformed intestinal epithelium in a Dap/p21 dependent fashion (Figure 3B,C). Inhibition of Mmp activity by ectopic expression of Timp, an endogenous inhibitor of Mmp proteolytic activity⁶¹, significantly reduced the number of NimC1-positive hemocytes (Figure 3D,E). Blocking Mmp activity and subsequent hemocyte recruitment also significantly reduced the size of the anterior hindgut area (Figure 3F,G). These findings suggest that senescent cells contribute to tumorigenesis by inducing Mmp1 expression and recruiting hemocytes. Reducing hemocyte recruitment, on the other hand, had no effect on the number of senescent cells within transformed tissue (Figure S3A,B). Furthermore, disrupting basement membrane integrity by ectopically expressing Mmp1 in an otherwise wildtype hindgut is sufficient to recruit hemocytes (Figure S3C,D), suggesting that senescence-induced Mmp1 expression is the primary driver of hemocyte recruitment to the KRAS TP53 PTEN APC transformed tissue. Notably, hemocyte recruitment to an otherwise wildtype epithelium is not sufficient to induce transformation (Figure S3E,F), highlighting the context-dependent roles hemocytes play during tumorigenesis.

These findings demonstrate that senescent cells trigger Mmp1 expression within KRAS TP53 PTEN APC transformed tissue, compromising basement membrane integrity and recruiting hemocytes. Given the broad Mmp1 expression in KRAS TP53 PTEN APC transformed hindguts and the relatively low number of senescent cells, we next tested whether senescent cells induce Mmp1 expression non-autonomously. As Mmps are direct transcriptional targets of JNK signaling⁶², we reasoned that senescent cells may be expressing and secreting the Drosophila TNFα ortholog and the sole JNK pathway ligand Eiger^63,64, to drive non-autonomous JNK pathway activation and Mmp expression. Using the reporter construct eiger-LacZ⁶⁵, we found that eiger is not expressed in the transformed epithelium or the overlaying muscle; hemocytes were the only eiger-LacZ positive cells in KRAS TP53 PTEN APC transformed hindguts (Figure S4A–C).

As senescent cells recruit hemocytes to the transformed tissue, any JNK signaling within the hindgut epithelium induced by hemocyte-derived Eiger would be downstream of senescence. However, we have previously shown that JNK signaling is also necessary to induce senescence (Figure 2B,I) and consequently hemocyte recruitment. These findings suggest a biphasic JNK pathway activation within the transformed epithelium, the first phase upstream of senescence, activated in a ligand-independent fashion in response to DDR, and required for the formation of senescent cells, Mmp1 expression, and hemocyte recruitment. Eiger-expressing hemocytes then initiate the second phase of JNK pathway activation and Mmp1 expression. To test this hypothesis, we knocked down the Eiger receptor Wengen (Wgn) in the hindgut epithelium, which would be expected only to affect the second, Eiger-induced phase of JNK signaling and Mmp1 expression. Consistent with our hypothesis, wgn knock-down eliminated most of the Mmp1 expression within KRAS TP53 PTEN APC transformed hindguts, only leaving a small domain that still exhibited a strong Mmp1 signal (Figure 4A). Significantly, we observed the remaining Mmp1 signal upon wgn knockdown to concentrate around senescent cells (Figure S4D), consistent with our hypothesis that the first phase of Mmp1 expression is triggered by senescent cells. We observed a similar pattern with JNK pathway activity, where after wgn knockdown, only a few pJNK-positive cells remained (Figure 4B). Notably, the remaining Mmp1-positive (Figure 4A) and pJNK-positive cells (Figure 4B) continued to exhibit strong signal intensity, indicating that these cells do not depend on hemocyte-derived Eiger for JNK pathway activation. In contrast, reducing the level of Bsk, the Drosophila c-Jun N-terminal Kinase (JNK), which is further downstream of the pathway and required for all JNK signaling, resulted in a much broader loss of pJNK throughout the hindgut (Figure S2B). We conclude that wgn knockdown eliminates the second phase, non-autonomous JNK pathway activation. In contrast, bsk knockdown removes both the DDR-induced first phase and the hemocyte-induced second phase of JNK signaling downstream of senescence. Consistent with this, we also observed a much stronger reduction in the anterior hindgut area upon bsk knockdown than wgn (Figure S4E,F).

Figure 4. Positive feedback loop between senescent cells and hemocytes promote KRAS TP53 PTEN APC induced tumorigenesis in larval hindguts.

(A,B) Mmp1 (white, A) and pJNK (white, B) staining in hindguts with indicated genotypes. (C,D NimC1 staining (white, C) and quantification of NimC1 positive cells (D) recruited to the transformed hindguts from larvae with indicated genotypes. (E,F) Mmp1 (white, E) and pJNK (white, F) staining in KRAS TP53 PTEN APC transformed hindguts (green) from larvae with indicated genotypes. (G) Quantification of the pJNK positive area in anterior hindguts with indicated genotypes. Error bars: Standard Error of the Mean., p: **: p≤0.01, *p: ≤0.05 (t-tests, PRISM Software) Scale bars: 50μM. (H) Working model: Reciprocal interactions between senescent cells in the transformed epithelium and the hemocytes recruited to it promote tumorigenesis through a JNK pathway mediated positive feedback loop (see main text for details).

To further explore the mechanism of the interaction between the transformed epithelium and hemocytes, we sought to genetically eliminate hemocytes in control and KRAS TP53 PTEN APC larvae using the Q-system, another independent binary targeted expression tool that can be used simultaneously with the Gal4-UAS system^66–68‘. Inducing the expression of the apoptotic reaper (rpr) gene in hemocytes using the Q-system^67,69 in KRAS TP53 PTEN APC larvae resulted in a significant reduction in hemocyte recruitment to the transformed tissue (Figure 4C,D). If Eiger from hemocytes was responsible for driving the second phase non-autonomous JNK signaling in the transformed epithelium, we reasoned that genetically depleting hemocytes in KRAS TP53 PTEN APC animals would phenocopy wgn knockdown in the hindgut epithelium. Consistent with our hypothesis, we found a significant reduction in the size of the Mmp1 positive and pJNK positive domains in KRAS TP53 PTEN APC hindguts from hemocyte-depleted larvae, comparable to results from wgn knockdown in the transformed hindgut epithelium (Figure 4E–G).

Overall, our findings highlight a positive feedback loop between senescent cells and hemocytes and biphasic JNK pathway activation as an important driver of tumorigenesis (Figure 4H). During intestinal transformation, concurrent activation of AKT, DDR, and JNK signaling triggers senescence in a small number of cells. Senescent-cell-driven Mmp1 expression disrupts the basement membrane, resulting in the recruitment of phagocytic hemocytes, which express the JNK pathway ligand Eiger. Hemocyte-derived Eiger triggers a second phase of JNK pathway activation and MMP expression through its receptor Wgn in the transformed hindgut epithelium, further compromising basement membrane integrity and promoting hemocyte recruitment. Disrupting this feedback loop by eliminating senescent cells or blocking the Wgn-dependent second phase of JNK pathway activation suppresses tumorigenesis.

DISCUSSION

Cellular senescence is a dynamic process regulated by complex interactions that integrate several short- and longer-range signals with intrinsic properties of individual cells, including their genetic makeup, disease state, or tissue of origin. Capturing and functionally exploring additional layers of complexity underlying the senescent cell fate and the cell non-autonomous effects of these cells during tissue homeostasis and disease development requires complex genetic manipulations. In this study, we leverage the powerful genetic tools available in Drosophila for a mechanistic exploration of triggers and consequences of senescence in a genetically complex model of colorectal cancer.

Multiple studies have shown that cells can respond to stress by undergoing cell death, cell repair, or senescence depending on the nature of the stress signal, cell type, and tissue microenvironment^16,70,71. The emergence of senescence depends on the dynamic interaction of various stressors acting on a tissue microenvironment. As most senescence triggers are typically broadly activated, elucidating how these signals cooperate in vivo to drive senescence in only a few cells has been difficult. Our study untangles the intricate interactions between multiple oncogenic and stress response pathways, including AKT signaling, JNK signaling, and the DDR pathway underlying the emergence of senescent cells in during intestinal transformation. We demonstrate that a cooperative epistatic interaction between these pathways leads to senescence and genetic targeting of any of these three pathways is sufficient to prevent the emergence of senescent cells.

Senescent tumor cells have been reported for multiple tumor types both in primary tumor specimens and in metastases^14,15, though the specific triggers of senescence in different genetic contexts appear to be different and poorly understood. For instance, oncogenic KRAS-driven senescence is associated with DDR induced by reactive oxygen species⁷². RAS/MAPK-PI3K coactivation, on the other hand, abrogates senescence and drives tumorigenesis in multiple experimental systems^73,74. While TP53 is a critical player in the induction of cell cycle arrest, senescence, and apoptosis in response to DNA damage and stress, TP53-independent mechanisms of senescence have also been reported in mammalian cells^75–77. In this study, we demonstrate the emergence of senescent cells in KRAS TP53 PTEN APC transformed epithelium, which includes coactivation of RAS/MAPK and PI3K pathways and loss of p53 function, a commonly observed genetic landscape in multiple tumor types. It has been shown that AKT-induced senescence in TP53 wildtype cells does not display DDR, SAHF, or high nuclear p16; instead, it depends on TP53⁷³. Notably, we find that Akt-dependent senescence in the absence of p53 function displays all these classical hallmarks of senescence. Our findings emphasize the importance of studying triggers and consequences of senescence in genetically complex and diverse cancer models to fully understand its nuanced role in tumor progression.

Recruitment of macrophages to sites of tissue damage by senescent cells is a welldocumented senescence-surveillance mechanism to prevent the accumulation of senescent cells and promote tissue repair^40,78. In multiple experimental systems, macrophage ablation leads to the accumulation of senescent cells, compromising tissue repair, remodeling, and embryonic development^79–82. Like senescent tumor cells, macrophages also play a crucial role in cancer, which can be tumor-promoting or tumor-suppressive depending on the tumor type, stage, genotype, and microenvironment^83,84. However, the specific mechanisms mediating interactions between these cell types during tissue remodeling and cancer remain to be fully elucidated.

Tumor-associated hemocytes have also been reported in Drosophila cancer models, and their dual role in tumorigenesis is conserved⁸⁵. Furthermore, hemocyte-derived Eiger has been previously shown to both promote and suppress tumorigenesis in Drosophila models^86–88, but a potential role for senescence in these contexts has not been investigated. We demonstrate that senescent cells are the primary drivers of hemocyte recruitment by compromising basement integrity in an Mmp-dependent manner. The TNFα ortholog Eiger secreted by these “tumor-associated hemocytes” is then co-opted to achieve more widespread JNK pathway activation and Mmp1 expression within the transformed epithelium. These findings demonstrate how a comparatively low number of senescent cells in transformed tissue can non-autonomously trigger and maintain a self-sustaining positive feedback JNK signaling loop in cooperation with hemocytes to allow tumor growth and progression.

Notably, hemocytes recruited to damaged tissues secrete other ligands to promote tissue homeostasis which may also be appropriated by transformed tissue to promote tumorigenesis. For instance, hemocytes recruited to the damaged Drosophila intestine have been shown to secrete the BMP ligand Dpp to promote intestinal regeneration⁸⁹. Future studies exploring potential roles additional hemocyte-secreted molecules in different cancer models will be necessary to fully elucidate the interactions underlying hemocytes, senescent cells and transformed tissue.

Targeting of senescent cells via senolytic compounds has been found to ameliorate disease states, including age-related disease as well as certain cancer types,^90,91, although their mechanisms of action are still not well understood. Fisetin has been recently reported to reduce tumor progression in cancer patients and animal models^92–94; while navitoclax was reported to inhibit rapid tumor growth and development⁹⁵. We found that the senescent-cell clearing and anti-tumor effects of senolytic compounds are conserved in our experimental system, making it a powerful platform to explore the mechanisms of action of this promising class of cancer therapeutics. Furthermore, a better mechanistic understanding of interactions between senescent cells and macrophages offers opportunities to discover additional druggable nodes for therapeutic targeting of these cell-cell interactions in future studies.

Limitations of the Study

While we were able to confirm that the cells we identified as senescent exhibit multiple well-established features of senescence, our molecular characterization of senescent cells was limited by antibody availability and compatibility. As a result, we were not able to determine whether there are molecular differences among senescent cells. More unbiased approaches like single cell transcriptional profiling will be necessary to investigate senescent cell heterogeneity. Furthermore, our study focused on studying senescence in the context of a single, albeit genetically complex, cancer model. Given the context dependent nature of senescence and the genetic heterogeneity of cancer, extending this work to other cancer models that capture additional commonly observed cancer genome landscapes will be needed to gain broader insights into the role of senescence in cancer and better understand interactions among cancer-driving genomic alterations underlying senescence.

STAR METHODS

RESOURCE AVAILABILITY

Lead Contact

Further information and requests for resources should be directed to and will be fulfilled by the lead contact, Erdem Bangi (ebangi@bio.fsu.edu)

Materials availability

Reagents generated in this study are available upon request. Requests should be directed to and will be fulfilled by the lead contact.

Data and code availability

• All data reported in this paper will be shared by the lead contact upon request.

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

All Drosophila stocks were maintained on standard Drosophila media at room temperature. We have previously described the 4-hit model which we refer to as KRAS TP53 PTEN APC in this study³⁰. UAS-based RNAi and overexpression lines were obtained from Bloomington Drosophila Stock Center (BDSC). The following stocks were used: UAS-aktIR (#31701, BDSC)³⁰, UAS-bskIR (#31323, BDSC)^96,97, UAS-dapIR (#36720, BDSC)⁹⁸, UAS-mmp1 (#58701, BDSC), UAS-wgnIR (#50594, BDSC)^99,100, UAS-timp (#58708, BDSC), RFP-HP1 (#30562, BDSC), UAS-p53¹⁰¹ and eiger-lacZ⁶⁵ (gifted by K. Basler), TREdsredT4 (#59012, BDSC), UAS-hep.CA (#58780, BDSC), UAS-akt (#8191, BDSC), UAS-tefuIR(#44073, BDSC)¹⁰², w;wg/ Cyo; srpHemo-QF2/ (Tm3, Ser) (#78365, BDSC), w; QUAS-rpr; MKRS/Tm6b, Tb (#92782, BDSC).

All crosses were set up on Bloomington’s semi-defined media. Experimental animals used in this study were generated and induced as previously described³⁰. The genotype for experimental animals labeled as KRAS TP53 PTEN APC and control in the figure panels are: w¹¹¹⁸ UAS-dcr2/w; UAS-ras^G12V UAS-p53^RNAi UAS-pten^RNAi UAS-apc^RNAi/+; byn-gal4 UAS-GFP tub-gal80^ts /+ and w¹¹¹⁸ UAS-dcr2/w; +/+; byn-gal4 UAS-GFP tub-gal80^ts/+, respectively.

Briefly, virgins were generated by heat-shocking w¹¹¹⁸ UAS-dcr2/Y, hs-hid; UAS-ras^G12V UAS-p53^RNAi UAS-pten^RNAi UAS-apc^RNAi; byn-gal4 UAS-GFP tub-gal80^ts /S-T, Cy, tub-gal80, Hu, Tb and w¹¹¹⁸ UAS-dcr2/Y, hs-hid; +; byn-gal4 UAS-GFP tub-gal80^ts /S-T, Cy, tub-gal80, Hu, Tb stock to kill all male progeny as previously described³⁰. 18–20 virgins were then crossed to 10–12 males of either w¹¹¹⁸ as controls or those from stocks listed in the previous section. Virgins from w¹¹¹⁸ UAS-dcr2/Y, hs-hid; UAS-ras^G12V UAS-p53^RNAi UAS-pten^RNAi UAS-apc^RNAi; byn-gal4 UAS-GFP tub-gal80^ts /S-T, Cy, tub-gal80, Hu, Tb and/or w¹¹¹⁸ UAS-dcr2/Y, hs-hid; +; byn-gal4 UAS-GFP tub-gal80^ts /S-T, Cy, tub-gal80, Hu, Tb were further crossed to males carrying i) UAS-based RNAi or overexpression transgenes, or ii) w; QUAS-rpr; srpHemo-QF2/S-T, Cy, tub-gal80, Hu, Tb a to generate experimental animals. Crosses were kept at 18°C to keep transgenes silent during embryonic development and prevent early larval lethality. After 3 days of egg-laying, parents were removed, and progeny were kept at 18°C for an additional 3 days to allow larval development. The transgenes were then induced by a temperature shift to 29°C for 3 days before dissection.

METHOD DETAILS

Immunohistochemistry and imaging:

Hindguts from experimental and control larvae, identified by the absence of the Tubby (Tb) marker, were dissected and processed as previously described³⁰. Dissections of 10–12 hindguts/genotype were performed in 1X PBS. Dissected tissue fixed with 4% paraformaldehyde, and processed for immunohistochemistry using our standard protocols^28,30. The following primary antibodies were used: mouse anti-NimC1^59,86(Gifted by Eva Kurucz, 1:30), mouse anti-Mmp1(1:1000, DSHB #3B8D12), mouse anti-phospho JNK (1:50, Cell Signaling Technology #9255), mouse anti-phospho AKT (1:1000, Cell Signaling Technology #4054), mouse anti-Histone 2A gamma variant, phosphorylated (1:100, DSHB #UNC93–5.2.1), rabbit anti-Histone H3 (tri methyl K9) (1:50, Abcam #ab8898), mouse anti-Dacapo (1:10, DSHB #NP1), rabbit anti-Dcp1(1:1000, Cell Signaling Technology #9578), mouse anti-β-gal (1:100, DSHB #40–1a), rabbit anti-β-gal (1:200, Thermo Fisher Scientific #A111–32). Alexa-conjugated goat-anti-mouse and anti-rabbit antibodies were used as secondary antibodies (1:1000, Thermo Fisher Scientific, #A-11031, #A-21052, #A-110356, #A-21071). All guts were imaged using SPE DM6 Leica Confocal Microscope at 40X magnification at 1.0 Zoom. Hindguts used to quantify the imaginal ring proliferation area were imaged at 10X magnification, 1.5X Zoom³⁰.

Drug Administration:

Drug concentrations used in this study are as follows: navitoclax (10uM Selleck Chemicals #S1001), fisetin (10uM, Selleck Chemicals #S2298), UC2288 (10uM, Millipore Sigma/ Sigma-Aldrich #532813). Drugs were diluted in Bloomington semi-defined media to achieve the indicated final concentrations in the food along with 0.1% dimethyl sulfoxide (DMSO) as previously described²⁹. Food with 0.1% DMSO alone was used as a vehicle-only control.

QUANTIFICATION AND STATISTICAL ANALYSIS

Scoring and Quantification:

Crosses for survival studies were set up in triplicates with 18–20 virgin females of KRAS TP53 PTEN APC crossed to 10–12 males at 29°C and parents were removed after 2 days of egg laying. KRAS TP53 PTEN APC virgins crossed to w¹¹¹⁸ males were used as baseline controls. The progeny were allowed to develop for the next 10 days at 29°C and survival to pupal stage was calculated by counting the number of experimental and control pupae as previously described³⁰. The size of the anterior hindgut area was measured using Image J³⁰. Cell number quantifications were performed on Leica SPE DM6 confocal microscope at 40X magnification (8–10 hindguts/genotype). Statistical analysis was done using PRISM software.

Statistical analysis:

Details of statistical analysis can be found in relevant sections of methods section and figure legends. Survival studies were performed in triplicate with >50 animals/biological replicate. Immunohistochemical analyses and quantifications of imaging data were performed using 8–10 hindguts/genotype. Survival studies were performed at least three times and immunohistochemical analyses were performed at least twice to independently confirm the results.

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Mouse anti-phospho-SAPK/JNK-pThr183/pTyr185 G9	Cell Signaling Technology	Cat #: 9255 RRID: AB_2307321
Mouse anti-NimC1	Gifted by Dr. Eva Kurucz	Kurucz et al.59
Rabbit anti-phospho-Histone-H3-pSer10	Sigma Aldrich	Cat #: H0412 RRID: AB_477043
Rabbit anti-phospho-AKT-pSer505	Cell Signaling Technology	Cat #: 4054 RRID: AB_331414
Mouse anti-Histone 2A gamma variant, phosphorylated	Developmental Studies Hybridoma Bank	Cat #: UNC93-5.2.1 RRID: AB_2618077
Mouse anti-MMP1	Developmental Studies Hybridoma Bank	Cat #: 3B8D12 RRID: AB_579781
Rabbit anti-beta galactosidase	ThermoFisher Scientific	Cat #: A111-32 RRID: AB_221539
Mouse anti-beta galactosidase	Developmental Studies Hybridoma Bank	Cat #: 40-1a RRID: AB_528100
Rabbit anti- Histone H3 (tri methyl K9)	Abcam	Cat#: ab8898 RRID: AB_306848
Mouse anti-Dacapo	Developmental Studies Hybridoma Bank	Cat#: NP1 RRID: AB_10805540
Goat anti-Mouse IgG Secondary Antibody, Alexa Fluor 568	Invitrogen by ThermoFisher Scientific	Cat #: A-11031 RRID: AB_144696
Goat anti-Mouse IgG Secondary Antibody, Alexa Fluor 633	Invitrogen by ThermoFisher Scientific	Cat #: A-21052 RRID: AB_2535719
Goat anti-Rabbit IgG Secondary Antibody, Alexa Fluor 568	Invitrogen by ThermoFisher Scientific	Cat #: A-11036 RRID: AB_10563566
Goat anti-Rabbit IgG Secondary Antibody, Alexa Fluor 633	Invitrogen by ThermoFisher Scientific	Cat #: A-21071 RRID: AB_2535732
Chemicals, peptides, and recombinant proteins
Phosphate Buffer Saline 10x	ThermoFisher Scientific	Cat #: J75889-k2
Triton X-100	Acros Organics	Cat #: 32737
Normal Goat Serum	Jackson Immunoresearch	Cat #: 005000121
Vecta Shield	Vector Labs	Cat #: H-1200
Paraformaldehyde 16%	ThermoFisher Scientific	Cat #: 28908
DMSO (Dimethyl Sulfoxide) 0.1%	Corning	Cat#: 25-950-CQC
Experimental models: Organisms/strains
D. melanogaster: W1118, dcr2/ Y, hs-hid; byn-Gal4 UAS-GFP tub-80ts; UAS-rasG12V UAS-p53RNAi UAS-ptenRNAi UAS-apcRNAi/S-T, Cy, tub-gal80, Hu, Tb	Datta et al.30	N/A
D. melanogaster: UAS-dap RNAi	Bloomington Drosophila Stock Center	BDSC#: 36720
D. melanogaster: UAS-akt RNAi	Bloomington Drosophila Stock Center	BDSC#: 31701
D. melanogaster: UAS-bsk RNAii II	Bloomington Drosophila Stock Center	BDSC#: 31323
D. melanogaster: UAS-wgn RNAi	Bloomington Drosophila Stock Center	BDSC#: 50594
D. melanogaster: UAS-tefu RNAi	Bloomington Drosophila Stock Center	BDSC#:44073
D. melanogaster: eiger-LacZ	Gifted by Konrad Basler	Muzzopappa et al.65
D. melanogaster: RFP-HP1	Bloomington Drosophila Stock Center	BDSC#: 30562
D. melanogaster: TREdsredT4	Bloomington Drosophila Stock Center	BDSC#: 59012
D. melanogaster: UAS-mmp1	Bloomington Drosophila Stock Center	BDSC#: 58701
D. melanogaster: UAS-p53		Ollmann et al.101
D. melanogaster: UAS-hep.CA	Bloomington Drosophila Stock Center	BDSC#: 58780
D. melanogaster: UAS-akt	Bloomington Drosophila Stock Center	BDSC#: 8191
D. melanogaster: UAS-timp	Bloomington Drosophila Stock Center	BDSC#: 58708
D. melanogaster: w; wg/ Cyo; srpHemo-QF2/(Tm3, Ser)	Bloomington Drosophila Stock Center	BDSC#: 78365
D. melanogaster: w; QUAS-rpr; MKRS/Tm6b, Tb	Bloomington Drosophila Stock Center	BDSC#: 92782
Software and algorithms
Image J	U. S. National Institutes of Health	ImageJ2: 2.3.0/1.53t
Prism 9.5.0	Graphpad	https://www.graphpad.com/
Other: Drugs Administration
Navitoclax (10uM)	Selleck Chemicals	Cat #: S1001
Fisetin (10uM)	Selleck Chemicals	Cat #: S2298
UC2288 (10uM)	Millipore Sigma/Sigma Aldrich	Cat #: 532813

Highlights.

• Senescent cells promote tumorigenesis in the Drosophila larval hindgut.

• Crosstalk among AKT, DDR, and JNK pathways drives senescence

• Senescent cells cooperate with hemocytes they recruit to drive tumorigenesis

• Hemocytes play a context-dependent, pro-tumorigenic role via biphasic JNK activation