Stem Cell

Inter-cell type interactions that control JNK signaling in the Drosophila intestine – Nature Communications

Time Stamp: June 27, 2024 8:00 PM
Source Node: 625224

Eiger is induced in progenitor cells by aging and acute gut damage

To investigate the function of Egr, we first examined its expression in adult fly midguts using an endogenous GFP-tagged Egr (a MiMIC line²⁹). In adults 1 day after eclosion, the expression of Egr-GFP was barely detected (Fig. 1a, a’). However, at 10 and 20 days after eclosion, a progressive increase in Egr-GFP was observed in gut progenitor cells labeled by esg-lacZ (Fig. 1b–d’). Further, we found that Egr could be markedly induced in progenitors in 1-day-old flies by enteric infection with P.e. (Fig. 1e, e’). Because P.e. infection causes rapid cell turnover in the gut, we also observed faint expression of Egr in newborn ECs (nECs) in this condition (Fig.1f, f’, Pdm1⁺ cells), suggesting that low levels of Egr-GFP were inherited from EBs. These data indicate that Egr expression is specific to progenitor cells, and is induced by aging and gut damage, consistent with previous observations that JNK activity is upregulated by aging and gut damage^18,19,20.

JNK activation is restricted to differentiated enterocytes and enteroendocrine cells

Given the inducible, progenitor-specific expression of Egr, we investigated the activation pattern of JNK signaling in the gut using the puc-lacZ^E69 reporter, a widely used downstream target of JNK^18,19,20. Under normal conditions at 8 days after eclosion, puc-lacZ^E69 showed weak expression in large cells (Fig. 2a, white arrowheads) that did not express egr-GFP (Fig. 2a, yellow arrowheads), namely ECs. In flies older than 5 days post eclosion, puc-lacZ^E69 was also observed in enteroendocrine (EE) cells identified by the EE marker, Prospero (Pros) (Fig. 2b, white arrows). We also examined another JNK pathway reporter, TRE-DsRed ³⁰, and found that this was also specifically expressed in ECs (Supplementary Fig. 1, white arrowheads). These data demonstrate that JNK is exclusively activated in ECs and EEs and remains inactive in progenitor cells under normal conditions. Since previous studies reported that JNK could be activated in progenitor cells under stress conditions^18,19, we examined the expression of puc-lacZ^E69 after enteric infection with Ecc15. Although infection could induce low levels of puc-lacZ^E69 expression in progenitors (Fig. 2c, right panels, yellow arrowheads), ECs showed much stronger puc-lacZ^E69 induction in this condition (Fig. 2c, right panels, white arrowheads). These data indicate that JNK activation in progenitors is significantly restricted in both normal and stress conditions. To learn about the mechanisms of cell type-specific JNK activation, we further investigated the functions of Egr and its receptors, Grnd and Wgn.

**Fig. 2: JNK activation is restricted to differentiated ECs and EEs.**

Progenitor cell-produced Eiger signals to Grindelwald in differentiated enterocytes

Previous studies have shown that hemocytes³¹ and fat body cells⁹ can also produce Egr and that those two cell types can actively communicate with the gut^32,33. Therefore, we next evaluated whether gut progenitor cells are the major source of Egr for JNK activation in the gut. RNAi-mediated knockdown of egr either in hemocytes (Hml^ts-driven, Supplementary Fig. 2a) or in fat body (Lpp^ts-driven, Supplementary Fig. 2b) had no repressive effect on damage-induced ISC hyperproliferation. This suggested that gut-specific Egr may play a more important role in JNK activation. Although Egr is produced by gut progenitor cells, overexpression of egr in these cells did not promote ISC proliferation (Fig. 3a, n)³⁴. Also, in contrast to other JNK pathway components (e.g., grnd, Hep^Act, and puc^RNAi), overexpression of egr in ECs (Myo1A^ts-driven) also failed to promote ISC mitoses (Fig. 3d). Hence Egr expression is not sufficient to promote ISC proliferation. However, RNAi-mediated knockdown of egr in progenitors did significantly repress the ISC proliferation caused by gut damage (Fig. 3b), and overexpressed egr increased the mitogenic effect of gut damage (Fig. 3c). These data indicate that Egr is more effective under stress conditions, raising the possibility that stress may somehow enhance the signaling capability of Egr. In addition, although the JNK reporter puc-lacZ^E69 showed activation in EEs (Fig. 2b), overexpressing egr or its receptor grnd in EEs did not have any obvious cell non-autonomous pro-mitotic effects on ISCs (Fig. 3e).

**Fig. 3: Progenitor cell-produced Egr signals to Grnd in differentiated ECs.**

Our investigation next aimed to determine which receptor (Grnd or Wgn) responds to Egr signaling in the gut. Analyses of RNA-sequence (RNA-seq) datasets from our laboratory revealed higher wgn mRNA levels compared to grnd in the adult midgut under normal conditions (Fig. 3f), but after P.e. infection grnd mRNA significantly increased, while wgn mRNA decreased (Fig. 3f). To test the requirements for these receptors, we examined the loss-of-function effects of grnd and wgn in the ISC–EB–EC lineage during stress conditions, when JNK is activated and sufficient to trigger ISC proliferation. RNAi-mediated knockdown of grnd in progenitors (esg^ts-driven), ECs (mex^ts-driven), or the EB–EC lineage (Su(H)GBE^ts-FlipOut-driven), all significantly decreased stress-induced ISC proliferation, whereas wgn depletion in any of these cell types did not show suppressive effects (Fig. 3g, h, Supplementary Fig. 3a–c). Using more specific Gal4 drivers (Dl^ts for ISCs and Su(H)GBE^ts for EBs) to further pinpoint the locus of Grnd function, we found that depletion of grnd in ISCs had no effect (Fig. 3i), whereas depletion in EBs significantly inhibited stress-induced ISC mitoses (Fig. 3j). Knockdown of wgn, however, had no inhibitory effect with either driver (Fig. 3i, j).

Next, we examined the gain-of-function effects of grnd and wgn in the gut. Overexpressing grnd using the esg^ts driver significantly increased ISC proliferation (Fig. 3k). However, overexpression of wgn using the same driver had no obvious pro-mitotic phenotype (Fig. 3l). In fact, ectopic wgn markedly decreased damage-induced ISC hyperproliferation (Fig. 3l). These data emphasize that Grnd, rather than Wgn, serves as the relevant receptor for triggering ISC divisions in the gut stem cell lineage, and suggest that the levels of grnd in EBs and ECs exert profound effects on JNK-dependent stem cell proliferation.

We further detailed the gain-of-function effects of grnd in the ISC–EB–EC lineage using more specific drivers. Consistent with grnd’s loss-of-function effects, overexpression of grnd in ISCs was not pro-mitotic (Fig. 3m), whereas overexpressing grnd in EBs (Fig. 3n), ECs (Fig. 3d), or the EB–EC lineage (Supplementary Fig. 3d, e) elicited strong pro-proliferative effects. These gain-of-function effects align with our observations in the loss-of-function assays and further affirm Grnd’s crucial role in EBs and ECs for JNK activation. Finally, we found that the gain-of-function phenotype of Grnd is Egr-dependent (Fig. 3k, n). Taken together, these data indicate that signaling between Egr, produced by progenitor cells, to Grnd in EBs and ECs, is important for ISC proliferation.

To extend this analysis, we tested Grnd’s ability to activate puc-lacZ^E69. Tellingly, overexpression of grnd in progenitors did not upregulate puc-lacZ^E69 in these cells (Fig. 3o, yellow arrowheads) but preferentially and strongly upregulated puc-lacZ^E69 in ECs (Fig. 3o, white arrowheads). This suggests that Grnd’s ability to activate JNK is confined to ECs and its activity is repressed in ISCs and EBs. However, a question arises regarding why overexpression of grnd in EBs, driven either by esg^ts (Fig. 3k) or Su(H)GBE^ts (Fig. 3n), could induce significant ISC mitoses. One plausible explanation is that, due to the transient nature of EBs, which rapidly differentiate into ECs during gut epithelial regeneration, overexpression of grnd in EBs results in nECs inheriting the EB-overexpressed Grnd. Consequently, the increased JNK activity caused by ectopic Grnd in these nECs may trigger additional mitogenic signaling to stimulate ISC proliferation in a cell non-autonomous manner.

Hyperactivation of JNK signaling causes apoptotic stem cell loss

We then asked why progenitor cells are refractory to JNK activation. It’s well known that overactivation of JNK is strongly pro-apoptotic in many tissues. However, in the fly gut, activation of JNK promotes stem cell mitoses^18,19,20,27. To investigate this, we tested the pro-mitotic effect of JNK activation by overexpressing a constitutively active form of the JNKK Hep, Hep^Act. Consistent with previous observations, overexpressing Hep^Act using Dl^ts, a weak ISC-specific driver, induced high levels of ISC proliferation (Fig. 4a). When we used the stronger esg^ts driver, which expresses in both ISCs and EBs, the pro-mitotic effect of Hep^Act appeared to be absent. However, co-expressing the caspase inhibitory protein P35 with Hep^Act could somewhat restore Hep’s pro-mitotic effect at an early timepoint (1 day; Fig. 4b). But over time (e.g., 3-day overexpression), P35 lost this rescue capability (Fig. 4b). These data suggest that high levels of JNK activity negate ISCs’ mitotic capability, and that this effect is dose-dependent. Further tests using the esg^ts driver showed that Hep^Act-expressing midguts had lost most of their pre-existing GFP⁻ ECs, which were replaced by newborn GFP⁺ ECs (Fig. 4c, d), and that progenitor cell numbers were strongly reduced over time (Fig. 4e). This suggests that Hep^Act expression triggered rapid cell turnover before causing progenitor loss. To validate this, we utilized the Gal4 Technique for Real-time And Clonal Expression (G-TRACE) system³⁵ to track the cell fate of Hep^Act-expressing ISCs.

The G-TRACE cell lineage-tracing system (UAS-RFP, UAS-Flp, Ubi^p63E > FRT-STOP-FRT > nEGFP) was activated by temperature shift using esg^ts. In this method, UAS-RFP and UAS-Flp are induced in progenitors by esg-Gal4. FLP recombinase excises the FRT-STOP-FRT cassette from Ubi^p63E > FRT-STOP-FRT > nEGFP, converting it to a ubiquitously expressed GFP marker, Ubi^p63E-nEGFP. As a result, all progenitor cells are marked strongly with RFP and weakly with GFP (Fig. 4f, yellow arrowheads in upper panels) while their progeny cells are marked with GFP only. Under normal conditions, due to active gut regeneration, nECs inherited some RFP signal from their ISC/EB precursors, thereby exhibiting a mild-RFP/strong-GFP double positive pattern (Fig. 4f, white arrowheads in upper panels). In contrast, older pre-existing ECs (oECs) exhibited an RFP/GFP double negative pattern (Fig. 4f, white arrows in upper panels). Remarkably, in Hep^Act overexpression guts RFP⁺ progenitor cells disappeared and the number of nECs significantly decreased (Fig. 4f, white arrowheads in lower panels). These data confirm that hyperactivation of JNK in progenitors causes stem cell loss and impairs gut regeneration. Indeed, high levels of the apoptosis marker, DCP-1, were evident in the Hep^Act-expressing progenitors (Fig. 4g, yellow arrowheads in right panels). These experiments demonstrate that low levels of JNK activity in ISCs promote ISC proliferation whereas high levels trigger apoptosis (Fig. 4h). This biphasic function provides a rationale for why JNK signaling is restricted to differentiated cells, namely because too much JNK activation in ISCs is cytotoxic.

Additionally, we explored whether JNK activation in ECs can also trigger apoptosis and whether it has dose-dependent functions in these cells too. In contrast to progenitor cells, ECs exhibited hypersensitivity to JNK activation. A brief period (1 day) of Hep^Act expression in ECs induced robust EC apoptosis, as indicated by DCP-1 staining (Supplementary Fig. 4a). However, over time (e.g., 3-day overexpression), there was no significant difference in the degree of EC apoptosis (Supplementary Fig. 4a, b), suggesting that JNK triggers EC apoptosis in a dose-independent manner.

JNK signaling in ISCs is restricted by N-glycosylation

Continuing our exploration, we aimed to decipher the mechanisms responsible for suppressing JNK activation in progenitor cells. A recent study highlighted the role of the Gish/CK1r kinase in restraining JNK activation by phosphorylating and destabilizing Rho1, a JNK activator³⁶. Gish is downregulated in midgut progenitors in older flies, leading to increased JNK activity during aging³⁶. However, knockdown of Gish in progenitors for 3 days did not significantly increase ISC proliferation³⁶, suggesting that other factors are involved. Another recent study reported that the N-glycosylation pathway components ALG3 and ALG9 repress JNK activity in fly imaginal discs¹². ALG3 and ALG9 act as alpha-1,3-mannosyltransferases that modify proteins during the process of N-glycosylation. Among the five mannosyltransferase-encoding genes in Drosophila (Alg2/CG1291, Alg3/CG4084, Alg9/CG11851, Alg11/CG11306, and Alg12/CG8412), our screening showed that RNAi-mediated knockdown of either Alg3 or Alg9 in progenitors increased ISC mitoses (Fig. 5a), implicating these two genes as potential regulators of JNK activity. To discern the cell types in which ALG3 and ALG9 exert their anti-mitotic functions, we employed specific Gal4 drivers – Dl^ts for ISC knockdown, Su(H)GBE^ts for EB knockdown, and Myo1A^ts for EC knockdown. Remarkably, depletion of Alg3 or Alg9 in each cell type resulted in ISC over-proliferation (Fig. 5c–e), indicating that ALG3 and ALG9 repress ISC proliferation, both cell autonomously and non-cell autonomously.

To determine whether ALG3 and/or ALG9 repress ISC proliferation by restricting JNK activity, we examined the induction of puc-lacZ^E69 after Alg3– or Alg9-depletion. Knockdown of either gene using esg^ts markedly induced puc-lacZ^E69 expression in progenitors (Fig. 5b) as well as in newborn (GFP⁺) ECs, confirming that ALG3 and 9 are important suppressors of JNK activity. We next investigated whether the ISC hyperproliferation caused by Alg3– or Alg9-depletion is JNK dependent. Expression of egr^RNAi significantly blocked ISC mitoses caused by Alg3^RNAi (Fig. 5c, d), and expression of Bsk^K53R (a dominant-negative form of Bsk/JNK) significantly suppressed ISC mitoses caused by Alg9^RNAi (Fig. 5c). Interestingly, inactivating JNK signaling in ECs with Bsk^K53R was insufficient to suppress Alg9^RNAi’s pro-mitotic effects (Fig. 5e), suggesting that ALG9 and/or ALG3 may have multiple repression targets in ECs in addition to JNK signaling that affect ISC proliferation. These data indicate that ALG3 and ALG9 repress ISC proliferation, at least in part, by inhibiting JNK activation. Since puc-lacZ^E69 induction could be detected in progenitor cells of Ecc15-infected guts (Fig. 2c), we asked whether ALG3 and ALG9 are regulated by gut damage. qRT-PCR analyses on fluorescence-activated cell sorting (FACS)-sorted progenitor cells (GFP⁺), showed that both Alg3 and Alg9 mRNA levels were significantly reduced in progenitors by enteric Ecc15 or P.e. infection (Fig. 5f). Consistent with this, our previous cell type-specific RNA-seq data revealed that Alg3 and Alg9 are expressed in all midgut cell types and that both are markedly decreased in EBs by P.e. infection³⁷. Therefore, we suggest that the downregulation of Alg3 and Alg9 could activate JNK during the gut damage response. In line with this, overexpressing Alg3 in progenitors significantly suppressed damage-induced ISC hyperproliferation (Fig. 5g), confirming that Alg genes can act to restrain the proliferative response during infection stress.

Beyond the Alg genes, we also investigated whether N-glycosylation in general is important for regulating ISC proliferation. We found that progenitor-specific overexpression of Pngl, the Drosophila ortholog of peptide N-glycosidase (PNGase) that catalyzes the de-glycosylation of glycoproteins, resulted in increased ISC proliferation (Fig. 5h), whereas overexpression of Pngl^C303A, a mutant form of Pngl that lacks the enzymatic activity³⁸, was not pro-mitotic (Fig. 5h). Similarly, overexpression of mNgly1, the mouse PNGase ortholog, also promoted ISC proliferation (Fig. 5h). These results support our proposal that the N-glycosylation normally restricts ISC proliferation.

We next explored the potential mechanisms by which ALG-mediated N-glycosylation restricts ISC proliferation. Using Drosophila wing imaginal discs, de Vreede et al. found that Grnd is N-glycosylated at asparagine 63 (N63) by ALGs and that this modification limits Grnd’s binding to Egr¹². As shown above, overexpression of wild-type grnd in ISCs did not promote ISC proliferation (Fig. 3m). However, overexpression of grnd^N63A, a mutant form of grnd that is refractory to N-glycosylation, did stimulate ISC proliferation under precisely the same conditions (Fig. 5i). Furthermore, western blot data showed that progenitor-expressed wild-type Grnd has a higher molecular weight than Grnd^N63A (Fig. 5j), consistent with Grnd being N-glycosylated at N63 in the fly midgut. In addition, ectopic expression of Pngl in progenitors activated JNK signaling in these cells, as indicated by puc-lacZ^E69 induction (Fig. 5k). These data corroborate our proposal that N-glycosylation of Grnd normally blocks the Grnd–Egr interaction and thereby inhibits the activation of JNK signaling. Upon infection, the downregulation of Alg3 and Alg9 may impair the N-glycosylation of Grnd, thereby increasing its interaction with Egr to facilitate the activation of JNK in both ECs and progenitor cells (Fig. 5l).

In addition to their role in ISC mitosis, ALG3 and ALG9 also appeared to regulate stem cell numbers in a non-autonomous manner. Notably, our observations revealed that the knockdown of Alg3 in EBs led to a significant increase in stem cell numbers, as evidenced by an increase in the number of cells expressing the ISC marker Dl-lacZ (Supplementary Fig. 5a, b). Previous studies have highlighted that JNK signaling induces symmetric divisions in ISCs by reorienting the mitotic spindles²⁶. Given that the pro-mitotic effect of Alg3 depletion in EBs is reliant on Egr (Fig. 5d), we suggest that ALG3 and ALG9 in EBs may restrain ISC symmetric divisions via a non-cell-autonomous signaling involving the JNK pathway.

JNK activation in ECs stimulates EGFR signaling in ISCs via Rhomboid

We next focused on how the JNK ligand Egr is regulated in the midgut. As shown above (Fig. 1), Egr expression is induced specifically in progenitors by aging and stress. To test whether this induction might be a secondary effect of ISC proliferation, we used an RNAi against String (Stg), a Cdc25 homolog required for ISC mitoses^39,40. As shown in Supplementary Fig. 6a, b, bacterial infection induced a striking upregulation of egr and massive ISC proliferation. RNAi-mediated knockdown of stg totally blocked P.e.-induced ISC mitoses (Supplementary Fig. 6b), but did not block the induction of egr (Supplementary Fig. 6a, c), indicating that the induction of Egr is not a consequence of ISC proliferation, and may be more directly controlled by signaling. As ECs serve as the primary sensors of damage⁴¹, we first checked whether stress-dependent upregulation of Egr in progenitors correlates with mitogen induction in ECs. During stress-induced gut damage responses, two major signaling proteins, Yorkie (Yki, the downstream transcription activator of the Hippo pathway) and JNK, are turned on in ECs to generate mitogens (e.g., the cytokines Upd3, Upd2, and the neuregulin Krn, etc.) that support ISC proliferation⁴¹. Interestingly, overexpression of yki in ECs strongly induced Upd3 in ECs, but failed to induce egr expression in progenitors (Fig. 6a). Furthermore, whereas activation of JNK signaling in ECs by puc^RNAi strongly upregulated both Upd3 in ECs and egr in progenitors (Fig. 6a), direct overexpression of upd3 in ECs did not promote egr induction in progenitors (Fig. 6a). This suggested that signal(s) other than the Upd cytokines are involved. Our previous work indicated that Rho, an intermembrane protease that promotes the cleavage and secretion of EGFR ligands⁴², could be upregulated in ECs by Hep^Act⁴³. Tellingly, overexpression of grnd or puc^RNAi in ECs induced rho expression in these cells (Fig. 6b). In addition, ISC hyperproliferation caused by grnd overexpression in ECs was repressed by rho depletion (Fig. 6c). These data suggest that Rho is a downstream target of Grnd/JNK signaling. Consistently, overexpression of rho in ECs induced a striking upregulation of egr in progenitors (Fig. 6a). These findings indicate that the Grnd-JNK-Rho signaling in ECs plays a crucial role in regulating egr expression in progenitor cells.

**Fig. 6: JNK activation in ECs upregulates Rho.**

Next, we tested the interaction of Rho with EGFR ligands. Four activating EGFR ligands (Spi, Krn, Gurken/Grk, and Vein/Vn) and one inhibitory ligand (Argos) have been identified in Drosophila⁴⁴. Three of these (Spi, Krn, and Grk) are produced as transmembrane precursors and require cleavage by Rhomboid proteases to activate their secretion⁴⁴. Grk has not been detected in the midgut⁴⁵, and Vn is produced in gut visceral muscle as a secreted protein that does not require Rho for activation⁴³. In gain-of-function tests, overexpression of rho in ECs strongly induced ISC proliferation (Fig. 6d), had a weaker effect when expressed in ISCs (Supplementary Fig. 7a), and had no significant pro-mitotic activity when overexpressed in EBs or EEs (Supplementary Fig. 7b, c). Epistasis tests in which we co-expressed rho together with Krn^RNAi and/or spi^RNAi showed that Rho-driven ISC mitoses depended on both ligands, and that Rho potentially acts in both ECs and progenitor cells (Fig. 6e, f). These results support a model which Grnd-JNK-Rho- Krn/Spi signaling in ECs controls the activation of EGFR, and proliferation, in progenitor cells.

Eiger is a downstream target of EGFR-MAPK signaling

In further tests, we found that overexpressing either Krn or a secreted variant of Spi (sSpi) in ECs was sufficient to strongly induce egr expression in progenitor cells (Fig. 6g, h). Given that EB- and EC-derived Spi and Krn activate EGFR/MAPK signaling in progenitor cells^40,43,45,46, we next examined whether egr is a target of the EGFR/MAPK pathway. Artificially activating MAPK/ERK signaling in progenitors by expressing Ras^V12S35 or Raf^GOF strongly induced egr expression (Fig. 7a). Conversely, depletion of Egfr in progenitors totally blocked P.e.-induced egr induction (Fig.7c) as well as ISC hyperproliferation (Fig. 7b). Similarly, knockdown of Ras in progenitors totally blocked egr induction by Ecc15 infection (Fig. 7d). These data indicate that Egr is a downstream target of EGFR/Ras/Raf signaling in progenitor cells. When considered together with our discovery that progenitor-derived Egr signals to Grnd in ECs, this implies that signaling between progenitor cells and ECs forms an Egr→Grnd→JNK→Rho→Krn/Spi→EGFR→Egr feedforward loop capable of maintaining and potentially amplifying JNK activity in ECs, and pro-proliferative EGFR signaling in progenitor cells (Fig. 7e).

**Fig. 7: Egr is a downstream target of EGFR-MAPK signaling.**

Finally, we used a unique experimental condition to further validate this signaling interaction, by testing a condition in which Egr was expressed specifically in progenitors and Grnd was coincidently suppressed specifically in ECs. In a previous study, we identified an SH3PX1-dependent autophagy-endocytosis network that specifically controls EGFR degradation and activity in progenitor cells of fly midgut⁴⁰. Homozygous SH3PX1 mutant flies exhibit exclusive activation of EGFR in progenitors, leading to ISC hyperproliferation, while loss-of-function of SH3PX1 in ECs and EEs has no impact on EGFR activity and ISC mitosis⁴⁰. Thus the SH3PX1 mutant effectively mimics specific activation of EGFR in progenitors. This is likely because EGFR is expressed preferentially in progenitors^40,47, such that blocking its degradation using the SH3PX1 mutation causes EGFR accumulation and activation only in these cells. As expected, SH3PX1^d1/10A mutants specifically induced egr expression in progenitors (Supplementary Fig. 8a), and concomitant knockdown of egr in these progenitors suppressed SH3PX1^RNAi-induced ISC hyperproliferation (Supplementary Fig. 8b), confirming egr as the downstream target of SH3PX1. Hence, the SH3PX1 mutant mimics the overexpression of Egr in progenitors. Importantly, depleting grnd in ECs effectively blocked the pro-mitotic phenotype in the SH3PX1^d1/HK62b mutants (Supplementary Fig. 8c), consistent with our proposal that progenitor-derived Egr signals to Grnd in ECs to indirectly promote ISC proliferation. Altogether, this study uncovers a paracrine JNK-EGFR-JNK feedforward loop that sustains ISC proliferation during stress-induced gut regeneration, with ALG3 and ALG9 serving as required brakes to this pro-proliferative loop (Fig. 7e).