Alcohol Interacts with Genetic Alteration of the Hippo Tumor Suppressor Pathway to Modulate Tissue Growth in Drosophila

Alcohol-mediated cancers represent more than 3.5% of cancer-related deaths, yet how alcohol promotes cancer is a major open question. Using Drosophila, we identified novel interactions between dietary ethanol and loss of tumor suppressor components of the Hippo Pathway. The Hippo Pathway suppresses tumors in flies and mammals by inactivating transcriptional co-activator Yorkie, and the spectrum of cancers associated with impaired Hippo signaling overlaps strikingly with those associated with alcohol. Therefore, our findings may implicate loss of Hippo Pathway tumor suppression in alcohol-mediated cancers. Ethanol enhanced overgrowth from loss of the expanded, hippo, or warts tumor suppressors but, surprisingly, not from over-expressing the yorkie oncogene. We propose that in parallel to Yorkie-dependent overgrowth, impairing Hippo signaling in the presence of alcohol may promote overgrowth via additional alcohol-relevant targets. We also identified interactions between alcohol and Hippo Pathway over-activation. We propose that exceeding certain thresholds of alcohol exposure activates Hippo signaling to maintain proper growth control and prevent alcohol-mediated mis-patterning and tissue overgrowth.


Introduction
Alcohol consumption is a significant risk factor in cancers of tissues that contact or metabolize alcohol (for example, upper aerodigestive tract and liver cancers) and also in other tissues such as breast cancer [1][2][3][4][5][6][7][8][9][10][11][12][13][14][15][16]. Alcohol-associated cancers are responsible for more than 3.5% of all cancer deaths, yet how alcohol causes cancer remains a major open question. Drosophila are amenable to rigorous functional genetic analysis and descriptive phenotypic characterizations. Drosophila models have been established to explore the role of alcohol in health contexts including models of fetal alcohol syndrome and alcohol addiction [17][18][19][20][21][22][23] as well as cancerrelevant phenotypes such as tissue overgrowth [24][25][26] in a whole-animal model making it a particularly relevant system to investigate the relationship between alcohol and cancer where systemic responses could underlie the pathogenesis of disease.
We identified novel interactions between dietary ethanol and the Hippo Tumor Suppressor Pathway, a signaling network highly conserved from flies to mammals. The Hippo Pathway acts as a master regulatory pathway to restrict growth and proliferation and to promote apoptosis, and its disruption is implicated in a number of cancers [27][28][29][30][31][32][33][34]. Hippo (Hpo; Mst1 and Mst2 in mammals) [35][36][37][38][39] is the upstream kinase in a core cassette in which activated Hpo kinase associates with Salvador (Sav1 or hWW45 in mammals) [40,41] and phosphorylates and activates downstream effector kinase Warts (Wts; Lats1 and Lats2 in mammals) [40,42,43] and Wts co-activator Mats (Mob1 in mammals) [44]. Wts phosphorylates and inhibits transcriptional co-activator Yorkie (Yki; YAP and TAZ in mammals) [45], a potent oncogene. Components of this core cassette can be regulated by distinct upstream factors to define the eventual biological outputs. For example, GPCR signaling regulates Wts directly [46], Sik kinases regulate Sav [47], and a set of FERM-domain proteins including Merlin (Mer) and Expanded (Ex) act upstream to activate Hpo by an as yet undefined mechanism [48].
We found that alcohol exposure enhanced overgrowth upon Hippo Pathway attenuation in multiple organs in Drosophila. Surprisingly, alcohol did not enhance overgrowth from overexpressing yki, suggesting the Hippo pathway may target ykiindependent growth regulators that are alcohol-responsive. We also found that high doses of alcohol enhanced phenotypes of hpo over-expression. Our studies reveal multiple interactions between alcohol and the Hippo Pathway and suggest a previously undescribed role for Hippo signaling to prevent tissue overgrowth upon alcohol exposure.

Drosophila tools
RNAi was achieved using inverted repeat alleles from the Transgenic RNAi Project for hpo (P{TRiP.JF02740}attP2, referred to here as hpo IRT ), for Mer (P{TRiP.JF02841}attP2, Mer IRT ), and for ex (P{TRiP.JF03120}attP2, ex IRT ), and from the National Institute of Genetics for wts (12072R-2, wts IRN ) and for yki (4005R-2, yki IRN ). Mosaic analysis utilized the FLP/FRT system [49] including the stocks y w eyFLP; FRT42D pW+ UbiGFP/SM6-TM6B and w; FRT42D hpo MGH1 . Over-expression of transgenes used stocks UAS ykiV5, UAS FLAGyki, UAS ykiS168AGFP.HA, and UAS hpo. Screen crosses and other experimental crosses were reared on Formula 4-24 Blue food (Carolina Biological) reconstituted from 1 gram of food flakes per 4 milliliters of the specified doses of ethanol or water. Previous studies established that alcohol doses in food decline over time [50]; we exploited this to try to recapitulate the human scenario of periodic alcohol exposure, not constant alcohol exposure, food in each vial was supplemented with 300 microliters of ethanol at the specified dose every other day.

Drosophila alcohol screen
To identify interactions between alcohol and overgrowth models, or alcohol and alterations in signaling pathways, we utilized both (1) Gal4/UAS [51] and (2) FLP/FRT systems [49] to generate contexts of tissue overgrowth in an ongoing screen (1). We crossed tissue-specific gal4 drivers (including wing drivers engal4, c765gal4, eye driver eygal4, etc.) to UAS transgenes that direct RNAi or over-expression of genes known to cause tissue overgrowth or participate in signaling cascades on food sources containing a range of ethanol or water only. We considered positive hits those genes whose overgrowth phenotypes were enhanced when larvae were reared in the presence of ethanol (2). We crossed eyFLP-containing stocks to corresponding FRT chromosomes containing mutations that cause tissue overgrowth. We considered positive hits those mutations whose over-representation in a mosaic eye was enhanced when larvae were reared in the presence of ethanol.

Quantification
Mosaic eyes were visually scored as: "+" for mild overrepresentation of mutant tissue (white) compared to wild-type tissue (red), "++" for moderate over-representation, and "+++" for strongest over-representation. To rule out unintentional observer bias, trials were conducted blind with lab members scoring eyes without knowledge of genotypes. Adult wings were mounted, photographed, and traced to give pixel area. For engal4 wings, we measured area posterior to vein L4. For c765gal4, total wing area is shown. Area comparisons utilized T-tests; mosaic eye and wing abnormality comparisons utilized chi-squared tests. The data shown in the figures are representative experiments which have been performed independently a minimum of three times.

Ethanol enhances tissue overgrowth associated with loss of the Hippo tumor suppressor
In an ongoing screen to elucidate the link between alcohol and cancer, we used the Gal4/UAS system [51] and the FLP/FRT system of mosaic analysis [49] to modulate levels of conserved growth regulators, tumor suppressors, and oncogenes in various tissues (eye, wing, etc.) in Drosophila exposed to a range of dietary ethanol. Wings from control flies (en>dcr2) reared on food containing 1-10% ethanol generally showed no size increase, although consistent with previous reports [21] there were small but statistically significant size decreases in some cases ( Figure 1A-B). Early in the wing screen, we identified the hpo tumor suppressor as a strong hit ( Figure 1C-G). RNAi to hpo in the posterior wing increased size of this posterior compartment by more than 30% (Figure 1C-D,G). Posterior wing overgrowth was statistically significantly further increased by about 10% in flies reared on food containing 2-5% ethanol ( Figure 1D-G) compared to hpo RNAi alone.
Ethanol doses of 1-10% decreased the percentage of eyes in the mild "+" category and increased the percentage of eyes that were scored as moderate and severe ("++" or "+++") ( Figure  1K). At 5% ethanol, approximately 50% of eyes were severe (compared to less than 15% reared on 0% ethanol); at 7.5% ethanol, this increased to over 60% of eyes.
We find it unlikely that RNAi off-target effects or alcohol regulation of the GAL4/UAS system cause this phenomenon or total wing area (C, E-F) normalized to controls (en>dcr2 for A-B, c765gal4/+ for C, E-F, engal4/+ for D) at 0% ethanol was graphed for each gene and for controls for food containing the indicated doses of ethanol (0%, 2%, and/or 5%) ethanol for (A) Mer RNAi in the posterior wing (* indicates p=1.08025E-05 for females at 5%, p=0.02382138 for males at 5%), (B) ex RNAi in the posterior wing (* indicates p=3.74229E-10 for females at 2%, p=0.005567613 for males at 2%, p=0.000441271 for females at 5%, p=3.00164E-05 for males at 5%), (C) Wts RNAi in the wing (* indicates p=1.64464E-7 for females at 2%, p=8.18059E-7 for females at 5%, p= 8.0633E-11 for males at 5%), (D) ykiV5 over-expression in the posterior wing (* indicates p=0.004919 for females at 2%, p=0.041432 for males at 2%), (E) FLAGyki over-expression in the wing (*indicates p=0.000884273 for control females at 5%), and (F) ykiS168A over-expression in the wing (* indicates p= 0.00038782 for ykiS168A females at 5%; also, in this experiments, * indicates p=0.012339878 for female control wings at 5% compared to control wings at 0%). P values are given for comparisons of the same genotypes reared on the indicated doses of ethanol to those reared on control food lacking ethanol. Experiments in A-E were conducted at 21°C, experiments in C-D were conducted at 25°C, experiments in E were conducted at 18°C, and experiments in F were performed at 16°C.  . Comparing control and Hpo over-expressing wings at 0% ethanol, * indicates p=0.0132429375674621 (females) and p=0.00003467107 (males). Comparing Hpo over-expressing wings at 0% to 5%, * indicates p=0.0000391195433611(females), p=0.03843393 (males). Comparing Hpo over-expressing wings at 0% to 10%, * indicates p=0.0000000059140932 (females) and p=0.000442676 (males). (I) Quantification of weakening and loss of segments within L2, L3, and L4 longitudinal veins. Comparing control and Hpo over-expressing wings at 0% ethanol, * indicates p=2.45092E-22 (females) and p=0.00000659349 (males). Comparing Hpo over-expressing wings at 0% to 5%, * indicates p=0.02424 (males). Comparing Hpo over-expressing wings at 5% to 10%, * indicates p=0.00550172 (females). Comparing Hpo over-expressing wings at 0% to 10%, * indicates p=0.00001108737(females) and p=0.04258  (2) alcohol enhanced over-representation of hpo MGH1 mutant eye tissue ( Figure 1H-K), a genetic system that used neither GAL4/UAS or RNAi tools. Moreover, these data indicate that this phenomenon is not specific to wing tissue but may represent a general response of hpo-deficient tissue to ethanol.
The ability of alcohol to enhance overgrowth was not universal. Over-expressing the oncogene Myc in the wing ( Figure 1Q) promotes wing overgrowth compared to a control wing ( Figure 1P). Dietary ethanol did not enhance this overgrowth ( Figure 1R-S).

Alcohol enhances organ overgrowth associated with expanded and warts but not Merlin or yorkie
To establish if enhancement of wing overgrowth applied to other components of the Hippo Pathway, we examined knockdown of upstream tumor suppressor components Mer and ex. Alcohol did not enhance overgrowth upon Mer RNAi in the posterior wing (Figure 2A), but statistically significantly enhanced overgrowth upon ex RNAi ( Figure 2B). Our data could reflect interaction with alcohol specific for Ex-directed signaling through the pathway or could result from the complicated way in which upstream inputs direct pathway outputs. Signaling through the pathway is complex and not strictly linear; for example, Ex promotes signaling through Hpo [48] and also binds and inhibits Yki directly [52]. Loss of Mer or ex singly has distinct phenotypes from loss of core components hpo or wts [53]. However, simultaneous loss of Mer and ex phenocopies loss of hpo or wts in other contexts [48]. The interaction between alcohol and ex knockdown but not Mer knockdown can be further resolved as future work in the field elucidates how upstream factors activate the pathway to define distinct biological outputs.
To address downstream components, we examined knockdown of wts and over-expression of yki. RNAi to wts in the posterior wing with engal4 led to such overgrowth that wing folding prevented accurate quantification. We therefore used weaker pan-wing driver c765gal4. Ethanol enhanced wing overgrowth from RNAi to wts in females ( Figure 2C).
Over-expressing yki in the posterior wing or across the whole wing promotes tissue overgrowth. We tested conditions that led to overgrowth similar to loss of hpo seen in Figure 1 for wildtype yki transgene UAS-ykiV5. Surprisingly, ethanol did not reproducibly enhance ykiV5-mediated overgrowth ( Figure 2D). Therefore we tested a distinct wild-type yki transgene, UAS-FLAGyki. As with RNAi to wts, overgrowth produced by the FLAGyki transgene using engal4 was too extensive to quantify, so we assessed overgrowth with c765gal4. Consistent with our findings for ykiV5, dietary ethanol did not enhance FLAGykimediated overgrowth ( Figure 2E). Because wild-type versions of yki would be subject to inhibition by endogenous Hippo signaling, we also tested the transgene UAS ykiS168AGFP.HA. The S168A mutation cannot be phosphorylated by Wts at the 14-3-3 site, so is insensitive to Wts-induced inactivation via translocation out of the nucleus [54]. Dietary ethanol did not enhance the overgrowth due to ykiS168AGFP.HA expression ( Figure 2F).
If alcohol-mediated enhancement of overgrowth upon loss of ex, hpo, or wts occurs via interaction with Yki protein or its targets, we would predict alcohol to enhance Yki overexpression-induced overgrowth. Therefore, our genetic interaction studies suggest that alcohol may interact with Hippo signaling at or downstream of hpo and wts, possibly parallel to yki. This is particularly unexpected because signaling through Yki is reported to be crucial to overgrowth from loss of Hippo Pathway tumor suppression. Our findings may reveal a role for Yki-independent Hippo Pathway targets in promoting growth upon alcohol exposure. Alternatively, alcohol may be acting in a less straightforward way to regulate Yki protein not revealed by our wild-type and mutant Yki over-expression studies.

Alcohol enhances Hpo over-expression phenotypes but not Yki loss phenotypes
How does alcohol enhance overgrowth due to loss of ex, hpo, and wts? A trivial explanation would be that alcohol further impairs signaling through the pathway. This explanation would predict that alcohol should therefore suppress the phenotypes of over-activating the pathway, such as by over-expressing hpo itself. In contrast, we observed that dietary ethanol enhanced hpo over-expression phenotypes. Over-expressing hpo in the wing reduced wing size and disrupted wing patterning ( Figure  3B, quantified in 3E, 3H-I) compared to controls ( Figure 3A). Flies reared in 5% ethanol-containing food showed a trend of enhanced wing size reduction while flies reared in 10% ethanol-containing food showed significantly enhanced wing size reduction ( Figure 3C-E). Because knockdown experiments in Figure 1 do not eliminate all Hippo signaling, the increased Hippo pathway activation induced in the presence of 10% ethanol may explain the perplexing result that 10% ethanol doses did not enhance wing overgrowth or as strongly enhance eye mutant tissue over-representation upon Hippo loss as lower doses ( Figure 1E-G, K). Exposure to 5% and 10% ethanol significantly enhanced mis-patterning caused by Hippo over-expression ( Figure 3F-I).
How can we reconcile that alcohol (1) enhanced overgrowth from impaired Hippo Pathway tumor suppression but (2) also enhanced the phenotype of hpo over-expression? Although in seeming conflict with our earlier results, the ability of ethanol to enhance hpo over-expression phenotypes is consistent with previous reports in the literature using in vitro systems. Hpo kinases Mst1 and Mst2 were identified in mammalian systems as stress response kinases [55] and are activated in cultured cells by oxidative stress [56][57][58]. Alcohol promotes oxidative stress; our findings may represent a validation of those in vitro effects in a physiological setting.
Does this enhancement of Hippo-induced wing size reduction occur through further downregulation of yki? If so, we would expect alcohol to also enhance the wing size reduction of knocking down (but not knocking out) yki because presumably the increased Hippo signaling would act to further down-regulate the Yki protein produced. Similar to hpo overexpression, RNAi to yki across the entire wing results in a smaller wing ( Figure 3K-M) compared to a control wing ( Figure  3J). Parallel exposure of flies to doses of 10% ethanol (a dose at which we saw effects on hpo over-expression phenotypes of wing patterning and wing size) did not enhance the wing size reduction due to RNAi of yki ( Figure 3N-P). This may suggest that that alcohol further enhances Hpo-mediated growth regulation through targets other than yki.

A role for Hippo Pathway Tumor Suppression in response to stress?
If Hippo signaling is activated by alcohol as part of a stress response, it seems reasonable to speculate that Hippo signaling is acting to prevent some of the deleterious effects of alcohol. If this is the case, then impairing Hippo signaling in the presence of alcohol would mean those deleterious effects of alcohol would take place. Taking our findings into account with this logic, we propose that in addition to tissue homeostasis regulated by Hippo signaling under normal conditions ( Figure  4A), alcohol does not promote the overgrowth of Drosophila eye and wing tissues in part because alcohol promotes activation of Hippo signaling ( Figure 4B). However, in contexts where Hippo signaling is impaired and tissue undergoes established Yki-dependent overgrowth ( Figure 4C), we propose alcohol can then also act to promote additional tissue overgrowth, possibly by interacting with Yki-independent targets of Hippo signaling ( Figure 4D) because yki overexpression is not sufficient to recapitulate this phenomenon.
The ability of alcohol to promote Hippo signaling in flies and for oxidative stress to promote Hippo signaling in mammalian systems [55][56][57][58] suggests there may be a highly conserved role for stress to activate the Hippo Pathway. This raises interesting questions regarding the origins of Hippo signaling to respond to alcohol. We speculate that activation of Hippo signaling by alcohol developed in a common ancestor as a mechanism to maintain proper patterning and growth regulation even upon exposure to environmental stressors; maintenance of this mechanism may have been selected for in species like Drosophila that develop in the presence of alcohol, such as on fermenting fruit or other species that encounter environmental exposure to related stressors, to prevent stress-mediated tissue overgrowth.
Our findings that alcohol enhances both Hippo Pathway loss of function and over-expression phenotypes are consistent with a role for dietary alcohol (i.e. whole animal exposure) to have specific effects on target organs undergoing Hippo Pathway modulation. This could be particularly relevant in prescribing lifestyle changes and for designing therapies for patients depending on whether or not their tumors maintain intact Hippo signaling. Moreover, our findings suggest that impaired Hippo Pathway tumor suppression may underlie the pathogenesis of specific alcohol-mediated cancers. Notably, there is striking overlap between Hippo-associated cancers [27][28][29][30][31][32][33][34] and alcohol-associated cancers, including colorectal, liver, and breast cancers [59][60][61][62][63][64].
Yki is the best-characterized target of the Hippo Pathway; its homologs YAP and TAZ are widely accepted to play an important role in cancer. Therefore, we were surprised that alcohol did not enhance tissue overgrowth upon yki overexpression. Alcohol may act in a novel way to promote Yki accumulation or activity. Indeed, alcohol promoted Hippo signaling which normally antagonizes yki; this could have As established, when Hippo signaling is intact (A), Yki is properly regulated and tissues maintain proper growth control. We propose that alcohol promotes activation of the pathway at or upstream of Hippo to target alcohol-relevant growth-promoting activities (B) to maintain proper growth control upon alcohol exposure. When Hippo signaling is impaired (C), Yki becomes overactive and promotes tissue overgrowth. We propose that when Hippo signaling is impaired in the presence of alcohol (D), Yki acts to promote overgrowth as established, and additional alcoholrelevant targets of the pathway act to promote further overgrowth. doi: 10.1371/journal.pone.0078880.g004 masked alcohol-mediated enhancement of wild-type yki phenotypes and resulted in the lack of observed interaction between dietary ethanol and wild-type yki over-expression ( Figure 2D). However (1), we did not observe any significant phenotypic enhancement of over-expressing hpo at dietary alcohol concentrations of 2% (not shown), where we did not see interactions between yki over-expression and ethanol and (2) alcohol did not enhance the organ size reduction caused by knockdown of yki. Therefore, we believe the simplest model to explain our data is that in the range of alcohol concentrations tested in our study, alcohol interacts with Yki-independent targets of Hpo and/or Wts. Our report may represent another context of Hippo Pathway functions that are Yki-independent including reported roles in F-actin regulation [65] and polarity [66]. Moreover, controversial reports propose YAP acts as a breast cancer tumor suppressor [67][68][69]. Thus, Hippo Pathway targets parallel to YAP could be particularly relevant to alcoholmediated breast cancers.