BNIP3 attenuates hepatocellular carcinoma by promoting lipid droplet turnover at the lysosome.

BNIP3 attenuates hepatocellular carcinoma by promoting lipid droplet turnover at the lysosome. Abstract 31 Hepatic steatosis is a major etiological factor in hepatocellular carcinoma (HCC). Work reported here 32 identifies BNIP3 as a suppressor of HCC that mitigates against lipid accumulation. 33 Loss of BNIP3 decreased tumor latency and increased tumor burden in a mouse model of HCC. This 34 was associated with increased lipid accumulation and elevated HCC tumor cell growth. Conversely, 35 exogenous BNIP3 decreased lipid levels and reduced HCC tumor cell growth. Mutant BNIP3 W18A that is 36 unable to promote mitophagy did not decrease HCC cell growth and was defective at reducing lipid levels. 37 Growth suppression by BNIP3 was not mediated by effects on fatty acid oxidation (FAO) or de novo 38 lipogenesis (DNL). Rather, BNIP3 suppressed HCC cell growth by promoting lipid droplet turnover at the 39 lysosome through a process we have termed “mitolipophagy” in which lipid droplets and mitochondria 40 are turned over together at the lysosome. Low BNIP3 expression in human HCC also correlated with 41 increased lipid content and worse prognosis than HCC expressing high levels of BNIP3. This work reveals 42 a role for BNIP3 and lipid droplet turnover at the lysosome in attenuating HCC. a we observed BNIP3 mitigated against lipid linked to high ACACA BNIP3 high of lipid associated with high ACACA 8f). These results indicate that high BNIP3 expression is associated with decreased lipid accumulation, less aggressive HCC and increased patient survival when linked to high ACACA expression while conversely low BNIP3 predicts increased lipid accumulation and worse patient outcome.


75
Loss of Bnip3 reduces HCC tumor latency and promotes HCC tumor growth.

76
To examine the role of BNIP3 in hepatocellular carcinoma (HCC), we injected 15 day old Bnip3 +/+ 77 and bnip3 -/mice with the chemical carcinogen diethylnitrosamine (DEN) that is known to induce HCC 78 with predicted latency in laboratory mice 23,24 . We harvested mice at 24, 32 and 40 weeks of age to assess 79 the effect of Bnip3 loss on both latency and growth of HCC. At 24 weeks of age, macroscopic liver lesions 80 were apparent on the surface of the bnip3 -/liver (Fig. 1a, top right, arrows) but not on the Bnip3 +/+ liver.

81
By 40 weeks of age, large tumors were obvious on both the Bnip3 +/+ and bnip3 -/livers, although the 82 lesions on the bnip3 -/liver were visibly larger than those in Bnip3 +/+ liver (Fig. 1a). Quantification of tumor 83 number (total per liver on serial sections through each liver) and tumor size (diameter) supported the 84 visual assessment that tumors form earlier and grow faster in DEN-treated bnip3 -/mice than wild-type 85 mice. Specifically, by 24 weeks of age, there were significantly more tumors detected in the DEN-treated 86 bnip3 -/mice than in DEN-treated wild-type control mice (Fig. 1b) although by 32 and 40 weeks of age, 87 the difference in tumor number was no longer significant (Fig. 1b). By contrast, for those tumors that were 88 detectable at 24 weeks, there was no significant difference in tumor size between wild-type and bnip3 -/-89 mice initially but by 32 and 40 weeks, the bnip3 -/tumors were significantly larger than those forming in 90 wild-type mice (Fig. 1c). In summary, loss of Bnip3 reduces tumor latency and increases tumor growth 91 rate of DEN-induced HCC in mice.

92
Interestingly, when we examined BNIP3 expression by immunohistochemistry in tumors forming 93 in wild-type mice, we noted that BNIP3 expression is elevated in HCC tumors (T) at 24 weeks compared 94 to adjacent normal (N) liver (Fig. 1d). By contrast, BNIP3 levels were lower in HCC tumors (T) compared 95 to adjacent normal liver (N) at 32 weeks and 40 weeks (Fig. 1d). The upregulation of BNIP3 in HCC 96 tumors at 24 weeks compared to adjacent normal liver, is likely mediated by elevated nuclear Hif1a 97 expression which is also elevated in HCC tumors at 24 weeks (Supp. Fig. 1). qPCR showed that Bnip3 98 mRNA isolated from primary tumor and adjacent normal tissue was down-regulated in tumors at 32 weeks 99 and 40 weeks of age (Fig. 1e) suggesting that loss of BNIP3 expression in wild-type HCC at later 100 timepoints is likely mediated by gene silencing since Bnip3 mRNA levels are decreased (Fig. 1e) despite In addition to loss of BNIP3 expression in wild-type tumors at 32 to 40 weeks of age, there was also a 109 change in the appearance of HCC tumors forming in wild-type mice. At 24 weeks of age, wild-type HCCs 110 contain small, tightly packed tumor cells, whereas by 32 and 40 weeks of age, HCC tumors contained 111 larger tumor cells with more "bubbles" suggesting increased lipid accumulation (Fig. 1d). To assess this, 112 we stained liver sections from wild-type and bnip3 -/-DEN treated mice with Oil red O (ORO) to determine 113 lipid content in tumors at 24, 32 and 40 weeks (Fig. 2a-c). Consistent with the pattern detected by H&E 114 staining ( Fig. 1d; Fig. 2a, top left), we observed that wild-type tumors at 24 weeks contained less lipid 115 compared to surrounding normal tissue (Fig. 2a, middle panel). By contrast, HCC tumors forming in bnip3 -116 /livers contained considerably more lipid at 24 weeks than did tumors in wild-type mice (Fig. 2a) which 117 was apparent by both H&E staining and by ORO staining. However, tumors growing in 40 week old wild-118 type mouse liver that had lost Bnip3 expression ( Fig. 1d-f), exhibited higher lipid content than surrounding 119 normal liver and as high as that detected in bnip3 -/tumors (Fig. 2b). Quantification of ORO staining 120 confirmed these observations (Fig. 2c). At 24 weeks, there was increased lipid content in bnip3 -/tumors 121 compared to wild-type, but these differences in lipid content diminished over time at 32 and 40 weeks, 122 as wild-type tumors lost Bnip3 expression and simultaneously accumulated more lipid (Fig. 2c). Increased 123 lipid content in bnip3 -/-HCC was associated with increased transcript levels of genes involved in fatty 124 acid synthesis (Fig. 2d), including fatty acid synthase (Fasn), acetyl CoA carboxylase 1 (Acaca), ATP 125 citrate lyase (Acly) and Stearoyl CoA desaturase-1 (Scd1). Immunohistochemistry for ACACA and FASN 126 indicated that HCC tumors at 24 weeks in both Bnip3 +/+ and bnip3 -/mice express higher levels of these 127 enzymes than surrounding normal tissue (Fig. 2e-f). In summary, our findings show that loss of Bnip3 128 either by genetic deletion in bnip3 -/mice, or via silencing at late stages of tumorigenesis in wild-type mice whereas EV and HA-BNIP3 W18A did not (Fig. 3b). Decreased mitochondrial staining was visible in HA-

143
BNIP3 expressing HCC cells in the presence (Fig. 3b) or absence of bafilomycin A1 (Supp. Fig. 2a), but 144 the overlap between LC3 and TOMM20 could only be detected in the presence of bafilomycin A1 (Fig.   145 3b) indicating that there was increased mitophagic flux when HA-BNIP3 was expressed but not with EV 146 or HA-BNIP3 W18A expression. We also noted that HA-BNIP3 induced mitochondrial fragmentation, that 147 was most striking when cells were bafilomycin A1 treated (Fig. 3b), but not in the absence of bafilomycin 148 A1 (Supp. Fig. 2a). Together these results indicate that BNIP3 promotes mitochondrial fragmentation and 149 preferential mitophagy of these fragmented mitochondria at the autolysosome, which is dependent on its 150 interaction with LC3, since this was not observed in either EV or HA-BNIP3 W18A expressing HCC cells.

151
Mitochondrial mass was also indirectly assessed by qPCR for mitochondrial genome copy 152 number (amplifying for Nd1 and Cytb) relative to nuclear DNA (amplifying beta-hemoglobin/Hbb) and 153 shown to be decreased in HCC cells expressing HA-BNIP3 but not EV or HA-BNIP3 W18A (Fig. 3c). Also, 154 consistent with decreased mitochondrial mass, we observed decreased oxygen consumption in HCC 155 cells expressing HA-BNIP3 but not EV or HA-BNIP3 W18A (Fig. 3d). Interestingly, we observed that 156 exogenous HA-BNIP3 decreased extracellular acidification of growth media following a glycolysis stress 157 test performed using the Seahorse analyzer (Fig. 3e) and suppressed glucose uptake by HCC cells in 158 culture (Fig. 3f). Consistent with these findings in vitro and with increased growth rate of HCC lacking 159 Bnip3, there was increased glucose uptake into HCCs in bnip3 -/liver compared to Bnip3 +/+ liver as 160 measured by FDG-PET analysis of live mice at 32 weeks of age (Supp. Fig. 3). Overall, these results

161
show that exogenous HA-BNIP3, but not EV or HA-BNIP3 W18A , promotes mitophagy and reduces 162 mitochondrial mass in HCC cells. In addition, BNIP3 reduces reliance on glucose to fuel either OXPHOS 163 or glycolysis.

164
Given the steatotic phenotype in vivo in HCC tumors growing in the absence of BNIP3, we 165 examined lipid content of HCC cells in culture expressing EV, HA-BNIP3 or HA-BNIP3 W18A by staining 166 cells with the lipophilic dye, BODIPY 493/503. We observed that expressing exogenous HA-BNIP3 167 decreased the number of BODIPY positive lipid droplets in HCC cells compared to EV control expressing 168 HCC cells (Fig. 4a -4c). Intriguingly, even though HA-BNIP3 W18A was unable to promote mitophagy, we 169 observed that this mutant form of BNIP3 retained ability to decrease lipid droplet number compared to 170 EV ( Fig. 4a-b). However, when we challenged these cells with oleic acid to further increase cellular lipid 171 content, HA-BNIP3 was significantly more effective at decreasing lipid droplet number than either EV or 172 HA-BNIP3 W18A (Fig. 4a, 4c). Nevertheless, HA-BNIP3 W18A retained partial ability to decrease lipid droplet 173 number suggesting that while mitophagy is involved in how BNIP3 limits lipid accumulation, a second mitophagy-independent role is also at play. We noted that while HA-BNIP3 W18A is defective for mitophagy,

175
it retains the ability to induce mitochondrial fragmentation raising the possibility that mitochondrial fragmentation contributes to how BNIP3 prevents lipid accumulation. Analysis of BODIPY staining of 177 HCC cells by Imagestream flow cytometric analysis confirmed results obtained by immunofluorescence 178 microscopy in showing decreased lipid droplet numbers when HA-BNIP3 was expressed but not EV, and 179 less so with HA-BNIP3 W18A (Fig. 4d, 4e). Imagestream analysis also showed HCC cells expressing HA-180 BNIP3 to be smaller in size that HCC cells expressing EV or HA-BNIP3 W18A (Fig. 4d, 4f). Previous work 181 has linked lower mitochondrial mass and function to decreased overall cell size 25,26 .

194
Given that exogenous HA-BNIP3 decreased mitochondrial mass (Fig. 3b, 3c) and lowered oxidation of 195 glucose in bnip3 -/cells (Fig. 3d), we were surprised to observe that HA-BNIP3 markedly increased FAO 196 of palmitate in bnip3 -/-HCC cells compared to EV or HA-BNIP3 W18A (Fig. 5a). HA-BNIP3 expressing HCC 197 cells were also more sensitive to etomoxir (ETO) which suppresses carnitine palmitoyl transferase-1 198 (CPT1) and blocks fatty acid uptake into the mitochondria to suppress FAO (Fig. 5b, Fig. 5d). Indeed, 199 ETO collapsed oxygen consumption by HA-BNIP3 expressing HCC cells using palmitate as substrate, 200 down to levels seen in HCC cells expressing EV or HA-BNIP3 W18A (Fig. 5b, 5d). By contrast ETO had no 201 effect on HCC cells expressing EV or HA-BNIP3 W18A (Fig. 5c, 5e). These results suggest that HA-BNIP3 202 is promoting FAO and thus we were interested to determine if this explained how HA-BNIP3 was able to 203 promote lipid turnover in HCC cells.

204
Using similar approaches to those described above, we examined whether inhibiting FAO with 205 ETO inhibited the ability of HA-BNIP3 to decrease numbers of BODIPY-positive lipid droplets in HCC 206 cells. Surprisingly, even in the presence of ETO that clearly disrupted FAO, HA-BNIP3 retained the ability 207 to decrease lipid droplet content in HCC cells, even when FAO was inhibited (Fig. 5f, 5g). This ability was 208 also still evident in HCC cells fed oleic acid in the presence of ETO (Fig. 5f, 5h), suggesting that while in HCC cells. Significantly, HA-BNIP3 W18A was also able to decrease lipid droplet number in the presence 211 of Etomoxir (Fig. 5f, Fig. 5g), even though this form of BNIP3 was unable to promote FAO of palmitate 212 (Fig. 5b, Fig. 5e). Again, this effect of BNIP3 W18A was overcome when cells were fed oleic acid to increase 213 lipid droplet content such that it was clearly less effective than wild-type BNIP3 at promoting lipid droplet 214 clearance (Fig. 5f, Fig. 5g).

215
Expression of certain genes involved in FAO were modestly increased in HCC cells expressing 216 HA-BNIP3 compared to either EV expressing or HA-BNIP3 W18A expressing cells (Fig. 5i)   Indeed, treatment of HA-BNIP3 expressing HCC cells with ETO had no effect on cell growth (Fig. 5j).

239
When we examined the effect of TVB-3664 on lipid content of HCC cells, we observed that it 240 eliminated lipid content almost completely independent of BNIP3 since it decreased lipid content in cells 241 expressing EV, HA-BNIP3 or HA-BNIP3 W18A to undetectable levels ( Fig. 6f, Fig. 6h). However, by inhibited ( Fig. 6g, Fig. 6h), arguing that the ability of BNIP3 to decrease lipid content in HCC cells was 245 independent of effects on DNL and/or FAO. Furthermore, the repressive effect of FASN inhibition on 246 growth of HA-BNIP3 expressing cells was synergistic with the effect of exogenous HA-BNIP3 (Fig. 6i)

247
suggesting that BNIP3 is suppressing tumor cell growth in a manner independent of lipid synthesis. In 248 addition, TVB-3664 inhibition of FASN was also able to inhibit growth of EV and HA-BNIP3 W18A 249 expressing HCC cells (Fig. 6i), consistent with the growth suppressive effects of TVB-3664 being BNIP3-250 independent. In summary, while TVB-3664 inhibits HCC cell growth, this is independent of BNIP3 and

262
BNIP3 in mitophagy, we examined a potential role for BNIP3 in promoting LD turnover via a mechanism 263 that integrates mitophagy with lipophagy. When we inhibited lysosomal lipases with Lalistat2 (LALi) 33 , 264 we observed increased numbers of LDs accumulating in HCC cells when HA-BNIP3 was expressed but 265 less with EV or HA-BNIP3 W18A (Fig. 7a, 7b) suggesting that BNIP3 relied on lysosomal lipases to elicit at 266 least part of its effect in decreasing LD numbers. ImageStream analysis confirmed that LALi only 267 increased BODIPY LD number significantly when HA-BNIP3 was expressed and less so when HA-268 BNIP3 W18A or EV was expressed (Supp. Fig. 4a, 4b). Interestingly, LALi appeared to increase LD numbers 269 but did not increase overall cell size of HCC cells expressing HA-BNIP3, (Supp. Fig. 4a, 4c) indicating 270 that reduced HCC cell size with HA-BNIP3 was not likely due to effects of BNIP3 on LD numbers. In 271 summary, these results suggested to us that BNIP3 was decreasing lipid content in HCC cells by 272 promoting LD turnover at the lysosome.

273
To examine this more carefully, we co-stained cells for BODIPY and lysotracker to examine 274 overlap of LDs with the lysosome. Experiments were performed in the presence of oleic acid and LALi to 275 allow such structures to accumulate. While a few overlapping LD/lysosomes were detected in EV and smaller green LDs were seen predominantly inside magenta-colored lysosomes while larger LDs did not 279 associate with lysosomes (Fig. 7c). Analysis of LDs, mitochondria and lysosomes in the different HCC 280 lines by transmission electron microscopy (TEM), revealed plentiful LDs in both EV and HA-BNIP3 W18A 281 expressing cells (Fig. 7d), and fewer LDs in the HA-BNIP3 expressing HCC cells, consistent with BODIPY 282 imaging of cells showing HA-BNIP3 decreasing LD number. There were also more mitochondria in EV 283 and HA-BNIP3 W18A expressing HCC cells (Fig. 7d), supporting conclusions from Fig. 3 that HA-BNIP3, 284 but not EV or HA-BNIP3 W18A , promotes mitophagy and lowers mitochondrial mass. Lysosomes were also 285 more evident in HA-BNIP3 expressing cells (Fig. 7d) and contained LD-like structures inside lysosomes 286 (Fig. 7e, middle) as also seen by fluorescence microscopy (Fig. 7e, left) as well as LDs associated with 287 mitochondria (Fig. 7e, right). Taken together, these data suggest that BNIP3 is promoting LD turnover at 288 the lysosome in conjunction with turnover of mitochondria.

289
Treatment of cells with LALi significantly decreased the growth of HCC cells expressing HA-BNIP3 290 but had no effect on the growth of cells expressing EV or HA-BNIP3 W18A (Fig. 7f), suggesting that BNIP3 291 suppresses HCC cell growth by promoting lysosomal turnover of LDs in a mitophagy-dependent manner.

292
LALi treatment also decreased palmitate oxidation suggesting that fatty acids liberated by lysosomal 293 lipases are fueling FAO in HCC cells (Fig. 7g). However, this effect was independent of BNIP3 consistent 294 with data described above (Fig. 5) indicating that altered FAO does not explain how BNIP3-dependent 295 LD turnover suppresses cell growth. At this time, it is not clear to us how LDs promote HCC cell growth 296 but LDs could act as reservoirs of phospholipids and other lipids used by growing cancer cells for 297 membrane expansion and other pro-growth functions. In summary, these data indicate that BNIP3 298 attenuates HCC cell growth by promoting LD turnover by a process we refer to as "mitolipophagy".

300
Low BNIP3 expression in human HCC correlates with worse overall survival.

301
We examined publicly available RNA-Seq data comparing transcript expression in human HCC 302 to healthy human liver tissue 38 , that showed expression of genes involved in fatty acid metabolism, 303 cholesterol metabolism and adipogenesis to be up-regulated in HCC compared to healthy liver (Supp.  data-sets, we observed that while lipid synthesis genes like FASN and ACLY were increased in HCC 306 compared to healthy tissue, BNIP3 was significantly decreased (Fig. 8a, 8b). Interestingly, PPARGC1A 307 that promotes mitochondrial biogenesis was also increased in HCC compared to healthy liver while 308 ACADM that promotes FAO was decreased (Fig. 8a). Linear regression analysis showed a highly CoA to malonyl CoA to promote lipid synthesis, and is frequently up-regulated in human cancers, 316 including HCC 42,43 . Interestingly, we found that high BNIP3 expression levels in combination with high 317 ACACA expression portended a highly significant increase in overall survival rates (Fig. 8c) in contrast 318 to the combination of high ACACA expression with low BNIP3 expression that had the worst prognosis 319 for overall survival (Fig. 8e). Moreover, when we examined the histology of liver sections from these HCC 320 patients, we observed that high BNIP3 mitigated against lipid accumulation linked to high ACACA 321 expression (Fig. 8d) in contrast to low BNIP3 where high levels of lipid were associated with high ACACA 322 expression (Fig. 8f). These results indicate that high BNIP3 expression is associated with decreased lipid 323 accumulation, less aggressive HCC and increased patient survival when linked to high ACACA 324 expression while conversely low BNIP3 predicts increased lipid accumulation and worse patient outcome.

327
Our work identifies a novel role for BNIP3 in limiting HCC by promoting lipid droplet turnover at the 328 lysosome (Fig. 9). This conclusion was reached after first interrogating the role of BNIP3 in rates of fatty 329 acid oxidation (FAO) and also in de novo lipogenesis (DNL). FAO in particular seemed likely to be

344
Nevertheless, this effect of BNIP3 in promoting FAO is unlikely to explain HCC growth suppression since 345 inhibition of FAO with Etomoxir failed to block the growth suppressive properties of BNIP3 (Fig. 5j).

356
Having excluded effects of BNIP3 on FAO and DNL as explaining how BNIP3 decreased both lipid 357 droplet (LD) number and HCC cell growth, we considered the possibility that BNIP3 was promoting somehow promoting lipophagy in which lipid droplets are turned over at the lysosome 32 . This became 361 more interesting when, unlike FAO and DNL inhibitors, we also observed a marked decrease in the ability 362 of BNIP3 to limit HCC tumor cell growth following treatment with LALi (Fig. 7f). Using LALi to inhibit 363 turnover of LDs that had been taken up by lysosomes, we observed that BNIP3 promoted increased LD 364 uptake by lysosomes compared to either EV or BNIP3 W18A , but also that it was the smallest LDs that were 365 taken up by the lysosome (Fig. 7c). Larger LDs appeared to be resistant to turnover at the lysosome. As 366 discussed above, mitophagy is preceded by mitochondrial fragmentation and there is preferential 367 turnover of smaller mitochondria causing larger more fused mitochondria to predominate 50,51 .

372
Thus, to reconcile our collective findings, we suggest that BNIP3 promotes LD turnover through 373 "mitoplipophagy" in which small LDs get turned over with associated smaller, fragmented mitochondria 374 (Fig. 9). Selective forms of autophagy imply that only the selected cargo gets turned over and certainly

449
To make cDNA, 1-2 μg of RNA was reverse transcribed using the High Capacity RNA-to-cDNA kit per sample using Taqman gene-specific fluorogenic probes (Applied Biosystems/Thermo Fisher).

453
Primers used for qPCR included in this manuscript are as follows:

562
All statistical analyses were carried out using GraphPad Prism of raw data. The data were analyzed using 563 one-way or two-way ANOVA with Tukey's post-test with a 95% confidence interval for data sets involving

687
BNIP3 promotes mitochondrial turnover at the autolysosome (mitophagy) and here we propose that BNIP3-dependent 688 mitophagy also promotes lysosomal turnover of mitochondrial-associated lipid droplets in a process termed "mitolipophagy", a 689 hybrid form of mitophagy and lipophagy (selective turnover of lipid droplets at the autolysosome). Fatty acids liberated from lipid 690 droplets do fuel fatty acid oxidation in HCC cells but this does not explain the growth suppressive effects of BNIP3 in HCC.

691
Rather we propose that lipid droplets promote tumor growth in other ways, such as serving as a reservoir for lipids to promote 692 plasma membrane growth and organelle biogenesis.