Mitochondrial genome variation affects humoral and cell-mediated innate immune responses and infection outcomes

The role of mitochondria in both adaptive and innate immune responses is increasingly recognized, but the role of mitochondrial DNA (mtDNA) variation as an immunomodulatory factor has received less attention. One reason for this is the difficulty of separating the effect of mtDNA from that of the nuclear genome. By utilizing the fruit fly Drosophila melanogaster, a powerful model system, we created cytoplasmic hybrids, aka. cybrid lines, where unique mtDNAs (mitotypes) were introgressed into a controlled isogenic nuclear background. We harnessed a panel of cybrid lines to study the effect of mtDNA variation on humoral and cell-mediated innate immune responses. Mitotypes exhibited heterogeneity in infection outcomes upon bacterial, viral and parasitoid infections. One mitotype of note (mtKSA2) was more immunocompetent when compared to other mitotypes. We performed transcriptomic profiling of uninfected and infected flies to find the mechanistic basis of the immunocompetence of the mtKSA2 line. We found that in uninfected flies mtKSA2 caused an upregulation of oxidative phosphorylation (OXPHOS) and tricarboxylic acid cycle (TCA) related genes and a downregulation of a set of antimicrobial peptides (AMPs). Upon infection, mtKSA2 flies produced transcriptomic changes that were infection type and duration specific. When we examined immune cells (hemocytes) in mtKSA2 larvae, we noted an increase in hemocyte numbers. These hemocytes were activated in the absence of infection, increased their production of ROS, and showed evidence of increased encapsulation efficiency upon parasitoid wasp infection. Overall, our results show that mtDNA variation acts as an immunomodulatory factor in both humoral and cell-mediated innate immunity and that specific mitotypes can provide enhanced protection against various infections.


Abstract 26
Mitochondria participate in various cellular processes including energy metabolism, apoptosis, stress 27 responses, inflammation and immunity. While mitochondrial dysfunction contributes to many 28 diseases, mitochondrial perturbations can also be beneficial, a phenomenon coined mitohormesis. We 29 investigated how mitochondrial variation contributes to the heterogeneity of infection outcomes and 30 whether moderate mitochondrial dysfunction could benefit immune-challenged individuals. We took 31 three approaches to model variation in mitochondrial oxidative phosphorylation (OXPHOS) in 32 As mitochondria contain their own maternally inherited circular genome, their functions rely on the 93 coordinated signaling between the nuclear and mitochondrial genomes, and mutations in one or both 94 genomes can lead to variation in mitochondrial functions. This variation, in turn, contributes to the 95 variation in a plethora of phenotypes shown in different organisms (for example [15][16][17][18]). In some 96 cases, mutations can hamper the mitochondrial function and result in mitochondrial diseases [19], the 97 severity of which vary depending on the mutation, the affected tissues [20] and on compensatory 98 mechanisms arising from (and depending on) the particular nuclear and mitochondrial genomes and 99 their interactions. Since mitochondrial dysfunction is involved in several pathologies, including 100 nuclear DNA [4]. We utilized Drosophila cybrid (cytoplasmic hybrid) strains where unique mtDNA 151 variants have been introgressed onto an isogenic nuclear background to separate the effect of variation 152 in the nuclear genome from that arising from the mtDNA [37]. A naturally occurring mtDNA variant 153 (called mtKSA2) was previously shown to cause the formation of melanotic nodules in Drosophila 154 cybrid larvae in different nuclear backgrounds [23]. This motivated us to examine if mtDNA variation 155 and specifically the mtKSA2 variant has an impact on the cell-mediated innate immune response and 156 the outcome of an infection. Using a parasitoid wasp L. boulardi infection, we induced the cell-157 mediated immune response in the larvae of four strains containing unique mtDNA variants, mtORT, 158 mtWT5A, mtBS1 and mtKSA2 (S1 Table) in an ORT nuclear background. 159 160 We measured the efficiency of the immune response against parasitoid wasps as a proportion of 161 melanised wasp larvae found in cybrid larvae reared at 25 °C and 29 °C. In general, fully (+) 162 melanised wasp larvae were scarce in all four strains when reared at 25 °C, ranging from 0 % in the 163 mitovariants mtORT and mtWT5A to 5 % in mtBS1 and 9 % in mtKSA2 ( Fig 1A, light purple). 164 Rearing the larvae at 29 °C improved the efficiency of the encapsulation response especially in 165 mtKSA2 larvae, as they were now able to melanise (both partially and fully) more wasp larvae than 166 the other cybrid strains at this temperature ( Fig 1A, dark purple). These results show that there is 167 mitovariant-specific variation in the immune response against parasitoid wasps, and that the mtKSA2 168 larvae are able to elicit a more efficient immune response compared to the other tested mitovariants. 169 170 mtKSA2 larvae show signs of enhanced aerobic glycolysis 171 Given the elevated parasitoid melanisation in the mtKSA2 larvae, we focused on this strain to explore 172 the underpinnings of mtDNA-mediated enhancement of cellular immunity. mtKSA2 harbours the 173 mutation D21N in the mtDNA encoded cytochrome b (cyt-b) gene of the OXPHOS cIII, which has 174 been predicted to destabilize the interactions between the subunits of cIII, based on crystal structure 175 models [23]. mtKSA2 also has another mutation (A75T) in the cIV gene Cytochrome c oxidase 176 subunit III (COIII), but the possible effects of this mutation have not been characterized [23]. We 177 reasoned that the cyt-b mutation may lead to impaired OXPHOS function, which, in turn, may lead 178 to a switch from mitochondrial respiration to favour cytosolic aerobic glycolysis as the source of 179 energy. If this is the case, the larvae may need to mobilize their storage sugars to meet the increased 180 glucose demand. Therefore, we measured the levels of the storage sugar glycogen in the cybrid larvae. 181 As female and male Drosophila may have differences in their metabolism [38], sexes were analysed 182 separately. mtKSA2 larvae, especially females, showed a trend for decreased levels of glycogen 183 storages when compared the other cybrid strains, but the differences were not statistically significant 184 ( Fig 1B-B'). Since differences in feeding among the larvae may affect the ability to built up glycogen 185 storages, we next checked the larval gut contents as a proxy of food intake. The trend for reduced 186 glycogen was not explained by reduced feeding, since mtKSA2 males had more gut content than the 187 other cybrids, and also mtKSA2 females had increased gut content compared to the mtORT females, 188 but not in comparison to mtBS1 or mtWT5A females (S1A- B Fig). Of note, the size of the larvae, 189 based on the protein content in the samples, did not statistically significantly differ between the cybrid 190 larvae, indicating that there are no major defects in their food consumption/metabolism that would 191 affect the growth of the animals (S1C Fig). But it is noteworthy that mtKSA2 and mtORT male larvae 192 were smaller than the respective females, which was not the case for mtBS1 and mtWT5A larvae 193 To further disentangle the glycolytic state of the larvae, we measured the expression levels of the 196 transcription factor similar (sima) and its target Lactate dehydrogenase (Ldh) in the cybrid larvae. 197 The expression of both genes has been shown to be increased in Drosophila when cells shift to utilize 198 aerobic glycolysis [39]. Sima expression was elevated in mtKSA2 male larvae when compared to the 199 other cybrids, and also in females in comparison to mtORT and mtWT5A (Fig 1C-C'). However, 200 there were no detectable changes in the expression of Sima's target gene Ldh among the cybrids (Fig  201   1D-D').  202   203 Taken together, mtKSA2 flies showed a trend towards smaller glycogen reservoirs and they had more 204 food content in their guts at the 3 rd instar stage than other cybrids, which may indicate that the 205 mtKSA2 flies utilize their glycogen storages more and possibly feed more as an attempt to replenish 206 the reduced glycogen storages. This, together with the increased expression levels of sima in the 207 mtKSA2 larvae suggests an increased utilization of aerobic glycolysis in these animals. 208 209 mtKSA2 larvae have a higher hemocyte count and elevated hemocyte ROS 210 As the encapsulation response against the parasitoid wasps requires hemocyte proliferation and 211 differentiation of a specialized immune-activated hemocyte type (the lamellocyte), we tested the 212 amounts and types of hemocytes in the cybrid larvae. We applied flow cytometry to assess the total 213 number of circulating hemocytes in uninfected and L. boulardi -infected cybrid larvae (S2A-C' Fig). 214 In general, we did not observe sex-differences in hemocyte numbers, apart from wasp-infected 215 mtKSA2 males, which had slightly more hemocytes than females at 29 °C (S2D-E' Fig). Therefore, 216 the hemocyte data was pooled for the subsequent analyses. We found that uninfected mtKSA2 larvae 217 had higher hemocyte counts than the other strains at both temperatures (Fig 2A), and at 29 °C this 218 was the case also after wasp infection (Fig 2B). Taking advantage of the morphological difference 219 between plasmatocytes (round cells) and lamellocytes (flat and discoidal cells), we sub-gated the total 220 hemocytes in two groups in an FSC-Area (FSC-A) vs. FSC-Height (FSC-H) plot. Plasmatocytes (or 221 any round cell) form a population roughly at a 45° angle in the FSC-A vs FSC-H plot due to having 222 an equal ratio between height and area. We gated the cells that deviated from this ratio and called the 223 gate the "activated hemocyte gate" (S2C-C' Fig). More mtKSA2 hemocytes were in this gate both in 224 uninfected and wasp-infected larvae reared at 29 °C than in the other cybrid strains (Fig 2A' & B'). 225 To test if the differences in the hemocyte numbers might be due to the differences in the hemocyte 226 viability, we analysed the proportion of living hemocytes (PI-negative) out of a total amount of 227 hemocytes detected in the main hemocyte gate. There were no sex-specific differences in hemocyte 228 viability and also the strain-specific differences were small (S2F-G' Fig). mtKSA2 larvae had a 229 slightly higher proportion of live hemocytes than mtORT larvae in all conditions except for in 230 uninfected larvae reared at 29 °C, where there was no difference (S2F-G' Fig). . We did not detect differences in the expression levels of PPO3 between the studied 236 strains ( Fig 2C). Therefore, it remains somewhat uncertain if the mtKSA2 mitovariant, besides 237 increasing the total hemocyte numbers, also induces lamellocyte formation, or perhaps affects some 238 other aspect of hemocyte activation that is beneficial for the cellular immune response. library RNAi strains and from 2 to 31% in larvae from the KK library strains (Fig 3C). ND-75 (cI) 286 knockdown caused only a mild melanotic nodule phenotype, resulting a nodule prevalence of 2 % in 287 the GD strain and 1 % in the KK strain. Another exception was the cII gene SdhD whose knockdown 288 in hemocytes caused nodules in 1 % in the GD strain and no nodules in the KK strain). 289

290
Having established a link between the formation of melanotic nodules and OXPHOS dysfunction in 291 hemocytes, and considering that the mitovariant mtKSA2 exhibited both nodules [23] and enhanced 292 immune response against parasitic wasps (Fig 1A), we hypothesized that both of these phenomena 293 could be explained by hemocyte differentiation. Therefore, we checked if silencing the OXPHOS 294 genes in hemocytes affected hemocyte numbers and/or types. We utilized the in vivo hemocyte 295 reporter system (eater-GFP and msn-mCherry) to identify five hemocyte populations based on the 296 levels of the GFP and mCherry fluorescence [34]. In uninfected msn-mCherry,eater-GFP;Hml D -297 GAL4;He-GAL4/w GD larvae (control), as expected mainly hemocytes expressing high levels of eater-298 GFP (plasmatocytes) were present (Fig 4A). Knocking down genes of OXPHOS complexes I, III, IV 299 and V resulted in hemocyte activation as uninfected larvae produced infection-specific hemocytes, 300 including lamellocytes, which express high levels of msn-mCherry and do not express the 301 plasmatocyte marker eater-GFP (Fig 4A-B). In this context complex II behaved as an outlier, as the 302 knockdown of SdhD did not induce hemocyte activation (Fig 4A-B). Most of the OXPHOS gene 303 knockdowns also resulted in an increase of total hemocyte numbers, but this effect was more 304 inconsistent than the hemocyte activation phenotype (including lamellocyte formation), which tended 305 to get stronger the further along the OXPHOS complex chain the knockdown was ( Fig 4B).   To verify that silencing the selected OXPHOS genes resulted in an alteration in mitochondrial 342 function, we measured the mitochondrial membrane potential (ΔΨm), as a readout of mitochondrial 343 activity in the hemocytes. For this, we used a TMRM dye and flow cytometry. As a control we treated 344 hemocytes from msn-mCherry,eater-GFP;Hml D -GAL4;He-GAL4/w GD larvae with carbonyl cyanide 345 3-chlorophenylhydrazone (CCCP), an uncoupler of oxidative phosphorylation in mitochondria, and 346 compared the signal to untreated hemocytes ( Fig 5A). We compared only the eater-GFP-expressing 347 plasmatocytes to avoid possible hemocyte type -specific effects related to the mitochondrial 348 membrane potential. When comparing the plasmatocytes of ND-75, ox or ATPsynCF6 silencing (msn-mCherry,eater-GFP;Hml D -GAL4;He-GAL4/UAS-OXPHOS RNAi) to the controls, we found that ox 350 and ATPsynCF6 knockdown showed significantly lower MitoProbe signal intensity, indicating loss 351 of membrane potential (Fig 5A'-A''). In the ND-75 knockdown plasmatocytes the reduction in the 352 MitoProbe signal was milder (and statistically not significant) than the reduction seen in ox and 353 ATPsynCF6 knockdown hemocytes. Also in the case of cII gene SdhD the knockdown did not result 354 in the loss of mitochondrial membrane potential but instead there was a small but non-significant 355 elevation of the membrane potential ( Fig 5B-B'). Next, we investigated whether a switch to aerobic glycolysis in hemocytes could be behind the 376 hemocyte activation seen accompanying OXPHOS perturbation. We compared the levels of glycogen 377 and trehalose in ND-75, ox and ATPsynCF6 knockdown larvae to those in the controls, sex 378 specifically. We hypothesized that in the knockdown larvae there would be a decrease in glycogen 379 levels, as the storage sugars are mobilized to feed glucose into accelerated glycolysis. This, in turn, 380 might be reflected in the levels of trehalose in the hemolymph, which is the main circulating sugar in 381 Drosophila larvae. We found that the glycogen levels in the OXPHOS knockdown larvae did not 382 significantly differ from the controls (Fig 6A-A'). The levels of hemolymph trehalose were increased 383 when ND-75 was silenced in hemocytes, but only in females, and decreased in male ox knockdown 384 larvae ( Fig 6B-B').  Fig 7A). Here, as an additional control, we also crossed the 412 OXPHOS RNAi strains to the w 1118 background (w 1118 /OXPHOS RNAi). Overall, the response against 413 wasps in the OXPHOS knockdown larvae was enhanced or stayed the same also when compared to 414 these RNAi strain -specific controls, with the exception of knockdown of the complex II ( Fig 7A). 415 Despite some strain and/or background -specific variation in the encapsulation response, overall 416 tendency was towards enhanced encapsulation response. On the contrary, when we knocked down 417 ND-75 and ATPsynCF6 from another tissue important for the Drosophila immune response, the fat 418 body, the melanisation response against wasps was drastically reduced ( Fig 7B). As the immune-419 enhancing effect was shown to be hemocyte-specific, we tested if hemocyte-specific OXPHOS we can phenocopy the effect of the mtKSA2 mutation by pharmacologically inhibiting the cIII 468 function with antimycin A. We aimed to induce mild mitochondrial perturbation to the cybrid nuclear 469 background strain ORT and the RNAi background strain w GD by feeding the larvae different 470 concentrations of antimycin A and checking the larvae for the presence of melanotic nodules as a 471 sign of phenocopy. Using the concentrations 4, 6 and 8 ug/ml gradually increased the percentages of 472 melanotic nodules in the ORT and w GD larvae. However, there was a significant background effect 473 on the sensitivity to the antimycin A treatment as ORT produced more nodules in each of these 474 concentrations when compared to w GD ( Fig 8A). As we aimed to cause a mild cIII perturbation, we 475 continued with the w GD line and an antimycin A concentration of 6 ug/ml. Feeding antimycin A led 476 to a reduced melanisation response against wasps ( Fig 8B) and did not have an effect on the hemocyte 477 counts in the w GD larvae (Fig 8C-C'). We also fed 6 ug/ml antimycin A to msn-mCherry,eater- Since melanotic nodules are a common sign of the activation of the fly immune cells, the hemocytes, 522 we next quantified and qualified the hemocytes in our three models. The uninfected mtKSA2 larvae 523 were shown to harbor more hemocytes than the other mitovariants, and the shape analysis of the 524 hemocytes indicated the mtKSA2 larvae have immune-activated hemocyte types, although we could 525 not indisputably verify this. In antimycin A treated larvae, we did not detect signs of hemocyte 526 activation. Of note, we observed that the development of the antimycin A -fed larvae that exhibited 527 melanotic nodules was delayed approximately one day. Since the hemocyte analyses were performed 528 on larvae that reached the late 3 rd instar stage at the same rate as the controls (on the 5 th day from egg-529 lay when reared at 29 °C), most antimycin A -fed larvae with visible melanised aggregates were 530 excluded from the analyses. In the case of OXPHOS complex knockdown in hemocytes, the 531 formation of melanotic nodules was indeed associated with the formation of immune-induced 532 hemocyte types, mainly activated plasmatocytes and lamellocytes. Besides the antimycin A -fed 533 larvae, the two other models did not cause a delay in development, indicating that the pharmacological 534 approach and the used consentrations were causing more drastic changes in the animals. This in turn 535 shows that although mild OXPHOS perturbations might be beneficial for the immune response, more 536 severe perturbations are harmful. Indeed, an antimycin A -treatment as well as OXPHOS cI and cV 537 knockdown in the fat body led to a delayed development time and weakend immune response while 538 mitovariant mtKSA2 and OXPHOS gene knockdowns in hemocytes that had no effect on the 539 development time resulted in an enhanced innate immune response. However, when the locomotion 540 activities and sleep patterns were studied in eight Drosophila cybrid lines, including those studied 541 here, it was shown that especially the males that had the mitovariant mtKSA2 were moving less and 542 sleeping more than the other cybrids [53]. This indicates that the activated immune system of the 543 mtKSA2 flies might be energetically costly and reflect on the activity levels of the mtKSA2 cybrids. 544 545 Hemocyte activation in dysfunctional OXPHOS is associated with a loss of mitochondrial membrane 546 potential and reduced apoptosis 547 When studying the silencing of OXPHOS complexes I, III and V in hemocytes in more detail, we 548 found that hemocyte activation was accompanied with, and potentially induced by, the loss of the 549 mitochondrial membrane potential. The mitochondrial membrane potential, the electronic charge 550 difference between the intermembrane space and the mitochondrial matrix, is mainly generated by 551 the action of the proton pumps functioning in OXPHOS (cI, cIII and cIV). We found that in addition 552 to cIII silencing, the membrane potential was reduced also when silencing the ATP synthase stalk 553 [39]. We show that mtKSA2 larvae have a tendency towards smaller glycogen reservoirs than the 643 other mitovariants and higher expression of sima. In larvae with OXPHOS gene knockdown in 644 hemocytes, we did not find any consistent evidence towards reduced glycogen storages, increased 645 levels of trehalose in the hemolymph or increased sima expression. But, we did find that Ldh 646 expression was upregulated in ox (cIII) and ATPsynCF6 (cV) knockdown hemocytes. Therefore, it is 647 possible that we captured a "resolution phase" of the effect of OXPHOS gene knockdown in 648 hemocytes and therefore could not detect the earlier signs of the switch to aerobic glycolysis. Since 649 increased Ldh expression was seen in samples also containing lamellocytes (ox and ATPsynCF6 KD), 650 this could mean that lamellocytes are expressing Ldh at higher levels than plasmatocytes.

RNA isolation from larvae 811
Pools of 10 3 rd instar larvae (males and females separately, three replicates per genotype) were 812 collected in Eppendorf tubes, snap frozen and stored at -80 °C. Total RNA was extracted using the 813 TRIzol (Ambion by Life Technologies) method. 125 µl of TRIzol reagent was added to frozen larvae 814 and the tissues were ground using a motor-driven grinder with a fitted pestle. Then, 25 µl of 815 chloroform was added and the samples were vortexed for 20 s, after which they were centrifuged for 816 15 min at 12000 g. The upper phase was pipetted into a new Eppendorf tube, 62.5 µl of isopropanol 817 was added, followed by vortexing for 15 sec and a 10 min incubation at RT. After centrifugation for 818 15 min at 12 000 g, the excess liquid was removed, and the remaining pellet was washed twice with 819 500 ul of 75 % EtOH (centrifuging 5 min at 7500 g). After the second wash, the pellet was air-dried 820 until the EtOH had evaporated, and the pellet was dissolved in 50 µl of nuclease free water. The RNA 821 concentration was determined using a NanoDrop ND-1000 spectrophotometer, and the RNA was 822 diluted to 10 ng/µl prior to storing at -80 °C. To make sure that we were able to differentiate between different levels of 838 ROS in hemocytes with this method, we run positive (high ROS) and negative (low ROS) controls. 839 In our hands, treating the bled hemocytes with Tert-butyl hydroperoxide (TBHP) as a positive control 840 and with antioxidant N-acetylcysteine (NAC) as a negative control did not produce clear differences 841 in the CellROX signal intensity. Therefore, we used hemocytes from L. boulardi -infected larvae as 842 a positive ROS control and larvae fed with NAC (concentration in the food 1mg/ml) starting from 843 the egg lay as a negative control. We were able to confirm that the highest proportion of CellROX-844 positive hemocytes were found in the untreated wasp-infected larvae, followed by descending 845 CellROX in NACuninfected, NAC + infected and finally NAC + uninfected larvae. The same was seen 846 when looking at the CellROX intensity in the samples (Supplementary Fig 2C). To measure the number of apoptotic cells, we used annexin V staining. During apoptosis, 893 phosphatidylserine is translocated from the inner side of the plasma membrane to the outer side. 894 Annexin V binds to phosphatidylserine, allowing the detection of apoptotic cells using annexin V 895 conjugated to a fluorophore. 20 male 3 rd instar larvae per sample were washed and then dissected in 896 100 µl of the annexin-binding buffer (10 mM HEPES, 140 mM NaCl, and 2.5 mM CaCl2, pH 7.4) to 897 release the hemocytes. The samples were kept on ice prior to staining. Then, the samples were stained 898 with 2.5 µl of the Annexin V Pacific Blue conjugate (Invitrogen) and incubated for 15 minutes at RT. 899 After the incubation, the samples were diluted with 400 µl of annexin-binding buffer and analysed 900 using a CytoFlex S flow cytometer. The following channels and gain settings (in parentheses) were

ROS detection in OXPHOS complex knockdown hemocytes 908
The CellROX deep red reagent did not work properly in our hands, and so to measure ROS in the 909 hemocytes together with OXPHOS gene knockdown, the CellROX green probe was utilized as 910 The hemolymph from 50 larvae (males and females separately) was pooled per sample (three replicate 937 samples per genotype) and centrifuged for 7 minutes at 2500 g at 4 °C to pellet the hemocytes. As 938 much of the PBS was removed as possible (the hemocyte pellet is not visible and is loosely attached 939 on the Eppendorf tube) prior to storing the samples at -80 °C. Total RNA was extracted with the 940 Single Cell RNA Purification Kit (Norgen Biotek) protocol that is optimized for extracting RNA from 941 low numbers of cells. Samples were treated with DNase I (RNAse free DNase I kit, Norgen Biotek) 942 as described in the manufacturer's protocol. The RNA quality was checked using a NanoDrop ND-943 1000 spectrophotometer (Thermo Scientific) and each sample was diluted to 10 ng/µl prior to storing 944 at -80 °C. 945

Gene expression 946
Quantitative Real-time PCR (qRT-PCR) was used to detect the expression levels of the genes of 947 interest (from hemocytes or whole larvae) and to quantify the efficiency of OXPHOS gene 948 knockdown in the larval hemocytes. For the qRT-PCR reactions, the iTaq Universal Sybr green One-949 step kit (Bio-Rad) was used. 10 µl reaction mix per sample was as follows: 5 µl of iTaq universal 950 females and 10 males of the strains ORT and w GD were mated, and the females were left to lay eggs 972 at 25 °C in regular food. After two days the late 1 st -early 2 nd instar larvae were transferred to food 973 vials containing antimycin A and placed at 29 °C. This was repeated several times to obtain enough 974 larvae for the assays. 100 3 rd instar larvae were dissected from ORT and w GD strains reared on the 975 three antimycin A concentrations to check for the presence of melanotic nodules. The effect on the 976 response against L. boulardi parasitoid wasps (n=100) and on hemocyte composition (n=30) were 977 studied using 3 rd instar w GD larvae fed with 6 ug/ml antimycin A. In addition, the hemocyte 978 composition was assessed from 3 rd instar hemocyte reporter larvae (msn-mCherry,eater-GFP;Hml D -979 GAL4;He-GAL4/w GD ) reared on 6 ug/ml antimycin A food at 29 °C to gain a more detailed picture 980 of the effect of antimycin A on hemocyte profiles.