Ankyrin2 is required for neuronal morphogenesis and long-term memory and interacts genetically with HDAC4

Dysregulation of HDAC4 expression and/or subcellular distribution results in impaired neuronal morphogenesis and long-term memory in Drosophila melanogaster. A recent genetic screen for genes that interact in the same molecular pathway as HDAC4 identified the cytoskeletal adapter Ankyrin2 (Ank2). Here we sought to investigate the role of Ank2 in neuronal morphogenesis, learning and memory, and to examine the nature of interaction with HDAC4. We found that Ank2 is expressed widely throughout the Drosophila brain where it localizes predominantly to axon tracts. Pan-neuronal knockdown of Ank2 in the mushroom body, a region critical for memory formation, resulted in defects in axon morphogenesis, and similarly reduction of Ank2 in lobular plate tangential neurons of the optic lobe disrupted dendritic branching and arborization. Conditional knockdown of Ank2 in the mushroom body of adult Drosophila significantly impaired long-term courtship memory, and this requirement for Ank2 was isolated to gamma (γ) neurons of the mushroom body. As overexpression of HDAC4 in γ neurons also impairs the formation of long-term courtship memory, this suggests that any functional relationship between these proteins during LTM likely occurs in γ neurons. We determined that the genetic interaction requires the presence of nuclear HDAC4 and is not dependent on a conserved putative ankyrin-binding motif present in HDAC4. In summary, we provide the first characterization of the expression pattern of Ank2 in the adult Drosophila brain and demonstrate that Ank2 is critical for morphogenesis of the mushroom body and for the molecular processes required in the adult brain for formation of long-term memories.


201
Co-immunoprecipitation 202 Whole cell extracts were prepared as per the western blotting method above. 203 Immunoprecipitation (IP) was performed with the Pierce Classic IP Kit (Thermo 204 Scientific) according to the manufacturer's instructions. Anti-Myc or anti-HA antibody 205 (1 L) was incubated overnight with 1 mg of lysate. Following elution in 2x sample 206 buffer, IP samples were processed for SDS-PAGE and western blotting with anti-HA or 207 anti-Myc alongside 30 g input samples. Anti--tubulin (1:500) was used as a loading 208 control.

210
RT-qPCR 211 elav-GAL4 females were crossed to UAS-Ank2 RNAi males to generate progeny in 212 which Ank2 was knocked down in all neurons; and progeny of elav-GAL4 crossed to 213 w(CS10) served as the control. To confirm knockdown, total RNA was extracted from 214 Drosophila heads from three independent crosses with the RNeasy Mini kit (Qiagen) 215 according to the manufacturer's instructions. cDNA was synthesized from 1 g of 216 total RNA with Transcriptor (Roche) as per the manufacturer's instructions. RT-qPCR 217 was conducted using SsoFast-EvaGreen (BioRad) reaction master on a Lightcycler II 218 480 instrument (Roche), following manufacturer's instructions. The following primers 219 were used: EF1a48Drev 5'-TACGCTTGTCGATACCACCG-3'. A 5-fold dilution of cDNA from control 222 flies was used as template to prepare a standard curve to confirm efficiency of the 223 PCR reactions. Relative quantification was conducted using 2 -ΔΔCt method, normalizing 224 to the housekeeping gene Ef1α48D (LIVAK AND SCHMITTGEN 2001). Ank2 expression was 225 reduced to 0.42  0.12 (meanstandard error) of that of the control, student's t-test 226 t(12)=4.74, p<0.001.

228
Courtship Suppression Assay 229 The repeat training courtship suppression assay was used to assess 24-hour long-term 230 courtship memory. This is an experience-dependent assay in which wild-type male 231 flies that have been previously rejected by a mated female will reduce their courtship 232 behavior towards a new mated female. During mating, the male pheromone cVA is 233 transferred to the female, and the presence of this pheromone on the female causes 234 the male to reduce his courtship towards her. Males that have previously experienced 235 rejection will suppress courtship towards another mated female due to an enhanced prior to testing. Each trained or sham male fly was then placed into a testing chamber 246 containing a mated wild-type female and was scored for the time spent performing 247 stereotypic courtship behaviors over the ten-minute period. A courtship index (CI) was 248 calculated as the proportion of the ten-minute period spent courting. A mean CI for 249 each group was determined, and from this a memory index (MI) was calculated by the 250 following equation: MI = 1-(CI of each trained fly/mean CI of sham group) 251 (n≥16/group). The MI was measured on a scale of 0 to 1, a score of 0 indicating 252 memory was no different than untrained sham controls. In all experiments, the scorer 253 was blind to the genotype of the flies. For assessment of immediate short-term 254 memory, the training session was reduced to one hour and flies were tested 255 immediately after training. For assessment of learning, the male was placed with a 256 mated female for an hour and the first ten minutes and last ten minutes were scored 257 for courtship behavior. The learning index was calculated as 1-(CI last 10 mins/CI first 258 10 mins). For statistical analyses, data was arcsine transformed to approximate a 259 normal distribution and one-way ANOVA with post-hoc Tukey's HSD test was used to 260 assess significance (p<0.05).

262
Scanning electron microscopy (SEM) 263 Flies were anaesthetised using FlyNap (Carolina Biologicals) and fixed in primary 264 modified Karnovsky's fixative (3% glutaraldehyde, 2% formaldehyde in 0.1 M 265 phosphate buffer, pH 7.2) with Triton X-100 by vacuum infiltration. They were then 266 placed in fresh fixative and incubated at room temperature > 8 hours, followed by 3 x 267 10 min washes in phosphate buffer (0.1 M, pH 7.2). Dehydration was carried out via a 268 graded ethanol series for ten to fifteen minutes at 25%, 50%, 75%, 95%, and 100% 269 ethanol, followed by a final one-hour incubation in 100% ethanol. The flies were then 270 critical point dried using CO2 and 100% ethanol (Polaron E3000 series II drying 271 apparatus). Heads were removed, mounted onto aluminium stubs and sputter coated 272 with gold (Baltex SCD 050 sputter coater). Imaging was performed with a FEI Quanta 273 200 Environmental Scanning Electron Microscope at an accelerating voltage of 20 kV.

274
To determine the sizes of each eye for comparison between genotypes, ImageJ 275 software was used to draw a line surrounding the eye and calculate the area in 276 arbitrary units. Statistical analysis was performed with one-way ANOVA with post-hoc 277 Tukey's HSD test with significance set at p<0.05.

279
To provide a semi-quantitative analysis of SEM images, a scoring system was 280 developed based on observations of phenotypes resulting from overexpression of one 281 or two copies of UAS-HDAC4 in a previous study (SCHWARTZ et al. 2016). No defects: 282 the eye appears wild-type -ommatidia are organised in a regular array with no fusion 283 and mechanosensory bristles are correctly positioned between each ommatidium. 284 Mild: Presence of one of the following phenotypes; between 5 -10 instances of an 285 abnormal number of interommatidial bristles, mild ommatidial disorganisation or 286 fusion of ommatidia in up to two areas. Moderate: all mild phenotypes were 287 collectively observed or one of the following phenotypes was observed; between 10 288 -20 instances of an abnormal number of interommatidial bristles, moderate 289 disorganisation or fusion of ommatidia in up to five areas. Major: all moderate 290 phenotypes were collectively observed or one of the following phenotypes was 291 observed; more than 20 instances of an abnormal number of interommatidial bristles, 292 major disorganisation of the ommatidial array, fusion of ommatidia in up to 10 areas 293 with few large areas of fusion or up to 50 collapsed ommatidia. Severe: all major 294 phenotypes were collectively observed or one of the following phenotypes was 295 observed; severe disorganisation, fusion in more than 10 areas or multiple large 296 patches, more than 50 collapsed ommatidia or severe collapsing of ommatidia 297 resulting in central hole-like cavities. Statistical analysis was assessed with the Fisher's 298 Exact test with significance set at p<0.05. 299 300 RESULTS 301 Characterisation of Ank2 expression in the brain 302 To date, the expression and localisation pattern of Ank2 has been described in the that Ank2-L localizes to axon tracts across the adult brain ( Fig 1C). Ank2-L did not co-310 distribute with the glial marker Reversed Polarity (Repo) (ALFONSO AND JONES 2002), 311 confirming its specific neuronal expression pattern (Fig 1D). Since the mushroom body through the mushroom body. We confirmed that Ank2-L and Nrg codistributed in 328 multiple axon tracts including the axons of the mushroom body, where both were 329 observed in the α, β and γ lobes (Fig 1E). Ank2-L was also concentrated in axon tracts 330 surrounding the calyx of the mushroom body (Fig 1F). 331 332 Ank2 is essential for axon and dendrite morphogenesis 333 We previously found that overexpression of HDAC4 impairs morphogenesis of 334 mushroom body axons, with deficits in axon branching and guidance observed as 335 missing α and/or β lobes, misdirected axons as well as the appearance of fused β 336 lobes, resulting from defects in axon termination across the midline (MAIN 2021). To 337 investigate whether Ank2 is also required for axon morphogenesis, the morphology of 338 the mushroom body in brains with reduced Ank2 was analyzed via detection of 339 Fasciclin II. This cell adhesion molecule is highly expressed in the α, β and γ lobes of 340 the mushroom body (CRITTENDEN et al. 1998) and is a commonly used marker to 341 visualise mushroom body lobe architecture (Fig 2A). Pan-neuronal knockdown of Ank2 342 with an inverted repeat hairpin that targets all long isoforms of Ank2 mRNA for 343 degradation resulted in a variety of phenotypic defects of the mushroom body, 344 including thin lobes, missing lobes and guidance abnormalities (Fig 2B-F). GAL4 activity 345 increases at higher temperatures and accordingly we observed more severe defects 346 when the temperature was raised during larval development (Table 1).

348
During projections ( Figure 2J, K) leading to reduced total branch length ( Figure 2L). These data 363 suggest that wild-type levels of Ank2 are required for both axon branching, guidance 364 and elongation was well as normal dendritic branching and arborization. 365 366 LTM requires Ank2 expression in the γ lobe of the mushroom body 367 We that of a sham male. A score of zero indicates that memory is impaired and no 378 different from untrained sham controls, whereas a higher memory index indicates 379 intact memory. This form of courtship memory has been recently described as "cVA-380 retrievable memory" to differentiate from the associative memory formed when 381 virgin females are used for testing, which uses different circuitry for memory retrieval 382 (RAUN et al. 2021). Learning and immediate short-term memory were unaffected by 383 pan-neuronal knockdown of Ank2 (Fig 3A,B). Long-term courtship memory is 384 dependent on an intact mushroom body, therefore unsurprisingly, pan-neuronal 385 knockdown of Ank2 during development resulted in a significant and severe loss of 386 LTM formation compared to control genotypes ( Fig 3C). This was not due to an effect 387 on courtship behavior as sham males of each genotype all spent approximately the 388 same percentage of time courting (87 to 89%, Figure 3D).

390
To avoid the developmental deficits resulting from decreased Ank2 expression and 391 allow assessment of the role of Ank2 specifically in adult memory processes, 392 knockdown of Ank2 was restricted to the mature brain with GAL80ts, a temperature 393 sensitive inhibitor of GAL4 activity (MCGUIRE et al. 2004) (Fig 3E). Flies were raised at 394 the permissive temperature of 19°C at which GAL80ts is active. Seventy-two hours 395 after eclosion, male flies from the F1 progeny were collected individually and 396 transferred to 30°C to inactivate GAL80ts and thus induce pan-neuronal RNAi 397 expression. After three days, males were tested in the courtship suppression assay. 398 Adult-specific knockdown of Ank2 in all neurons resulted in impairment of LTM 399 formation ( Fig 3G) and when knockdown of Ank2 was restricted to the adult 400 mushroom body (Fig 3F) with OK107-GAL4, this impairment remained (Fig 3H) a differential requirement for Ank2 in specific mushroom body subtypes by restricting 405 expression of Ank2 RNAi to each subtype individually (Fig 3F). Expression with the α/β 406 and γ neuron driver MB247-GAL4 abolished LTM formation (Fig 3I). Knockdown in the 407 α/β neurons or α'/β' neurons did not significantly alter LTM (Fig 3J,K) whereas 408 knockdown in γ neurons with NP1131-GAL4 prevented LTM formation (Fig 3L). The γ 409 neuron driver 1471-GAL4 did not quite impair memory to significant levels (Fig 3M), 410 however this is a much weaker than NP1131-GAL4 (ASO et al. 2009). We therefore 411 tested an additional γ neuron driver R16A06-GAL4 which drives very strong expression 412 in the gamma lobe, but minimal expression elsewhere in the brain (JENETT et al. 2012), 413 and employed a second independent Ank2 RNAi line that also targets all long forms of 414 Ank2 mRNA, which together resulted in a significant reduction in LTM (Fig 3N).

416
Taken together these data show that Ank2-L is required for normal mushroom body 417 development and in the adult brain, wild-type levels of Ank2-L are required in the γ 418 lobes for normal LTM formation. This is strikingly similar to the phenotypes resulting 419 from manipulation of HDAC4 expression in that it is also required in the γ lobe for LTM The N-terminal region of HDAC4 contains an ankyrin repeat binding domain consisting 424 of a PxLPxI/L motif (Fig 4A) increased HDAC4 expression does not alter the level of Ank2 protein (Fig 4D,E).

441
Nuclear HDAC4 mediates the genetic interaction with Ank2 442 We previously showed that knockdown of Ank2 in the eye resulted in a mild 443 developmental impairment that was significantly enhanced when combined with 444 HDAC4 overexpression, indicating the two genes interact in the same molecular 445 pathway (SCHWARTZ et al. 2016). The HDAC4-induced impairments in eye development 446 are largely a consequence of nuclear accumulation of HDAC4, as expression of a 447 mutant variant of human HDAC4 that is sequestered in the nucleus resulted in a more 448 severe phenotype than overexpression of wild-type human HDAC4, whereas the 449 phenotype resulting from expression of a cytoplasm-restricted mutant was mild (MAIN 450 2021). To that end, we next investigated whether the genetic interaction between 451 Ank2 and HDAC4 in the eye is also dependent on the nuclear presence of Drosophila 452 HDAC4. In addition, we sought to determine whether the genetic interaction is 453 dependent on the putative ankyrin repeat-binding motif region of HDAC4. If not, this 454 would provide further confirmation that the genetic interaction between HDAC4 and 455 Ank2 is not through direct physical binding.

457
We first confirmed our previous observations by co-expressing DmHDAC4 and Ank2 458 RNAi under the control of GMR-GAL4, which drives expression in post-mitotic cells 459 posterior to the morphogenetic furrow (FREEMAN 1996), and examining the phenotype 460 of adult eyes. We also raised the temperature to 27°C to increase GAL4 activity and 461 thus increase the degree of knockdown in order to examine the resulting phenotypic 462 defects in eye development in more detail. Control flies (GMR-GAL4 crossed to the 463 background w[CS10] strain) showed predominantly normal ommatidial alignment and 464 no evidence of fusion (Fig 5A), whereas eyes with reduced Ank2 expression had fused, 465 collapsed and misaligned ommatidia lacking some interommatidial bristles (Fig 5B).

466
Overexpression of DmHDAC4 also resulted in missing/disorganised bristles and 467 misaligned and fused ommatidia (Fig 5C). Combined expression of DmHDAC4 and 468 Ank2 RNAi resulted in a more severe rough eye phenotype consisting of major areas 469 of ommatidial fusion, severe misalignment with hole-like cavities within the 470 ommatidia, and missing/disorganised bristles (Fig 5D). The severity of phenotypes was 471 scored (Table S1) and the percentage of eyes displaying the most severe phenotype 472 was significantly higher when DmHDAC4 and Ank2 RNAi were combined (Fig 5I). In 473 addition to the severe eye phenotypes observed above, the eyes were also 474 significantly physically smaller than those expressing either Ank2 RNAi or DmHDAC4 475 individually (Fig 5J). Together these synergistic phenotypes are consistent with our 476 previous findings (SCHWARTZ et al. 2016) and provide further confirmation that Ank2 477 and HDAC4 genetically interact together to influence eye development.

479
To further investigate the mechanism of this interaction, we examined whether it was 480 dependent on the nuclear activity of HDAC4. A nuclear-restricted mutant of HDAC4 481 (DmHDAC4 3SA) resulted in a more severe phenotype than wild-type HDAC4, with 482 increased disorganisation and fusion of ommatidia and a reduced number of bristles 483 (Fig 5E), which we have previously observed (MAIN 2021). DmHDAC4 3SA interacted 484 synergistically with Ank2 (Fig 5F), confirming that nuclear activity of HDAC4 is required 485 for the genetic interaction. To address whether the putative ankyrin binding motif in 486 Drosophila HDAC4 is required we substituted residues that have been shown to be 487 important for this interaction in mammals; P48 L50 P51 and I53 of the PxLPxI/L motif 488 (Fig 4A) to alanines to create the DmHDAC4 ΔAnk mutant. Expression of this mutant 489 also resulted in a moderate rough eye phenotype consisting of ommatidial fusion and 490 misalignment ( Figure 5G), indicating that the presence of this motif is not required for 491 the HDAC4 overexpression-induced eye defects. This mutant still interacted 492 genetically with Ank2, resulting in a significantly more severe rough eye phenotype 493 and reduced eye size when expressed in combination with Ank2 (Fig 5H), therefore 494 this interaction does not depend on the presence of the putative ankyrin-binding 495 motif of HDAC4, which is consistent with the lack of physical interaction with Ank2.
Here, we provide the first characterization of the expression pattern of Ank2 in the 499 adult Drosophila brain and demonstrate that Ank2 is critical for normal development 500 of the mushroom body and for formation of long-term memories. 501 502 Ank2 expression was observed throughout the brain and localized predominantly to 503 axon tracts, including the lobes of the mushroom body where it colocalized with Nrg. 504 As the mushroom body is a critical structure for memory formation (HEISENBERG et al. Together these data support the evidence for a functional relationship between Ank2, 531 Nrg and Moe in mushroom body development. 532 533 Pan-neuronal knockdown of Ank2 also severely impaired 24-hour LTM without 534 affecting courtship behavior, learning, or immediate courtship memory. It was 535 previously found that decreased expression of Ank2 in the mushroom body with 536 OK107-GAL4 did not result in learning deficits but caused significant impairment to we observed on knockdown of Ank2, and that Ank2 has been implicated in synapse 550 stability (HORTSCH et al. 2002;KOCH et al. 2008), it is unsurprising that memory would 551 be impaired, and from these data it cannot be determined whether Ank2 plays a 552 specific role in memory or whether the defects are a result of impaired morphogenesis 553 of the mushroom body. To that end, in order to dissociate developmental effects from 554 the molecular processes required for LTM in an adult brain, knockdown of Ank2 was 555 restricted to mature neurons with GAL80. A specific deficit in 24 hour LTM was 556 observed when Ank2 was knocked down in the adult mushroom body, and subsequent 557 testing of GAL4 drivers that restrict expression to specific mushroom body subtypes 558 revealed that knockdown in just the γ neurons was sufficient to impair memory. We 559 previously showed that overexpression of HDAC4 in γ neurons also impaired the between these proteins also occurs during LTM formation. 565 566 The identification that Ank2 is required in the γ neurons of the mushroom body for 567 normal cVA-retrievable memory is consistent with current models of the circuitry that 568 facilitates    Immunohistochemistry with anti-FasII on whole mount brains reveals morphological defects of the mushroom body resulting from pan-neuronal expression of Ank2 RNAi driven by elav-GAL4. Knockdown was confirmed by RT-qPCR (as described in the methods). All images are frontal confocal projections through the mushroom body. Scale bar = 50 μm. A. Wild-type mushroom body stained with anti-Fas II to highlight the α, β and γ lobes. B. Thin α lobes (arrowheads), a prematurely terminated β lobe (arrow) and a missing γ lobe (asterisk). C. β lobes (arrows) have crossed the midline and appear fused (circle). D. A thin (arrowhead) and a prematurely terminated α lobe (arrow). E. Misoriented β lobe (arrow) which projects anteriorly rather than its usual medial orientation (asterisk). A prematurely terminated α lobe is also present (arrow). F. None of the lobes have elongated (asterisks) and axon stalling is observed whereby the axons grow in a ball-like structure (arrow). Grey asterisk indicates the midline. G-J. Immunohistochemistry on whole mount brains with anti-GFP driven by 3A-GAL4 to detect Lifeact in LPTCs whole mount brains. All images are confocal projections through the optic lobe of the brain. G. The dendritic arbor of the six neurons comprising the vertical system of LPTCs in a wild-type brain is visualised with anti-GFP. Scale bar = 50 µm.   Learning and memory were assessed with the courtship suppression assay. The controls included in each assay are the GAL4 driver (plus tubP-GAL80ts where indicated) crossed to CS, and UAS-Ank2 crossed to CS, such that the progeny are heterozygous for either the driver or the RNAi. A-D. elav-GAL4 and UAS-Ank2 RNAi flies were crossed to achieve panneuronal knockdown of Ank2 in progeny. A. Learning was unaffected knockdown of Ank2 (ANOVA, F (2,47) =0.002, p=0.252). B. Immediate memory was also unaffected (ANOVA, F (2,45) =0.044, p=0.819). C. Ank2 knockdown impaired long-term memory (ANOVA, F (2,60) =7.31, p<0.001; post-hoc Tukey's HSD, **p<0.01). D. Courtship activity was not impaired by pan-neuronal knockdown of Ank2 (ANOVA, F (2,51) =0.14, p=0.870). E-K. Knockdown of Ank2 in the adult mushroom body impairs LTM. Ank2 was knocked down in specific regions of the brain by crossing Ank2 RNAi to the indicated driver line and tubP-Gal80ts. E. Schematic diagram depicting the induction of expression in the adult mushroom body. Expression was restricted to the adult brain by raising flies at raised at 19°C, at which temperature GAL80 represses GAL4. After eclosion, when flies were 3-5 days old, the temperature was raised to 30°C for 72 hours, after which training commenced. At this temperature GAL80 is inactivated, allowing GAL4 to induce transgene expression. Twenty-four hours after training, the flies were equilibrated to 25°C for one hour prior to testing. F. Schematic diagram labelling the lobes of the mushroom body in which Ank2 was knocked down. G. Pan-neuronal knockdown of Ank2 in the adult brain impairs long term memory (ANOVA, F (2,54) =0.317, p<0.01; post-hoc Tukey's HSD, *p<0.05). H. Similarly, memory is also impaired when knockdown of Ank2 is restricted to the mushroom body (ANOVA, F (2,52) =0.922, p<0.001; post-hoc Tukey's HSD, **p<0.01, *p<0.05). I. When knockdown is restricted to the a/b and g neurons of the mushroom body, long-term memory is still disrupted (ANOVA, F (2,51) =0.923, p<0.0001; post-hoc Tukey's HSD, **p<0.01). J. Reduction of Ank2 in just the a/b neurons has no significant effect on long-term memory (ANOVA, F (2,41) =0.025, p=0.819). K. There is also no impairment when Ank2 is reduced in the a'/b' neurons (ANOVA, F (2,51) =0.122 p=0.372). L. Ank2 is required in the g lobes, as knockdown with NP1131-GAL4 impairs LTM (ANOVA, F (2,46) =0.312, p<0.01; post-hoc Tukey's HSD, *p<0.05). M. The weaker g lobe driver 1471-GAL4 reduced LTM, however this was not quite significant (ANOVA, F (2,59) =0.210, p=0.056). Knockdown of Ank2 with the stronger g lobe driver R16A06-GAL4 did impair LTM significantly (ANOVA, F (2,33) =18.57, p<0.0001, post-hoc Tukey's HSD, **p<0.01).