RNA-targeted therapy corrects neuronal deficits in PACS1 syndrome mice

Neurodevelopmental disorders (NDDs) are frequently associated with dendritic abnormalities in pyramidal neurons that affect arbor complexity, spine density, and synaptic communication 1,2. The underlying genetic causes are often complex, obscuring the molecular pathways that drive these disorders 3. Next-generation sequencing has identified recurrent de novo missense mutations in a handful of genes associated with NDDs, offering a unique opportunity to decipher the molecular pathways 4. One such gene is PACS1, which encodes the multi-functional trafficking protein PACS1 (or PACS-1); a single recurrent de novo missense mutation, c607C>T (PACS1R203W), causes developmental delay and intellectual disability (ID) 5,6. The processes by which PACS1R203W causes PACS1 syndrome are unknown, and there is no curative treatment. We show that PACS1R203W increases the interaction between PACS1 and the α-tubulin deacetylase HDAC6, elevating enzyme activity and appropriating control of its posttranscriptional regulation. Consequently, PACS1R203W reduces acetylation of α-tubulin and cortactin, causing the Golgi to fragment and enter developing neurites, leading to increased dendrite arborization. The dendrites, however, are beset with diminished spine density and fewer functional synapses, characteristic of ID pathology. Treatment of PACS1 syndrome mice with PACS1- or HDAC6-targeting antisense oligonucleotides restores neuronal structure and synaptic transmission, suggesting PACS1R203W/HDAC6 may be targeted for treating PACS1 syndrome neuropathology.

microscopy ( Fig. 1c). Golgi positioning was rescued by K 40 Q-α-tubulin but not K 40 R-α-tubulin, suggesting the reduced level of Ac-Lys 40 -α-tubulin in PACS1 syndrome patient cells causes both the Golgi fragmentation and disorganized MTs.
Acetylation of α-tubulin is regulated by the class II lysine deacetylase HDAC6 and the class III enzyme SIRT2 8 . Co-immunoprecipitation (co-IP) studies showed that the R203W mutation increased the interaction between PACS1 and HDAC6 but had no effect on the interaction with SIRT2 (Figs. 1d and e).
Next, PACS1 R203W fibroblasts and their parental controls were treated with selective inhibitors of HDAC6 or SIRT2 activity 9 . Only inhibition of HDAC6 rescued Golgi positioning (Fig. S2a). The importance of PACS1 R203W and HDAC6 to the Golgi fragmentation was confirmed with siRNA knockdown. Depletion of either protein restored Golgi positioning in the PACS1 R203W patient cells (Figs. 1f and S2b).
The physiologic interaction between PACS1 and HDAC6 was established by co-IP of the endogenous proteins (Fig. 1g). We next investigated how PACS1 R203W and HDAC6 interact to affect enzyme activity. Of the 18 HDACs, only HDAC6 possesses two catalytic domains, CD1 and CD2 (Fig. 1h). CD2 accounts for the protein's physiological tubulin deacetylation activity whereas the biological function of CD1 is largely unknown 10 . Co-IP between PACS1 and a battery of HDAC6 deletion mutants showed that PACS1 interacts with the HDAC6 CD1 region. A reciprocal mapping strategy showed HDAC6 interacts with the PACS1 FBR, which harbors the disease-causing R203W change and binds numerous client proteins as well as trafficking adaptors ( Fig. 1i) 6,11 . In addition, HDAC6 interacts with the PACS1 CTR, a region of unknown function but contains the sequence that defines the PACS-1 family (Pfam #PF10254). Surprisingly, the R203W change reduced the interaction between HDAC6 and a truncated PACS1 containing only the FBR (Fig. S2c), suggesting the mutation allosterically increases the interaction between HDAC6 and full-length PACS1 R203W (see panel 1D). The effect of PACS1 R203W on HDAC6 activity was assessed in an in vitro assay, following isolation of HDAC6 from cells expressing HDAC6 alone or together with PACS1 or PACS1 R203W (Fig. 1j). HDAC6 activity was greater following co-expression with PACS1 R203W compared to PACS1, consistent with the reduced acetylation of α-tubulin in patient cells. Together, these findings suggest the R203W mutation increases the multi-domain interaction between PACS1 R203W and HDAC6 to aberrantly potentiate deacetylase activity.
To assess the impact of PACS1 R203W in vivo, we used CRISPR/Cas9 gene editing to knock-in Creinducible, HA-tagged human PACS1 or PACS1 R203W at the murine Rosa26 safe harbor locus 12 (Fig. S3a).
The resulting R26 P1 and R26 P1R203W lines were crossed with Emx1 Cre mice to induce cassette expression in excitatory pyramidal neurons of the hippocampus and cerebral cortex, a key site of autism/ID pathophysiology 2,13 . Western blot and immunohistochemical staining showed Emx1 Cre faithfully induced expression of HA-tagged PACS1 or PACS1 R203W at levels similar to endogenous PACS1 in the cortex and hippocampus ( Figs. 2a and b). As premature activation of HDAC6 can interfere with radial migration of cortical neurons 14 , coronal sections were also stained for SATB2 (upper layer marker) and CTIP2 (BCL11B, layer 5 marker), revealing that Emx1 Cre -induced PACS1 R203W does not appear to disturb cortical layering ( Fig. S3b).
A close examination of the CA1 neurons in the R26 mice revealed that PACS1 concentrated in the cell body whereas PACS1 R203W appeared to redistribute along the apical dendrites (Fig. 2c). In developing neurons, the Golgi ribbon disassembles into mini-stacks, which migrate into the growing dendrites to support membrane traffic and possibly serve as secondary MTOCs 15 . As PACS1 R203W triggers Golgi fragmentation ( Fig. 1), we analyzed the effect of the R203W mutation on Golgi positioning and dendrite morphology in dissociated CA1 hippocampal neurons. In R26 P1 neurons, the Golgi remained in the cell body together with PACS1, as observed in vivo (Figs. 2d and S3c ). By contrast, in R26 P1R203W neurons, the Golgi fragmented and deployed to the base of the primary neurite and collected, with PACS1 R203W , in varicosities along the developing axon and dendrites, which were also longer than those of R26 P1 neurons.
We next evaluated the impact of PACS1 R203W /HDAC6 on dendrite arborization in vivo. Specifically, R26 P1 and R26 P1R203W mice were co-injected at post-natal day 1 (P1) with Cre-inducible AAV-FLEX-tdTomato, to space fill neurons, together with an HDAC6-specific antisense oligonucleotide (H6ASO) or a negative control ASO (nASO) (Fig. 2e). A single injection of the H6ASO was found to effectively and durably deplete cerebral Hdac6 mRNA and protein and correspondingly increase the level of Ac-Lys 40 -α-tubulin (Figs. 2e and S3d). Sholl analysis of the tdTomato + CA1 neurons at P18 revealed that PACS1 R203W increased dendritic arbor complexity, both in dendrite length and the number of branch points (Fig. 2f). Importantly, the H6ASO reversed the overbranching, resulting in a dendritic arbor indistinguishable from that detected in PACS1 mice.
Examination of the tdTomato + secondary apical dendrites revealed PACS1 R203W markedly reduced spine density and induced accumulation of varicosities, characteristic of ID and neurodegenerative disorders (Fig. 3a) 16 . The H6ASO fully restored spine density to levels observed in PACS1 mice, suggesting the PACS1 R203W /HDAC6 complex functions in dendritic spines. In support of this possibility, confocal analysis of DIV18 hippocampal neurons revealed a pronounced co-localization of HA-tagged PACS1 or PACS1 R203W with the glutamatergic spine maker, PSD95 (Fig. 3b). In addition, endogenous PACS1 and HDAC6 were found to fractionate with synaptosomes and to co-localize in PSD95 + dendritic spines (Figs. 3c and S4a), consistent with a report that the Pacs1 mRNA belongs to a small subset of neuronal transcripts that localize to hippocampal synapses 17 . Notably, PACS1 R203W reduced the level of acetylated cortactin (Ac-cortactin), which has a structural role in synapses and is an HDAC6 substrate (Fig. 3d) 18 . As Ac-cortactin stabilizes the PSD95 scaffold that organizes cell-surface AMPA receptors (AMPARs) 19 , these findings suggest PACS1 R203W and HDAC6 traffic to dendritic spines where they trigger excessive deacetylation of cortactin to disturb synapse organization.
The lower spine density in R26 P1R203W mice suggested that the R203W mutation reduces the number of functional synapses. To test this possibility, brain slices were prepared from juvenile R26 P1 or R26 P1R203W mice that had been treated on P1 with H6ASO or nASO (Fig. 2e), and miniature postsynaptic current (mPSCs) were recorded from L2/3 pyramidal neurons within the medial prefrontal cortex (mPFC) 20 . PACS1 R203W markedly reduced AMPAR-mediated glutamatergic excitatory mPSC (mEPSC) amplitude and frequency but had no effect on GABAergic inhibitory mPSCs (mIPSCs, Figs. 3e and S4b). These findings are consistent with earlier reports that HDAC6-dependent deacetylation of presynaptic α-tubulin reduces mEPSC frequency whereas deacetylated cortactin reduces clustering of the PSD95 postsynaptic scaffold, which, in turn, suppresses mEPSC amplitudes 18,19 . Importantly, the H6ASO treatment in R26 P1R203W mice fully restored mEPSC amplitude and frequency to levels observed in R26 P1 mice, which were unaffected by HDAC6 depletion (Fig. 3e). The reduced mEPSCs implies a broader impact of PACS1 R203W /HDAC6 on neuronal communication, considering that mEPSCs are not only a measure of functional synapses, but also modulate action potential firing and influence the maturation and stability of synaptic networks; reductions in mEPSCs in ID interfere with local translation 21,22 . Thus, the ability of the H6ASO to restore mEPSCs in R26 P1R203W mice suggests targeting HDAC6 may significantly benefit synaptic networks in PACS1 syndrome.
The R26 knock-in strategy, while accelerating our ability to identify a key molecular pathway underpinning PACS1 syndrome, prevented testing whether PACS1 R203W itself can be safely and effectively targeted with ASOs, which would eliminate all molecular pathways disturbed by the disease mutation.
We, therefore, generated PACS1-deficient mice (Pacs1 Δ4bp/Δ4bp (Pacs1 KO )) to determine whether loss of Pacs1 affects viability (Fig. S5a). Mating of heterozygous Pacs1 +/Δ4bp (Pacs1 HET ) mice demonstrated that loss of Pacs1 reduced survival at weaning, and the surviving pups were smaller than their WT littermates but were viable and fertile (Table S1 and Fig. S5b). Western blot analysis revealed that loss of Pacs1 increased the levels of Ac-cortactin in brain (Fig. S5c). Similarly, the level of Ac-Lys 40 -α-tubulin was higher in Pacs1 KO embryonic fibroblasts (MEFs) than in WT MEFs (Fig. S5d). Interestingly, loss of Pacs1 was coupled to an increased expression of its paralogue, Pacs2, both in the brain and in MEFs, suggesting PACS2 buffers an essential function of PACS1 (Figs. S5d and e). In support of this possibility, crossing Pacs1 HET ;Pacs2 HET double heterozygotes failed to produce double knockout pups (Table S2). Next, the effect of Pacs1 loss on synaptic transmission was evaluated in the L2/3 mPFC. Loss of Pacs1 had no effect on mEPSC or mIPSC amplitude or frequency (Figs. S5f and g). Together, these findings suggest Pacs1 loss does not significantly disturb excitatory or inhibitory synapses, supporting development of a PACS1 ASO strategy. They further suggest that PACS1 is an in vivo HDAC6 modulator, and that PACS2 compensates, to some extent, for loss of PACS1.
We found that a standard germline knock-in of R201W (equivalent to human R203W) caused embryonic lethality. Therefore, we adapted a modified gene trap strategy to conditionally express the knocked-in mutation from the endogenous Pacs1 locus (Fig. S6a). We also generated Pacs1 WT mice harboring the gene trap while retaining Arg 201 (Pacs1 M/+ ), to control for haploinsufficiency. The genetrapped lines were crossed with Emx1 Cre mice, permitting a direct comparison with the R26 lines, and which correctly removed the traps (Fig. S6a). Next, the effect of the R201W mutation on basal synaptic transmission was evaluated in L2/3 mPFC pyramidal neurons from the Emx1 Cre -induced Pacs1 R201W/+ and Pacs1 M/+ mice. PACS1 R201W reduced mEPSC amplitude and frequency but had no effect on mIPSCs, similar Recent studies suggest PACS1 forms a functional hub with PACS2 and WDR37, connecting the cellular chaperome to the membrane trafficking machinery 11,23,24 . Indeed, recurrent missense mutations in each gene cause overlapping neurodevelopmental deficits 5,25,26 . In support of this model, ASO depletion of Pacs1 profoundly reduced WDR37 protein levels (Fig. 4b). Similar results were observed in Pacs1 KO mice ( Fig. S5e). Interestingly, PACS1 and WDR37 modulate calcium signaling by the ER, which is the cellular store that supplies calcium for action potential-independent mEPSCs 21,27 . However, an aberrant PACS1 R203W /WDR37 interaction alone is unlikely to be the sole underlying cause of PACS1 syndrome since the reduced mEPSC phenotype observed in R26 P1R203W mice was rescued by HDAC6 ASO, despite the presence of PACS1 R203W and WDR37 (Fig. 2e). Interestingly, PACS1 R201W suppressed PACS2 levels, which were restored by the P1ASO (Fig. 4b). Considering Pacs2 buffers the loss of Pacs1 (Table S2), these findings suggest Pacs1 R201W prevents induction of a compensatory role for PACS2. Strikingly, the Pacs1 ASO also reduced HDAC6 protein levels only in the Pacs1 R201W/+ mice (Fig. 4b), suggesting the benefit achieved by the P1ASO on mEPSCs resulted, at least in part, from the coupled reduction of Pacs1 R201W and HDAC6.
Our studies suggest that a pathogenic interaction between PACS1 R203W and HDAC6 underpins PACS1 syndrome. The R203W change increases the multi-domain interaction between PACS1 and HDAC6, potentiating its deacetylase activity and pirating its posttranscriptional regulation. Together, PACS1 R203W and HDAC6 increase dendrite complexity and dendritic varicosities but reduce spine density and the number of functional synapses (see Fig. 4g). Speculatively, this PACS1 R203W /HDAC6-induced dendritic overbranching, coupled with reduced spine density, suggests PACS1 R203W may induce early onset excitotoxic epileptic seizures that resolve postnatally when spine density is reduced 32,33 . Indeed, PACS1 syndrome children frequently present with ID and early onset epilepsy that subsequently resolves, and both pathologies are associated with reduced spine density 16,34 . The PACS1 R203W -associated varicosities suggest a possible PACS1 R203W /HDAC6-induced degenerative process that is frequently described in epilepsy, infantile neurobehavioral failure, and neurodegenerative disorders [35][36][37] . The co-localization of PACS1 R203W with HDAC6 or p62 in aggresome-like structures supports this possibility. Finally, our study suggests PACS1 syndrome may be treatable with targeted therapies, including HDAC6 inhibitors or RNAbased strategies that deplete HDAC6 or PACS1 9,38 . We anticipate similar strategies may be applied to combat the many other recently identified NDDs also caused by recurrent missense mutations.  Table S3.
ICV injections were performed on genotyped P1 pups as described 44 . Pups were cryo-anesthetized for 8 min, and the injection site was located ~0.25 mm lateral to the sagittal suture and 0.5-0.75 mm rostral to the neonatal coronary suture. ASOs diluted to 20 μg/μl in PBS (no calcium or magnesium) containing 0.01% Fast Green were drawn into a Hamilton syringe with a 32g needle, positioned perpendicular to the skull surface, and impaled to a depth of 2 mm into the right hemisphere to deliver a single dose of 40 µg (2 µl injection) of ASO into the ventricle. After 10 seconds, the needle was slowly retracted to prevent backflow. The pups were gently warmed on a heating pad until fully recovered, returned to the nest, and monitored daily.   Table S3. All reactions were performed in triplicate. qPCR primers used in this study are listed in Table S4.
Brain harvest and lysate preparation. Animals were anesthetized with isoflurane and then decapitated.
(j) Cells expressing HDAC6-FLAG alone or together with HA-tagged PACS1 or PACS1 R203W were lysed, and HDAC6-FLAG was immunoprecipitated with M2 agarose. Bound proteins were eluted with FLAG peptide, and equal amounts of HDAC6 from each sample .were used to measure deacetylase activity using the Fluor-de-Lys substrate. Data are mean ± SD, n=3.       (Bottom) Quantification of Golgi fragmentation and dispersal, as described in Fig. 1. Data are mean ± SEM, n= 30-100 cells for each condition from 3 independent experiments. (b) Confocal images of 160 cells nucleofected with the NS, PACS1 or HDAC6 siRNAs described in Figure 1F. (c) HCT116 cells co-expressing HDAC6-V5 and the FLAG-tagged PACS1 1-266 (construct D in Figure 1I) or R203W-PACS1 1-266 constructs were harvested and FLAG-tagged proteins captured with M2 agarose. Co-precipitating HDAC6-V5 was detected by Western blot. Data are mean ± SD, n = 3.