Preclinical evaluation of antisense oligonucleotide therapy in a mouse model of HNRNPH2-related neurodevelopmental disorder

Summary

Mutations in HNRNPH2 cause an X-linked disorder characterized by developmental delay, intellectual disability, motor and gait disturbances, and seizures. Murine models that reproduce key clinical features of HNRNPH2-related neurodevelopmental disorder suggest that it may result from a toxic gain of function of the mutant protein or a complex loss of normal HNRNPH2 function with impaired compensation by its homolog, HNRNPH1. In this study, we tested gapmer antisense oligonucleotides (ASOs) that target murine Hnrnph2 in a non-allele-specific manner. The lead ASO reduced Hnrnph2 mRNA and protein levels while inducing compensatory upregulation of Hnrnph1 in both WT and Hnrnph2 mutant mouse brains. A single intracerebroventricular injection of the Hnrnph2 ASO into neonatal mutant Hnrnph2 mice rescued molecular and audiogenic seizure phenotypes and improved motor and cognitive functions. ASO treatment at the juvenile stage also rescued audiogenic seizures and motor deficits. In contrast, Hnrnph2 ASO administration did not improve survival, body weight, or hydrocephalus. In human iPSC-derived neurons, a human-specific HNRNPH2 research ASO reduced HNRNPH2 and upregulated HNRNPH1 mRNA levels. Mechanistically, we demonstrate that HNRNPH1 expression is regulated by alternative splicing and that HNRNPH2 modulates this process. These findings provide preclinical proof of concept for HNRNPH2 ASO therapy and offer insights into its underlying molecular mechanism.

One Sentence Summary:

ASO-mediated Hnrnph2 knockdown induces Hnrnph1 upregulation and rescues phenotypes in a mouse model of HNRNPH2-related neurodevelopmental disorder.

Introduction

HNRNPH2-related disorder is a rare neurodevelopmental disorder (NDD) caused by mutations in HNRNPH2, which is located on the X chromosome (1). HNRNPH2 is part of the HNRNP F/H family of RNA-binding proteins, which also includes HNRNPH1, HNRNPH3, HNRNPF, and GRSF1. Together, these proteins play a critical role in RNA maturation by regulating alternative splicing, 5′ capping and 3′ polyadenylation of RNAs, and RNA export (2). HNRNP F/H proteins contain a proline-tyrosine nuclear localization sequence (PY-NLS) within their glycine-tyrosine-arginine-rich (GYR) domain, which binds karyopherin β2 (Kapβ2) to regulate their nucleocytoplasmic transport (3).

The first six HNRNPH2-NDD cases reported in 2016 were all females carrying one of three de novo mutations. Subsequent studies have expanded the genotypic spectrum to include 11 distinct de novo variants (4) as well as some maternally inherited cases (5, 6). Furthermore, although HNRNPH2 mutations were initially thought to be embryonically lethal in males, several male patients have since been identified (7-9), but they remain in the minority. Most individuals identified to date carry a nonsynonymous single nucleotide variant within or adjacent to the NLS of HNRNPH2, with the 2 most common missense variants, R206W and R206Q, located within the NLS. Another variant located within the NLS, P209L, has been found in only one patient to date.

The phenotypic spectrum of HNRNPH2-NDD varies considerably between individuals. This can be attributed in part to sex/dosage effects and the location of specific mutations (4, 6, 10). However, other factors may play a role, as female patients carrying the same variant show considerable differences in their symptom severity. In this regard, skewed X-inactivation has emerged as a possible contributor to phenotypic variation (5). Most HNRNPH2-NDD patients exhibit some combination of developmental delay, intellectual disability, language impairment, motor function deficits, and growth and musculoskeletal problems. More minor features associated with the disorder include facial dysmorphia, acquired microcephaly, epilepsy, gastrointestinal disturbances, neuropsychiatric diagnoses, and cortical visual impairment. Although rare, premature death due to early stroke or seizure has been reported (4, 11). Currently, there is no FDA-approved treatment addressing the underlying mechanism of HNRNPH2-NDD, and management of the disorder remains focused on symptomatic treatment (11).

In our previous study, we developed and characterized two knockin mouse models of HNRNPH2-NDD, one carrying the most prevalent mutation, R206W, and the other carrying the rare P209L mutation. Both models faithfully recapitulated several features of the human clinical syndrome, including reduced survival in male mice, impaired motor and cognitive functions, and increased susceptibility to audiogenic seizures. These phenotypes showed mutation- and dosage-dependent effects, with P209L mice being more severe than R206W mice, and hemizygous males more severely affected than heterozygous females (12). In parallel, we characterized an Hnrnph2 KO mouse, which were phenotypically normal. Interestingly, we found that expression of Hnrnph1 was significantly increased in Hnrnph2 KO mice but not in Hnrnph2 R206W or P209L knockin mice. HNRNPH1 is a paralog of HNRNPH2, with 96% homology at the amino acid level, and the proteins are believed to play similar, and potentially redundant, roles in RNA processing and splicing (13). The expression of the two genes is similar in terms of spatial CNS expression, but different with regard to developmental regulation (12). Notably, mutations in HNRNPH1 have been linked to a neurodevelopmental syndrome very similar to HNRNPH2-NDD (14, 15). Together, these data suggested that increased Hnrnph1 expression compensates for the loss of Hnrnph2 in KO mice, likely accounting for the lack of any observable phenotypes. The fact that this increased expression is not observed in Hnrnph2 knockin mutant mice suggested two possible mechanisms as drivers of disease. The first possibility is a toxic gain of function. Indeed, we observed a decrease in nuclear levels of mutant HNRNPH2 protein in both mouse brains and human cells, with a concomitant increase in cytoplasmic expression. However, the level of mislocalization was modest, with most of the mutant protein remaining in the nucleus. Instead, our data favored an alternative pathological mechanism of complex loss of function of HNRNPH2 driven by failure of HNRNPH1 compensation. Importantly, both of these possible gain-of-function and loss-of-function mechanisms are predicted to respond positively to therapies designed to deplete expression of HNRNPH2 while simultaneously upregulating HNRNPH1. In the current study, we tested this hypothesis by treating mice with a gapmer antisense oligonucleotide (ASO) designed to target murine Hnrnph2 mRNA in a non-allele-specific manner for RNase H degradation. We found that intracerebroventricular (ICV) injection of the Hnrnph2 ASO was well tolerated and resulted in dose-dependent decreases in Hnrnph2 expression and increases in Hnrnph1 expression in the mouse brain. We further found that both neonatal and juvenile treatment with the ASO rescued or improved multiple phenotypes of Hnrnph2 mutant P209L and R206W knockin mice. These results provide proof of principle that knockdown of Hnrnph2 to sufficient levels leads to a compensatory increase in Hnrnph1 expression, is well tolerated, and is a potential therapeutic strategy for HNRNPH2-NDD. Importantly, as the ASO targets both WT and mutant Hnrnph2, this strategy may prove effective for HNRNPH2-NDD patients regardless of their specific mutation.

Results

ASO targeting mouse Hnrnph2 in a non-allele-specific manner knocks down Hnrnph2 and upregulates Hnrnph1.

To develop ASOs that effectively knock down Hnrnph2 in our mouse model of HNRNPH2-NDD, we designed and tested more than 450 gapmer ASOs targeting murine Hnrnph2 in a non-allele-specific manner for degradation by RNase H. ASOs were initially screened in primary mouse cortical neurons at 20 μM, and the most active ASOs were subsequently evaluated in a 4-point dose-response curve ranging in concentration from 1.2 to 32 μM. The top 19 most efficacious ASO candidates in vitro were subsequently selected for an in vivo 8-week activity and tolerability assessment in WT mice. Briefly, adult C57BL/6J female mice received a single bolus ICV injection of 700 μg Hnrnph2 ASO or vehicle (PBS) control and were assessed for neurological function and behavior, weight gain, and markers of neuroinflammation, including Aif1, IBA1, and CD68 (table S1). The ASOs demonstrated variable efficacy, with greater cortical Hnrnph2 mRNA reduction associated with greater Hnrnph1 mRNA upregulation (fig. S1A). The most potent Hnrnph2 ASO, ASO1, knocked down Hnrnph2 mRNA levels by 96% (fig. S1B) while increasing Hnrnph1 mRNA by 63% (fig. S1C). ASO1 (hereafter referred to as Hnrnph2 ASO) also showed a favorable tolerability profile, including no significant effect on markers of neuroinflammation or body weight (fig. S1, D and E), and was selected for further studies.

Potency and duration of action of Hnrnph2 ASO in mouse brain.

To test the potency of the Hnrnph2 ASO in vivo, we performed ICV injections of WT C57BL/6J pups with a single bolus of 1.875-30 μg of non-targeting control ASO or Hnrnph2 ASO at postnatal day (PND) 2-4, followed by whole brain harvest at age 3 weeks (Fig. 1A). The Hnrnph2 ASO significantly reduced Hnrnph2 mRNA levels in a dose-dependent manner as measured by ddRT-PCR, starting with the lowest dose tested (Fig. 1B). As previously demonstrated (12), depletion of Hnrnph2 increased Hnrnph1 mRNA levels in a dose-dependent manner, starting from the second lowest dose tested (Fig. 1C). As all doses of control ASO and Hnrnph2 ASO were well tolerated in this cohort, the highest dose (30 μg) was selected for further neonatal treatment studies.

Fig. 1. Potency, duration, distribution, and tolerability of neonatal Hnrnph2 ASO treatment in C57BL/6J mice.

(A) Design of studies used to generate data in (B-M). All panels show data collected from C57BL/J6 mice at 3 weeks of age (except as indicated in F, G, and K) and with 30 μg dose of ASO (except as indicated in B). (B-C) ddRT-PCR of whole brain tissues showing Hnrnph2 and Hnrnph1 mRNA levels with increasing doses of ASO. n = 3-4 mice per dose. (D-E) Immunoblot of whole brain tissues showing HNRNPH2 and HNRNPH1 protein levels in indicated lysate fractions. n = 4 mice per treatment. (F-G) ddRT-PCR of whole brain tissues showing Hnrnph2 and Hnrnph1 mRNA levels at indicated time points. n = 2-4 mice per age. (H) ddRT-PCR of whole brain tissues showing Hnrnph2 and Hnrnph1 mRNA levels in mice treated with ASO or vehicle. n = 4 mice per treatment. (I) Left, low-magnification image of mouse brain showing ASO distribution (scale bar, 300 μm). OB: Olfactory bulb. NCX: Neocortex. T: Thalamus. SC: superior colliculus. Cb: Cerebellum. P: Pons. Right, higher-magnification images (inset scale bar, 50 μm) of hippocampus. (J) Staining of neocortex with antibodies against ASO, HNRNPH2, and the neuronal marker NeuN. Scale bar, 7 μm. (K) Kaplan-Meier curves for indicated genotypes and treatments. (L) Body weight for indicated genotypes and treatments. n = 8-12 mice per treatment. (M) ddRT-PCR of whole brain tissues showing Gfap, Aif1, and Cd68 expression. n = 3-4 mice per treatment. For all statistical analyses, ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05. For panels (D), (E), (H), (M): 1-way ANOVA with Tukey's multiple comparison. For panels (B), (C), (F), (G), (L): 2-way ANOVA with Tukey's multiple comparison. All bar graphs show data as mean ± SEM.

The changes in Hnrnph2 and Hnrnph1 mRNA levels with 30 μg Hnrnph2 ASO were mirrored by changes in protein expression as measured by Western blot of nuclear and cytoplasmic fractions of whole brain tissue at 3 weeks of age. Due to its low expression levels, HNRNPH2 protein was detectable by an HNRNPH2-specific antibody (12) (fig. S2) only in nuclear fractions, where it was significantly reduced in Hnrnph2 ASO-treated samples compared to control ASO- and vehicle-treated samples (Fig. 1D and fig. S2). In contrast, HNRNPH1 was readily detected in whole-cell RIPA fractions using an HNRNPH1-specific antibody (fig. S2) and showed a significant increase following Hnrnph2 ASO treatment (Fig. 1E and fig. S2), consistent with their respective changes in mRNA levels.

To determine the duration of action of the Hnrnph2 ASO, we performed a single-bolus ICV injection of 30 μg control ASO or Hnrnph2 ASO in C57BL/6J pups at PND 2-4, followed by whole brain harvest at 3-16 weeks of age (Fig. 1A). We observed significant knockdown of Hnrnph2 mRNA until the maximum time tested (16 weeks of age), although the level of knockdown began to decline at 12 weeks as compared to 3 weeks (Fig. 1F). Similarly, Hnrnph1 mRNA was significantly increased compared to control ASO treatment up to 16 weeks of age, with no apparent reduction in upregulation at later time points as compared to 3 weeks (Fig. 1G). As expected, no significant differences were observed between the 30 μg control ASO and vehicle treatments (Fig. 1H). These data demonstrate that the effect of the Hnrnph2 ASO is stable for at least 12 weeks in mouse brain after a single early postnatal ICV injection.

Distribution of Hnrnph2 ASO in mouse brain.

To determine the distribution of ASO in the mouse brain, we used immunofluorescent staining with a polyclonal antibody that selectively recognizes the phosphorthioate backbone of ASOs (16, 17). This antibody showed broad distribution of the Hnrnph2 ASO and control ASO across the brain (Fig. 1I). Higher magnification revealed a punctate staining pattern in Hnrnph2 ASO- and control ASO-treated samples, with no staining evident in vehicle-treated brains (Fig. 1I). Costaining with specific markers for neurons, oligodendrocytes, microglia, and astrocytes confirmed the presence of ASO in all four major cell types of the CNS (fig. S3A). Lastly, triple staining with the anti-ASO antibody, an HNRNPH2-specific antibody, and neuronal markers revealed downregulation of HNRNPH2 signal in neurons of the neocortex (Fig. 1J), hippocampus, and cerebellum (fig. S3, B and C) that were positive for the Hnrnph2 ASO. Together, these data demonstrate that a single ICV bolus injection at PND 2-4 results in broad distribution of the Hnrnph2 ASO in the mouse brain and efficient reduction of HNRNPH2 protein in all major CNS cell types.

Hnrnph2 ASO is well tolerated.

To assess the safety of neonatal treatment with the Hnrnph2 ASO, we performed ICV injection of C57BL/6J pups with a single bolus of 30 μg control ASO, Hnrnph2 ASO, or vehicle at PND 2-4. Pups were followed for survival until 3 weeks of age, at which point they were weighed and whole brains harvested. We found no significant difference in survival or body weight between any of the treatment groups for males and females (Fig. 1, K and L). As concerns have been raised regarding neuroinflammation in response to CNS administration of ASOs (18), we next examined the expression of reactive gliosis markers by ddRT-PCR using whole brain tissues collected at 3 weeks of age. Hnrnph2 ASO-treated brains showed no significant difference in mRNA levels of the astrocytic marker Gfap or the microglial marker Aif1 compared to control ASO- or vehicle-treated samples (Fig. 1M). Although Cd68 transcript levels were significantly increased in Hnrnph2 ASO-treated brains, Cd68 is more closely related to lysosomal activity than to overall microglial activation, whereas Aif1 is a more direct marker of microglial activation (19). Importantly, histological analyses revealed no overt signs of neuroinflammation or systemic toxicity in brain (table S2) or any organs (table S3) following administration of either control or Hnrnph2 ASO. Together, these data indicate that treatment with 30 μg control ASO and Hnrnph2 ASO in neonatal WT C57BL/6J mice is well tolerated. As there were no significant differences between the control ASO-treated and vehicle-treated groups, all further neonatal treatment studies were conducted with the non-targeting control ASO and Hnrnph2 ASO only (no vehicle groups) to reduce the number of animals needed to complete the study.

Neonatal treatment with Hnrnph2 ASO does not influence the survival and body weight phenotypes of mutant Hnrnph2 mice.

We next evaluated the effect of the Hnrnph2 ASO in mutant Hnrnph2 mice. For these and all subsequent experiments, we evaluated phenotypes in at least two genotype/sex groups. Notably, these mouse models have a consistent pattern of phenotype severity across all phenotypes tested: P209L hemizygous male mice are the most severely affected, followed in decreasing severity by R209W hemizygous mice and P209L heterozygous female mice; R206W heterozygous females show no significant phenotype in most tests (12). Thus, we designed our experiments to assess the effects of Hnrnph2 ASO in the most severely affected group that was testable (i.e., P209L hemizygous males in nearly all assays), along with at least one more mildly affected group.

Following the same protocol as was used for WT C57BL/6J mice, we treated Hnrnph2 P209L hemizygous males and WT littermates with a single bolus ICV injection of 30 μg control ASO or Hnrnph2 ASO at PND 2-4 and collected whole brains at 3 weeks of age (Fig. 2A). Hnrnph2 mRNA levels were significantly reduced and Hnrnph1 significantly increased in Hnrnph2 ASO-treated samples in both WT and P209L mutant mice (Fig. 2B), demonstrating that Hnrnph2 ASO effectively targets both WT and mutant alleles. Next, we treated Hnrnph2 P209L and R206W mice with Hnrnph2 ASO using the same regimen and assessed its impact on body weight and survival up to 8 weeks of age. In control ASO-treated male mice, pairwise comparisons indicated that P209L and R206W mice had significant differences in survival compared to WT mice, as previously reported (12); however, the statistical significance of these differences was lost after correction for multiple comparisons (fig. S4, A and B). In the Hnrnph2 ASO-treated group, P209L mice showed significantly decreased survival compared to WT mice even after correction for multiple comparisons (fig. S4A), whereas in R206W mice there was no significant difference in survival between Hnrnph2 ASO-treated mutant or WT males, nor was there a significant effect for Hnrnph2 ASO treatment compared to control ASO treatment in WT or R206W mice (fig. S4B). Consistent with our previous report, we found no mutation-dependent differences in survival of heterozygous P209L female mice compared to their WT littermates, regardless of ASO treatment (fig. S4C). Together these results demonstrate that neonatal Hnrnph2 ASO treatment does not rescue reduced survival of Hnrnph2 mutant mice.

Fig. 2. Neonatal treatment with Hnrnph2 ASO rescues the molecular and audiogenic seizure phenotypes of juvenile Hnrnph2 P209L mice.

(A) Design of studies used to generate data in (B-L). All panels show data collected from P209L mutant mice or littermates at 3 weeks of age treated with 30 μg dose of ASO. (B) ddRT-PCR of whole brain tissues showing Hnrnph2 and Hnrnph1 mRNA levels. (C) Seven upregulated genes in both Hnrnph2 mutant mice and human HNRNPH2 mutant iPSC-derived neurons. (D) ddRT-PCR of whole brain tissues showing expression levels of genes listed in (C). (E) Volcano plots showing differentially expressed genes with significantly downregulated (blue) and upregulated (red) genes indicated. (F) Top 10 most upregulated and 5 most downregulated genes by log2R from middle plot in (E). (G) Effects of Hnrnph2 ASO on upregulated (left), downregulated (middle), and unchanged (right) genes at the individual gene level. (H) RNA-seq-derived differential alternative splicing events grouped by Hnrnph2 ASO treatment (left) and P209L mutation (right). SE: skipped exon. A3SS: alternative 3′ splice site. RI: retained intron. A5SS: alternative 5′ splice site. MXE: mutually exclusive exon. Overlap between Hnrnph2 ASO-dependent and P209L mutation-dependent events is shown. (I) No overlap between P209L mutation-dependent events identified by differential gene expression (DGE) and differential alternative splicing (DAS) analyses. (J) Left, volcano plot showing the 185 significant skipped exon (SE) events in control ASO-treated P209L mutant mice with significantly increased (red) and decreased (blue) events. Right, volcano plot showing the effect of Hnrnph2 ASO treatment on P209L mutation-dependent skipped exon events. (K-L) Audiogenic seizure susceptibility scoring (K) and assessment (L) of indicated mice. For panel B and D, 2-way ANOVA with Sidak's multiple comparison, n = 3-4 mice per treatment. For panel L, non-parametric Scheirer–Ray–Hare test with Mood’s median test, control ASO WT males n = 20, control ASO P209L males n = 30, Hnrnph2 ASO WT males n = 20, Hnrnph2 ASO P209L males n = 29, control ASO WT females n = 26, control ASO P209L females n = 27, Hnrnph2 ASO WT females n = 30, Hnrnph2 ASO P209L females n = 19. For all statistical analyses, ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05. Data shown as mean ± SEM.

Similarly, although control ASO-treated Hnrnph2 P209L and R206W males, but not P209L females, showed reduced body weight compared to WT littermates at 8 weeks, this phenotype was not rescued by Hnrnph2 ASO treatment (fig. S4, D to F), suggesting that this aspect of the phenotype may be driven by peripheral factors that are not rescued by central administration of ASO. Importantly, there was no significant difference in body weight between control ASO and Hnrnph2 ASO-treated WT or Hnrnph2 mutant mice, indicating that neonatal treatment with the Hnrnph2 ASO is well tolerated in R206W males and P209L mice.

Neonatal treatment with Hnrnph2 ASO rescues the molecular phenotype of juvenile mutant Hnrnph2 mice.

Next, we investigated whether the Hnrnph2 ASO was able to rescue the molecular phenotype of Hnrnph2 P209L and R206W male mice previously identified by RNA-seq (12). Using whole brain samples from Hnrnph2 P209L and R206W hemizygous males and WT littermates treated with a single bolus ICV injection of 30 μg control ASO or Hnrnph2 ASO at PND 2-4, we assessed the mRNA levels of seven differentially expressed genes identified in our original study (Fig. 2C). These genes (Ctnna2, Tnpo2, Shank1, Ddn, Ppfia3, Phactr3, and Sipa1l1) were chosen based on their differential expression in 8-week-old Hnrnph2 R206W and 3-week-old P209L hemizygous male mice, and in human HNRNPH2 R206W, P209L, and R206Q induced pluripotent stem cell (iPSC)-derived neurons, while not being significantly changed in Hnrnph2 KO hemizygous male mice or human HNRNPH2 KO iPSC-derived neurons (12). Using ddRT-PCR, we found that neonatal Hnrnph2 ASO treatment restored the expression of all 7 genes in Hnrnph2 P209L mutant males to WT levels at 3 weeks of age (Fig. 2D). A similar rescue of gene expression was observed in Hnrnph2 R206W mutant males following neonatal Hnrnph2 ASO treatment (fig. S5, A to D). Notably, in WT mice, Hnrnph2 ASO treatment did not change the mRNA levels of any of the seven genes tested, likely due to functional compensation by upregulation of Hnrnph1 in response to reduced Hnrnph2 (Fig. 2D and fig. S5D).

To examine the broader impact of neonatal ASO treatment, we performed RNA-seq analyses on whole brains from 3-week-old Hnrnph2 P209L and R206W male mice treated with a single bolus ICV injection of 30 μg control ASO or Hnrnph2 ASO at PND 2-4. In WT mice, only three genes, including Hnrnph2 and Hnrnph1, were differentially expressed following Hnrnph2 ASO treatment compared to control ASO (Fig. 2E, left), supporting our hypothesis of functional compensation by Hnrnph1 upregulation in response to reduced Hnrnph2 and consistent with the previously reported absence of phenotypes in Hnrnph2 KO mice (12). Moreover, the lack of significant changes in the expression of other genes following Hnrnph2 ASO treatment supports the specificity and safety of the ASO. Within the control ASO-treated groups, P209L male mice had 41 differentially expressed genes compared with WT mice, with 36 genes upregulated and only 5 genes downregulated (Fig. 2E, middle, and Fig. 2F), consistent with a prior report that HNRNPH1/2 primarily act as negative regulators of target gene expression (20). Importantly, there was only 1 significantly differentially expressed gene in P209L samples compared to WT samples after Hnrnph2 ASO treatment (Fig. 2E, right). At the individual gene level, the expression of both upregulated and downregulated genes in P209L mice was restored to WT levels with Hnrnph2 ASO (Fig. 2G), indicating that neonatal Hnrnph2 ASO treatment broadly rescues the molecular phenotypes observed in juvenile Hnrnph2 P209L mice.

We obtained similar results from juvenile Hnrnph2 R206W male mice after neonatal Hnrnph2 ASO treatment. Consistent with their milder behavioral phenotype compared to P209L mutant mice (12), R206W mutant mice had only 14 differentially expressed genes compared to WT mice in the control ASO-treated group, with 13 genes significantly upregulated, and only 1 gene downregulated (fig. S5, E middle, and F). All of these expression changes were restored to WT levels following Hnrnph2 ASO treatment (fig. S5G). We note that Hnrnph1 transcript levels were not significantly increased in this RNA-seq analysis of samples treated with Hnrnph2 ASO, although a significant increase was detected by ddRT-PCR assay from the same samples (fig. S5B). This discrepancy could be due to the targeted nature and higher sensitivity of ddRT-PCR, which is better able to detect subtle changes in transcript abundance compared to the more global approach of RNA-seq. Taken together, these results demonstrate that neonatal Hnrnph2 ASO treatment is effective in correcting the gene expression profile of both Hnrnph2 P209L and R206W mutant juvenile males, bringing the expression of all significantly differentially expressed genes back to WT levels.

Given the important role of HNRNPH1/2 in regulating alternative splicing (2), we also analyzed the RNA-seq data for differential alternative splicing (DAS) events. In contrast to differential gene expression (DGE) (Fig. 2, E to G and fig. S5, E to G), Hnrnph2 ASO treatment in WT mice resulted in a substantial number of DAS events. This may reflect incomplete compensation for HNRNPH2 specific splicing events by HNRNPH1 and/or the introduction of novel, HNRNPH1-specific splicing events due to its upregulation. The most common DAS event in Hnrnph2-treated WT mice was skipped exons (SE), accounting for 65% of all events, with 81 SE events upregulated and 41 downregulated (Fig. 2H, left). In the control ASO-treated group, P209L mutant mice also exhibited a substantial number of DAS events compared to WT mice, again with SE events (81%) being the most common (Fig. 2H, right). Notably, there was no overlap between differentially expressed genes and differentially spliced genes, suggesting that direct alternative splicing of the identified DGE genes is not the primary mechanism underlying changes in gene expression by the P209L mutation (Fig. 2I).

We next assessed whether Hnrnph2 ASO treatment could rescue the DAS events induced by the P209L mutation by examining individual splicing events. To do this, we focused on SE events, as they represent the majority of DAS events. Control ASO-treated P209L mutant mice had 185 significantly altered SE events compared with control ASO-treated WT mice, including 105 upregulated and 80 downregulated events (Fig. 2J). Hnrnph2 ASO treatment restored 104 of these SE events (~56%) to levels not significantly different from WT, while 56 events (~30%) remained differentially spliced (Fig. 2J). Twenty-five SE events (~14%) were either not detected by RNA-seq or filtered out from the final dataset during data analysis due to low read counts or extreme percent spliced in (PSI) values (21).

Control ASO-treated R206W mutant male mice exhibited 549 DAS events compared to control ASO-treated WT mice, with SE events comprising the majority (459, ~84%) (fig. S5H), a substantial increase in total DAS events that is likely attributable to batch effects, given a similar increase in control ASO-treated WT mice in this experiment. Only 1 gene, Stmn4, was both differentially expressed and differentially spliced (fig. S5I). This minimal overlap between DEG genes and DAS genes suggests that direct alternative splicing of the DEG genes is not the primary mechanism underlying changes in gene expression by the R206W mutation. Following Hnrnph2 ASO treatment, approximately 52% of significant SE events were restored to levels not significantly different from WT, 35% remained differentially spliced, and 14% were not detected or filtered out from the final dataset during data analysis (fig. S5J). Together, these data suggest that Hnrnph2 ASO treatment is effective in rescuing a substantial proportion of the alternative splicing abnormalities caused by both the P209L and R206W mutation, although some splicing defects persist.

Neonatal treatment with Hnrnph2 ASO rescues the audiogenic seizure phenotype of juvenile Hnrnph2 P209L mice.

To further investigate the potential therapeutic effects of knocking down Hnrnph2, we tested whether neonatal Hnrnph2 ASO treatment could rescue behavioral deficits associated with mutant mice. In our initial characterization of Hnrnph2 mutant mice, we identified increased susceptibility to audiogenic seizures as a robust behavioral phenotype. Audiogenic seizures are generalized seizures triggered by exposure to high-intensity sound and serve as a model to study sensory hypersensitivity and sudden unexpected death in epilepsy (SUDEP; (22, 23)). In that study, Hnrnph2 P209L hemizygous males were the most severely affected, followed by Hnrnph2 R206W hemizygous males, and Hnrnph2 P209L heterozygous females with a mild phenotype (12). Hnrnph2 R206W heterozygous females did not have a significant increase in audiogenic seizure susceptibility compared to their WT littermates (12). To assess the impact of ASO treatment on susceptibility to audiogenic seizure, we treated P209L hemizygous males, P209L heterozygous females, and R206W hemizygous males, as well as their WT littermates, with a single bolus ICV injection of 30 μg control ASO or Hnrnph2 ASO at PND 2-4. Mice were maintained until 3 weeks of age, at which point they were tested for audiogenic seizure susceptibility (Fig. 2A). We found that neonatal Hnrnph2 ASO treatment rescued the increased audiogenic seizure susceptibility in juvenile Hnrnph2 P209L hemizygous males, with no statistically significant difference between WT and mutant mice in the Hnrnph2 ASO-treated group (Fig. 2 K and L). The milder phenotype of Hnrnph2 P209L heterozygous females was also improved following neonatal Hnrnph2 ASO treatment (Fig. 2, K and L). Although the seizure score of control ASO-treated R206W hemizygous males tended to be increased compared to their WT littermates, it failed to reach statistical significance and neonatal Hnrnph2 ASO treatment had no significant effect (fig. S5, K and L).

We previously identified an increased incidence of hydrocephalus as a prominent phenotype in Hnrnph2 mutant mice (12). Although there is no evidence of hydrocephalus in patients with HNRNPH2 mutations, acquired microcephaly is present in approximately half of cases (4). Although the two conditions are typically viewed as mutually exclusive, cases of microcephalic hydrocephalus suggest potential overlap in their pathogenesis, including impaired fetal neural stem cell proliferation (24). In our mouse models, hydrocephalus exhibited incomplete penetrance and variable presentation: some mice presented with domed heads, whereas others had normal skulls but ventricular enlargement by MRI scan (12). Using the same neonatal treatment regimen as mice tested for audiogenic seizure (Fig. 2A), we found that Hnrnph2 ASO treatment did not rescue the increased incidence of hydrocephalus in Hnrnph2 P209L hemizygous males (fig. S6A). Hnrnph2 P209L heterozygous females did not show an increased incidence of hydrocephalus compared to their WT littermates, regardless of treatment (fig. S6A). Although control ASO- and Hnrnph2 ASO-treated R206W hemizygous males tended to have a higher incidence of hydrocephalus compared to their WT littermates, the difference was not significantly different after correcting for multiple comparisons (fig. S6B). As several studies have implicated defects in early brain development in the pathogenesis of congenital hydrocephalus (25-27), it is possible that neonatal ASO treatment may be too late to rescue the hydrocephalus phenotype in Hnrnph2 mutant mice.

Neonatal treatment with Hnrnph2 ASO improves motor function of adult Hnrnph2 P209L male and female mice.

We next assessed the effect of neonatal Hnrnph2 ASO treatment on motor phenotypes we previously identified in Hnrnph2 mutant knockin mice. To assess the impact of Hnrnph2 ASO treatment across a range of phenotype severities, we treated P209L hemizygous males (severe phenotype), P209L heterozygous females (mild phenotype), and their WT littermates with 30 μg control or Hnrnph2 ASO by single bolus ICV injection at PND 2-4. At 8 weeks old, mice were tested for balance and coordination, muscle strength, and gait changes (Fig. 3A). In the rotarod, WT mice performed significantly better than P209L mice when treated with control ASO (Fig. 3B and fig. S7A). However, after Hnrnph2 ASO treatment, the difference between WT and P209L mice was no longer significant in both males (Fig. 3B) and females (fig. S7A), suggesting that Hnrnph2 ASO treatment improves motor deficits in P209L mice. Balance beam performance was also improved in Hnrnph2 P209L hemizygous males (Fig. 3C) and rescued in Hnrnph2 P209L heterozygous females (fig. S7B): whereas control ASO-treated mutants showed increased latency to cross and more hind paw slips compared to WT littermates, Hnrnph2 ASO treatment reduced hind paw slips to WT levels, but latency to cross was not rescued in Hnrnph2 P209L hemizygous males (Fig. 3C). In females, Hnrnph2 ASO treatment restored these measures to WT levels (fig. S7B). For grip strength, Hnrnph2 P209L males exhibited slightly weaker muscle strength compared to WT in the control ASO-treated group, though this difference did not reach statistical significance. In Hnrnph2 ASO-treated males, muscle strength remained impaired (Fig. 3D). Of note, we observed that Hnrnph2 ASO-treated WT males showed significantly higher grip strength compared to control ASO-treated WT males, suggesting a possible treatment effect. Grip strength was improved in Hnrnph2 ASO-treated Hnrnph2 P209L females (fig. S7C). Finally, reduced stride length was restored in Hnrnph2 P209L males, with no significant difference between Hnrnph2 ASO-treated mutants and WT littermates, and a significant increase compared to control ASO-treated mutants (Fig. 3E). No significant stride length reduction was observed in Hnrnph2 P209L females compared to their WT littermates and Hnrnph2 ASO treatment had no effect (fig. S7D).

Fig. 3. Neonatal treatment with Hnrnph2 ASO improves motor and cognitive function in adult Hnrnph2 P209L and R206W male mice.

(A) Design of studies used to generate data in (B-K). (B-E) Effect of neonatal Hnrnph2 ASO treatment on rotarod performance (B), balance beam performance (C), grip strength (D), and stride length (E) of P209L male mice. (F-K) Effect of neonatal Hnrnph2 ASO treatment on elevated plus maze behavior (F), open field behavior (G), and Morris water maze performance (H-K) in female P209L mice. In (F), heatmaps show the average animals’ center point for the groups for the 5-minute test. In (G), heatmaps show the average animals’ center point for the groups for the 20-minute test. For panels B-H and K, ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05 by 2-way ANOVA with Sidak's multiple comparison. For panel I, **P < 0.01, *P < 0.05 control ASO WT males vs. control ASO R206W males, ##P < 0.01, #P < 0.05 Hnrnph2 ASO WT males vs. Hnrnph2 ASO R206W males, linear mixed effects model with random intercept. Post-hoc comparisons adjusted by FDR within each time point. For (B-E), control ASO WT males n = 22, control ASO P209L males n = 16, Hnrnph2 ASO WT males n = 33, Hnrnph2 ASO P209L males n = 18. For (G-K), control ASO WT males n = 27, control ASO R206W males n = 21, Hnrnph2 ASO WT males n = 23, Hnrnph2 ASO R206W males n = 24. All graphs show data as mean ± SEM.

Neonatal treatment with Hnrnph2 ASO improves cognitive function in adult Hnrnph2 R206W male mice.

The major and most common feature of HNRNPH2-NDD is intellectual disability, while anxiety and other neuropsychiatric symptoms are also often reported and of considerable concern to patients’ parents (4). In our original phenotypic characterization of the Hnrnph2 mouse lines, we found that R206W males have mildly impaired decision making related to approach/avoid conflict and impaired spatial learning and memory (12). In that study, we elected not to perform these assays in the Hnrnph2 P209L males due to their high mortality rate at age 8 weeks as well as concerns regarding their physical ability to participate in a water maze test. Although we did not assess the cognitive phenotype of Hnrnph2 P209L heterozygous female mice in the original phenotyping study, based on the pattern seen for all other tests, it was predicted to be milder than that of the Hnrnph2 R206W hemizygous males. To assess the effect of Hnrnph2 ASO treatment on the impaired decision making and spatial learning and memory previously identified in Hnrnph2 R206W males and predicted in Hnrnph2 P209L females, we injected hemizygous R206W males, heterozygous P209L females, and their WT littermates with a single ICV bolus of 30 μg control or Hnrnph2 ASO at PND 2-4. Mice were maintained until 8 weeks of age, when they were tested first in the elevated plus maze, followed by the open field test, and lastly the Morris water maze (MWM) (Fig. 3A).

In the elevated plus maze, total distance traveled, a measure associated with locomotor activity, was not significantly different between WT and R206W male mice when treated with control ASO; however, with Hnrnph2 ASO treatment, R206W males showed a significant increase compared to WT (Fig. 3F). R206W males spent significantly more time in the open arms and less time in the maze center compared to WT mice when treated with control ASO, indicative of impaired decision making related to approach/avoid conflict. In Hnrnph2 ASO-treated mutants, these parameters were restored to WT levels (Fig. 3F). However, interpretation of these results is confounded given the increased general activity in Hnrnph2 ASO-treated mutants as indicated by increased total distance traveled.

In the open field test, there were no significant differences in total distance traveled, a measure of locomotor activity, between mice regardless of genotype and treatment (Fig. 3G). With control ASO treatment, R206W mutant males spent significantly less time in the open field center compared to WT males, suggestive of increased anxiety in the mutants. While Hnrnph2 ASO-treated mutant mice spent significantly more time in the center zone compared to control ASO-treated mutant mice, the interpretation of this result is confounded by the finding that treatment with the Hnrnph2 ASO resulted in increased time in the center zone in both WT and mutant males compared to their control ASO counterparts (Fig. 3G).

We next assessed spatial learning and memory in the MWM. To ensure that the mice possessed the basic sensory-motor functions and visual ability required to complete the task, we first performed a cued version of the MWM. Swim speed, a parameter associated with locomotion, was significantly reduced in Hnrnph2 ASO-treated male mice, regardless of genotype (Fig. 3H). However, percent time spent in the thigmotaxis zone of the pool, which is associated with anxiety, and latency to find the visible platform were similar in all groups (Fig. 3H), demonstrating that ASO treatment did not alter the basic abilities, strategies, and motivation that is required for completing the task. In the training phase of the MWM, which assesses spatial learning, control ASO-treated R206W males showed a significant increase in latency to find the hidden platform and cumulative distance from the hidden platform compared with their WT littermates on days 2, 3, and 4 (Fig. 3I), despite similar swim speed and percent thigmotaxis time (Fig. 3J), suggesting a mild deficit in spatial learning. Hnrnph2 ASO treatment did not significantly improve this deficit (Fig. 3I). In the probe trial of the MWM to assess memory retention, control ASO-treated R206W males showed a significant increase in latency to first enter the platform location and a significant reduction in the number of platform location crossings compared to their WT littermates, suggesting a deficit in spatial memory (Fig. 3K). This deficit was improved by Hnrnph2 ASO treatment, with no significant difference in either parameter between Hnrnph2 ASO-treated R206W males and WT littermates (Fig. 3K), suggesting that although Hnrnph2 ASO treatment did not rescue all parameters in the MWM test, it was effective in restoring spatial memory in R206W mutant male mice.

P209L heterozygous females exhibited minimal cognitive or behavioral deficits (figs. S7, E to I). The only significant difference between WT and P209L heterozygous females was increased time spent in the open arms of the elevated plus maze (fig. S7E), indicative of increased anxiety-like behavior. This phenotype was improved by Hnrnph2 ASO treatment, with treated mice showing open arm times comparable to WT controls (fig. S7E). Hnrnph2 ASO treatment had effects in other tests that did not show any mutation-dependent phenotypes, including increases in the total distance traveled and time in center zone in the open field test (fig. S7F), decreased swim speed and increased latency to find the platform in the MWM cued phase (fig. S7G), and isolated effects in day 4 of the MWM training phase (fig. S7H).

Juvenile treatment with Hnrnph2 ASO rescues molecular and audiogenic seizure phenotypes and improves the motor phenotype of adult Hnrnph2 P209L mice.

We next tested an extended therapeutic window of Hnrnph2 ASO treatment. To do so, we first tested the potency, duration of action, and tolerability of a juvenile ASO treatment protocol in WT C57BL/6J mice. Mice received a single bolus ICV injection of 30-200 μg of non-targeting control ASO or Hnrnph2 ASO at PND 27-29, followed by whole brain harvest at age 8 weeks (Fig. 4A). Similar to results obtained from our neonatal (PND 2-4) Hnrnph2 ASO treatment protocol, mice treated with Hnrnph2 ASO at a juvenile stage showed significantly reduced Hnrnph2 mRNA levels and increased Hnrnph1 mRNA levels in a dose-dependent manner (Fig. 4B and 4C). This knockdown of Hnrnph2 mRNA and upregulation of Hnrnph1 mRNA persisted until the maximum time tested (16 weeks of age) (Fig. 4D and 4E). No significant differences were observed between 200 μg control ASO and vehicle treatments (Fig. 4F). Similar to the findings with neonatal treatment, juvenile ASO-treated mice showed upregulated Cd68 transcript levels; however, assessment of other reactive gliosis markers revealed no apparent neuroinflammation following CNS administration of ASOs at this stage (Fig. 4G), consistent with results observed in neonatally treated mice. As there were no significant differences between the control ASO-treated and vehicle-treated groups, all further juvenile treatment studies were conducted with the non-targeting control ASO only (no vehicle groups) to reduce the number of animals needed for the study.

Fig. 4. Juvenile treatment with the Hnrnph2 ASO rescues the audiogenic seizure and motor phenotype of Hnrnph2 P209L mice.

(A) Design of studies used to generate data in (B-G). In (B-G), all panels show data collected from C57BL/J6 mice at 8 weeks of age (except as indicated in D) and with 200 μg dose of ASO (except as indicated in B). (B-C) ddRT-PCR of whole brain tissues showing Hnrnph2 and Hnrnph1 mRNA levels with increasing doses of ASO. n = 3-4 mice per dose. (D-E) ddRT-PCR of whole brain tissues showing Hnrnph2 and Hnrnph1 mRNA levels at indicated time points. n = 2-4 mice per age. (F) ddRT-PCR of whole brain tissues showing Hnrnph2 and Hnrnph1 mRNA levels in mice treated with ASO or vehicle. n = 4 mice per treatment. (G) ddRT-PCR of whole brain tissues showing Gfap, Aif1, and Cd68 expression. n = 3-4 mice per treatment. (H) Design of studies used to generate data in (I-N). (I-J) Audiogenic seizure susceptibility in untreated (I) and ASO-treated (J) female mice. For (I): WT females n = 10, P209L females n = 11. For (J): control ASO WT females n = 17, control ASO P209L females n = 11, Hnrnph2 ASO WT females n = 16, Hnrnph2 ASO P209L females n = 16. (K-L) Effect of juvenile ASO treatment on rotarod (K) and balance beam (L) performance. For (K): control ASO WT females n = 16, control ASO P209L females n = 11, Hnrnph2 ASO WT females n = 23, Hnrnph2 ASO P209L females n = 19. For (L): control ASO WT females n = 15, control ASO P209L females n = 10, Hnrnph2 ASO WT females n = 23, Hnrnph2 ASO P209L females n = 19. (M-N) Effect of juvenile ASO treatment on grip strength (M) and stride length (N) of female mice. Control ASO WT females n = 16, control ASO P209L females n = 11, Hnrnph2 ASO WT females n = 23, Hnrnph2 ASO P209L females n = 19. All graphs show data as mean ± SEM. For all statistical analyses, ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05 by 2-way ANOVA with Tukey's multiple comparison (B-E), 1-way ANOVA with Tukey's multiple comparison (F-G), non-parametric Mann-Whitney U test (I), non-parametric Scheirer–Ray–Hare test with Mood’s median test (J), or 2-way ANOVA with Sidak's multiple comparison (K-N).

To assess the effect of juvenile administration of Hnrnph2 ASO on audiogenic seizure susceptibility and motor function, we treated P209L hemizygous males, P209L heterozygous females, and their WT littermates with 200 μg control ASO or Hnrnph2 ASO by single bolus ICV injection at PND 27-29. Mice were followed until 8 weeks of age, at which point they were weighed and tested for either audiogenic seizure susceptibility or motor function (Fig. 4H). In this larger cohort, several mice suffered seizures immediately after ICV injection, some of whom died as a result, while others recovered. For Hnrnph2 P209L males and WT littermates, there was no significant effect on the incidence of post-ICV injection seizures for genotype or treatment after pairwise comparisons between groups (fig. S8A). Similar results were obtained for Hnrnph2 P209L females and WT littermates; however, after correction for multiple comparisons, control ASO-treated female mice were significantly more likely to experience post-ICV seizures, with P209L female mutants being slightly more susceptible (fig. S8B). Thus, juvenile ASO treatment may induce post-ICV injection seizures, and in female mice, the incidence was significantly increased with control ASO treatment as compared to Hnrnph2 ASO treatment. To minimize potential confounding effects of acute ASO toxicity on the phenotypes being evaluated, we excluded all mice that experienced post-ICV seizures from subsequent behavioral analyses.

Like neonatal treatment, juvenile treatment resulted in significantly reduced whole brain Hnrnph2 mRNA levels and significantly increased Hnrnph1 mRNA levels in Hnrnph2 ASO-treated samples in both WT and P209L mutant male mice (fig. S8C), confirming that the Hnrnph2 ASO effectively targets both WT and mutant alleles. Importantly, all seven genes previously identified as differentially expressed in Hnrnph2 P209L mutant males were restored to WT levels following juvenile Hnrnph2 ASO treatment at 8 weeks of age, while no changes were observed in WT mice (fig. S8D). RNA-seq analysis of whole brains from 8-week-old Hnrnph2 P209L male mice treated with either control or Hnrnph2 ASO further confirmed these findings. In WT mice, Hnrnph2 was the only differentially expressed gene following Hnrnph2 ASO treatment, with no significant increase in Hnrnph1 transcript levels by RNA-seq (fig. S8E, left), although ddRT-PCR from the same samples detected its upregulation (fig. S8C). These results support the specificity and safety of the Hnrnph2 ASO in both neonatal and juvenile stages.

In the control ASO-treated groups, adult P209L male mice exhibited 42 differentially expressed genes compared with WT mice, predominantly upregulated (fig. S8E, middle, and fig. S8F), mirroring the DGE patterns observed in younger mice. Following Hnrnph2 ASO treatment, no significant differences in gene expression remained between P209L and WT samples (fig. S8E, right), with both upregulated and downregulated genes returning to WT levels (fig. S8G).

Analysis of DAS events revealed a substantial number of splicing changes in WT mice treated with Hnrnph2 ASO, consistent with neonatal treatment. The majority were SE events (64%), with 69 upregulated and 54 downregulated (fig. S8H). Control ASO-treated P209L mutant mice also exhibited numerous DAS events compared to WT mice, again primarily SE events (64%) (fig. S8H). Minimal overlap was observed between differentially expressed genes and differentially spliced genes, suggesting that changes in alternative splicing do not directly underlie the observed gene expression differences (fig. S8I).

To evaluate the rescue of splicing abnormalities, we focused on SE events. Control ASO-treated P209L mutant mice had 91 significantly altered SE events compared with WT mice, including 60 upregulated and 31 downregulated events (fig. S8J). Juvenile Hnrnph2 ASO treatment restored 54 SE events (~59%) to WT levels, while 17 events (~19%) remained altered (fig. S8J); 20 events (~22%) were either not detected by RNA-seq or filtered out from the final dataset during data analysis due to low read counts or extreme PSI values (21). Together, these data demonstrate that juvenile Hnrnph2 ASO treatment, like neonatal treatment, is effective in correcting both gene expression and alternative splicing abnormalities caused by the P209L mutation, although some splicing defects persist.

Next, we assessed whether 8-week-old mice exhibit audiogenic seizure susceptibility similar to that observed in 3-week-old mice. Both Hnrnph2 P209L hemizygous males (fig. S8K) and heterozygous females (Fig. 4I) showed a significant increase in audiogenic seizure susceptibility compared to their WT littermates. Similar to the 3-week-old male mice that received neonatal treatment, juvenile Hnrnph2 ASO treatment rescued this increased susceptibility to audiogenic seizures in 8-week-old Hnrnph2 P209L hemizygous males (fig. S8L) and improved it in hemizygous females (Fig. 4J).

We next assessed whether juvenile Hnrnph2 ASO treatment can improve motor function. Due to the reduced survival of Hnrnph2 P209L hemizygous male mice at age 8 weeks and the large number of mice required for behavioral assays, we elected to perform this experiment in Hnrnph2 P209L heterozygous female mice. We found that the latency to fall from the rotarod in adult P209L females was improved by juvenile Hnrnph2 ASO treatment (Fig. 4K). In contrast to neonatal control ASO-treated P209L females, juvenile control ASO-treated P209L females did not have a significant deficit in balance beam performance compared to their WT littermates. Hnrnph2 ASO treatment slightly but significantly increased total latency to cross, but not number of hind paw slips (Fig. 4L), in P209L females compared to their WT littermates. Finally, in contrast to neonatal control ASO-treated P209L females, juvenile control ASO-treated P209L females did not show a significant reduction in grip strength or stride length, and this was not changed by Hnrnph2 ASO treatment (Fig. 4, M and N).

HNRNPH2 protein regulates the expression of HNRNPH1 by modulating its splicing.

In our previous study, we showed that Hnrnph1 total mRNA levels were significantly increased in the cortex of male Hnrnph2 KO mice, but not in male Hnrnph2 R206W or R209L mice. We also identified changes in alternative splicing in the brains of these mice (12). Upon further investigation of the RNA-seq data from that study, we found that alternative splicing of Hnrnph1 was significantly different in Hnrnph2 KO brains, but not in Hnrnph2 P209L or R206W brains. Specifically, compared to their WT littermates, Hnrnph2 KO mice, but not Hnrnph2 R206W or R209L mice, had decreased skipping of an Hnrnph1 exon (Fig. 5A and fig. S9A). Several studies have identified this exon (exon 5 in murine transcript Hnrnph1-202 and human transcript HNRNPH1-203) and surrounding introns to be involved in regulation of Hnrnph1 expression. HNRNPH1 exon 5 is an essential exon, which, if skipped, generates a frameshift and premature termination codon in a downstream exon (Fig. 5B), resulting in nonsense-mediated decay (NMD) of the resulting transcript (28, 29). Using a primer set that distinguishes between Hnrnph1 transcripts including and excluding exon 5 (fig. S9B), we found that primary cortical neurons from Hnrnph2 KO mice express higher levels of the exon 5-including transcript (314 bp) together with lower levels of the exon 5-excluding transcript (175 bp) compared to those from WT mice (fig. S9, C and D). Additionally, the non-productive, exon 5-skipped transcript (175 bp) increased in both WT and Hnrnph2 KO cells treated with the NMD inhibitor cycloheximide (CHX), confirming that this transcript is primarily degraded via NMD (fig. S9, C and D).

Fig. 5. HNRNPH2 protein regulates the expression of HNRNPH1 by modulating its splicing in human cells and iPSC-derived neurons.

(A) Differential alternative splicing of Hnrnph1 in cortex of Hnrnph2 KO mice. (B) Representation of exon/intron organization in Hnrnph1 and HNRNPH1. (C) Schematic outputs from the minigene reporter where differential splicing results in inclusion or exclusion of HNRNPH1 exon 5. The out-of-frame transcript (right) results in the introduction of a premature termination codon in exon 6 and translation of a truncated peptide. (D) Left, immunoblot showing HA expression (representing productive splicing) in the presence or absence of GFP-HNRNPH2 overexpression in HEK239T cells. Arrow indicates HA-positive minigene expression; asterisk indicates a nonspecific band. Right, quantification of expression via densitometry. (E) Left, immunoblot showing HA expression (representing productive splicing) in presence or absence of RNAi targeting endogenous HNRNPH2. Arrow indicates HA-positive minigene expression; asterisk indicates a nonspecific band. (F) ddRT-PCR showing HNRNPH2 and HNRNPH1 total transcript levels in HEK293T cells after HNRNPH2 siRNA knockdown. (G) ddRT-PCR showing levels of HNRNPH1 mRNA including or lacking exon 5 in HEK293T cells after HNRNPH2 siRNA knockdown. (H) ddRT-PCR showing HNRNPH2 and HNRNPH1 total transcript levels in WT human iPSC-derived neurons after HNRNPH2 ASO treatment. (I) ddRT-PCR showing levels of HNRNPH1 mRNA including or lacking exon 5 in WT human iPSC-derived neurons after HNRNPH2 ASO treatment. (J) Levels of HNRNPH1 mRNA lacking exon 5 in WT human iPSC-derived neurons after treatment with cycloheximide. Data are shown as mean ± SD for all panels. For all statistical analyses, ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05 by student t-test (E and J) or ordinary one-way ANOVA with Dunnett's multiple comparison (D, F-I).

We next examined the effect of Hnrnph2 mutations. In control ASO-treated mouse brains, there was no significant difference in the level of Hnrnph1 non-productive splicing between WT and Hnrnph2 P209L as measured by ddRT-PCR, mirroring the previously published RNA-seq results (fig. S9A). Hnrnph2 ASO treatment, however, decreased this non-productive splicing in both WT and P209L mice (fig. S9E). Together, these results show that Hnrnph2 KO or significant knockdown by Hnrnph2 ASO, but not P209L mutation alone, results in a decreased percentage of non-productive splicing of Hnrnph1.

The HNRNPH1 essential exon is flanked by detained introns, a class of retained introns that have been proposed to serve as a pool of nucleus-detained, polyadenylated mRNA for the cell to quickly and efficiently fine-tune gene expression to respond to environmental stimuli and stress (30, 31). In Mantle cell lymphoma, mutations in the intron surrounding the essential exon of HNRNPH1 disrupt binding of HNRNPH1 protein to intronic motifs and increase the inclusion of this exon, leading to increased HNRNPH1 protein expression and inferior clinical outcomes (32). Given the high homology between HNRNPH1 and HNRNPH2 proteins and the emerging theme of self- and cross-regulation among RNA-binding protein paralogs through modulation of unproductive splicing (33-35), we hypothesized that HNRNPH2 might regulate the expression of HNRNPH1 by promoting the skipping of its essential exon. To test this, we expressed a minigene containing the genomic sequence for HNRNPH1 from exons 3 through 7, including all intronic sequences (Fig. 5C) (32) in HEK293T cells, and examined the effect of overexpressing HNRNPH2 protein on the HA-tagged HNRNPH1 peptide. We found that expression of GFP-HNRNPH2 significantly reduced expression of HA-HNRNPH1, reflecting increased non-productive splicing (i.e., essential exon skipping) (Fig. 5D, arrow). Additionally, GFP-HNRNPH2 overexpression was associated with a marked trend toward reduced endogenous HNRNPH1 protein levels (fig. S9F). Conversely, inhibiting endogenous HNRNPH2 expression by RNAi promoted productive splicing (i.e., essential exon inclusion), as reflected by increased levels of the HA-tagged peptide (Fig. 5E). This effect was also observed at the endogenous HNRNPH1 transcript level: RNAi-mediated KD of HNRNPH2 led to higher total HNRNPH1 transcript levels (Fig. 5F) accompanied by increased inclusion and decreased skipping of exon 5 in a dose-dependent manner (Fig. 5G). Together, these data suggest that HNRNPH2 protein, like HNRNPH1 protein, regulates the expression of HNRNPH1 by modulating its splicing, thus providing molecular insight into how the Hnrnph2 ASO upregulates Hnrnph1 expression.

Research ASO targeting human HNRNPH2 knocks down HNRNPH2 and upregulates HNRNPH1 likely by modulating its alternative splicing.

To test whether the compensatory upregulation of Hnrnph1 in response to Hnrnph2 ASO treatment observed in the mouse brain is conserved in human neuronal cells, we treated human WT iPSC-derived neurons with a research ASO targeting human HNRNPH2 for degradation by RNase H (fig. S9, G and H). Whereas control ASO had no effect (fig. S9I), HNRNPH2 ASO treatment from day 7 to 21 of neuronal maturation resulted in a significant knockdown of HNRNPH2 together with increased total HNRNPH1 mRNA levels in a dose-dependent manner (Fig. 5H). This increase in HNRNPH1 mRNA was also reflected at the protein level, with endogenous HNRNPH1 protein levels increasing following HNRNPH2 ASO treatment (fig. S9J). Digital droplet RT-PCR with splicing-specific assays (32) showed increased inclusion and decreased skipping of exon 5 in HNRNPH1 transcripts after HNRNPH2 ASO treatment (Fig. 5I). Since exon 5-skipped transcripts are subject to NMD, as evidenced by a threefold increase after CHX treatment (Fig. 5J), the true level of nonproductive transcripts generated is conceivably higher. In summary, our data suggest that the compensatory upregulation of HNRNPH1 upon HNRNPH2 ASO treatment is conserved in human cells and that HNRNPH2 modulation of alternative splicing of HNRNPH1 is likely involved in the mechanism of compensation.

Discussion

HNRNPH2-NDD has a significant impact on the quality of life of patients and their families. Patients typically develop symptoms early in life, including infant feeding difficulties and failure to thrive. Major features of the disorder include developmental delay/intellectual disability, motor and language impairments, and growth and musculoskeletal abnormalities. Facial dysmorphia, acquired microcephaly, behavioral and psychiatric disorders, epilepsy, gastrointestinal disturbances, and cortical visual impairment are also common, though not present in all patients (36). Variation in the clinical spectrum is speculated to result from location of the pathogenic mutation in the HNRNPH2 protein (6), skewed X chromosome inactivation in the case of females (5), and sex (4). Finally, although rare, early death due to stroke or seizures has been reported (4), highlighting an urgent need for a treatment targeting the root cause of the disease.

Our previous study characterizing Hnrnph2 KO and mutant knockin mice revealed a mechanism for genetic compensation by Hnrnph1. Whereas homologous knockin of human HNRNPH2 R206W or P209L mutations into the mouse Hnrnph2 gene resulted in a highly penetrant, disease-relevant phenotype, KO of Hnrnph2 was well tolerated. Loss of HNRNPH2 protein in KO mice triggered upregulation of Hnrnph1, the autosomal conserved paralog of Hnrnph2. In contrast, modestly reduced levels of mutant HNRNPH2 protein in the nucleus due to impaired nuclear import failed to trigger Hnrnph1 upregulation. Together, these data suggested either a toxic gain-of-function mechanism for mutant cytoplasmic Hnrnph2 or a complex loss-of-function mechanism with failed Hnrnph1 compensation. We hypothesized that both mechanisms would respond positively to a therapeutic strategy designed to deplete HNRNPH2 protein while also triggering compensatory upregulation of HNRNPH1.

In the current study, we designed and tested gapmer ASOs targeting Hnrnph2 in a non-allele-specific manner in a mouse model of HNRNPH2-NDD. We found that a single ICV injection of our lead ASO produced long-lasting reductions in Hnrnph2 expression in whole brain on both the mRNA and protein level accompanied by compensatory increases in the levels of Hnrnph1 mRNA and total HNRNPH1 protein. The ASO showed widespread distribution across the mouse brain and was well tolerated, with no significant effect on survival, body weight, or signs of neuroinflammation. Both neonatal and juvenile treatments with Hnrnph2 ASO restored the expression of differentially expressed genes to WT levels and to a lesser extent corrected differentially spliced genes. Importantly, RNA-seq analyses consistently identified seven commonly upregulated genes that we had previously characterized, and these genes were restored to WT levels by Hnrnph2 ASO treatment in both Hnrnph2 P209L and R206W mice. These findings suggest that Hnrnph2 ASO treatment effectively rescues key molecular phenotypes associated with these mutations. Strikingly, both neonatal and juvenile treatment paradigms completely rescued the severe audiogenic seizure susceptibility of Hnrnph2 P209L male mice and improved the milder seizure phenotype of Hnrnph2 P209L females. Furthermore, some facets of motor function were also rescued or improved by neonatal or juvenile Hnrnph2 ASO treatment in adult Hnrnph2 P209L male and female mice. We also found that neonatal treatment with Hnrnph2 ASO rescued the anxiety and/or impaired decision-making phenotype of hemizygous R206W males and heterozygous P209L females and significantly improved spatial memory of R206W males. Unfortunately, due to their high mortality rate at age 8 weeks, as well as concerns regarding their physical ability to perform the tasks, we were unable to perform these tests on hemizygous P209L males. Together, these findings suggest that ASO-mediated HNRNPH2 knockdown could offer a promising therapeutic approach for HNRNPH2-NDD patients.

Notably, the body weight, survival, and hydrocephalus phenotypes present in some Hnrnph2 mutant mice as reported in our previous paper (12) were not rescued by Hnrnph2 ASO treatment. Although reduced life expectancy is not an established feature of the disorder, premature death due to early stroke or seizure has been reported in patients (4, 11) and is of major concern to families. In contrast, more than half of patients are reported to have experienced difficulty with feeding and weight gain (4). Although we have not identified the specific cause of reduced body weight and survival in Hnrnph2 mutant mice, it is conceivable that these phenotypes could be driven, at least in part, by peripherally expressed mutant Hnrnph2 that is not rescued by central administration of the ASO. An ASO that is administered peripherally may be able to treat these and other peripheral symptoms and should be considered as an alternative or adjunctive treatment strategy. Similarly, the exact cause of the increased incidence of hydrocephalus in our Hnrnph2 mutant mice is unknown, although we did rule out aqueduct blockage in our original phenotyping study. It is possible that the observed hydrocephalus resulted from changes in early brain development (25, 26) that occurred prior to ASO administration and were not reversible. Therefore, symptoms that originate prenatally or early postnatally such as hydrocephalus may not be responsive to postnatal treatment and this should be recognized as a limitation of the proposed postnatal ASO-mediated treatment of HNRNPH2-NDD.

Although most individuals with HNRNPH2-NDD present with symptoms before the age of 12 months (11), it typically takes several years to obtain a genetic diagnosis. Therefore, unless widespread newborn screening becomes available, any future disease-altering treatment would likely be administered well after the newborn stage. This is a concern for potential disease-altering treatments of NDDs, as the early childhood period is characterized by the greatest neuroplasticity and is therefore the most beneficial time for intervention (37). We are encouraged, however, by the developmental expression profile of HNRNPH2: both the mouse and human gene appear to be expressed at relatively low but consistent levels throughout development, from early embryonic stages to adulthood. In contrast, HNRNPH1 is most highly expressed during early embryonic stages, declining during later fetal and neonatal stages before stabilizing during late infancy into adulthood (12, 38). This pattern suggests that HNRNPH1 may govern early brain developmental processes that are gradually shared with HNRNPH2 at later and/or post-developmental stages and raises the possibility that treatment targeting HNRNPH2 mutations may prove to be beneficial even in later developmental stages.

While the time scales characterizing human and altricial rodent brain development differ substantially, the sequence of key events are remarkably conserved. Rodents are born with a relatively immature CNS and experience greater neurogenesis with accelerated brain development and growth postnatally. In contrast, the brains of human neonates are much more developed, although certain processes such as neocortical myelination and synaptic maturation extend well into adulthood (39). Thus, although it remains unclear exactly how the mouse neurodevelopmental stages translate to the clinical treatment window for HNRNPH2-NDD, our juvenile ASO treatment data suggest that patients may benefit from this therapeutic strategy even after the newborn stage.

In summary, this preclinical study demonstrates the therapeutic potential of non-allele-specific ASO-mediated knockdown of HNRNPH2 and suggests that this approach may be well tolerated in patients. Advancing beyond these preclinical proof-of-concept studies will require detailed toxicological assessments and further optimization of ASO chemistry and delivery to ensure both efficacy and safety. Reflecting this progress, an ASO-based clinical trial in pediatric HNRNPH2-NDD patients was recently launched based on this work. Importantly, this approach may be of benefit to all HNRNPH2-NDD patients, regardless of their specific mutation.

Materials and Methods

Study design

In this study, we tested whether an ASO that depletes murine Hnrnph2 while at the same time upregulating Hnrnph1 can rescue phenotypes of Hnrnph2 mutant P209L and R206W knockin mice, thereby setting the stage for an ASO-based treatment approach for HNRNPH2-NDD. Murine Hnrnph2 ASOs were first screened in vitro in primary mouse cortical neurons for efficacy in knocking down Hnrnph2 and upregulating Hnrnph1, followed by in vivo safety assessment in adult WT mice. Subsequent in vivo experiments on potency, duration of action, and distribution of the lead ASO were performed in WT mice, followed by testing for effects on molecular and behavioral phenotypes of Hnrnph2 mutant mice. Both neonatal (PND 2-4) and juvenile (PND 27-29) treatments were tested with the lead ASO. Finally, we performed in vitro experiments to explore alternative splicing as a potential mechanism whereby HNRNPH2 controls the expression of HNRNPH1. We found that this mechanism is likely involved in upregulation of HNRNPH1 in human WT iPSC-derived neurons upon knockdown of HNRNPH2 by a human HNRNPH2 research ASO. For all in vivo studies, experimental sample sizes were based on previously published phenotyping data and the number of animals per treatment group is included in the figure legend. Data collection was stopped once predetermined Hnrnph2 mutant group sizes were reached. For ddRT-PCR experiments and western blots, 3-4 biological repeats were run. For rescue studies, litters of mice were randomly assigned to treatment groups and dosed with the same compound (control ASO or Hnrnph2 ASO). Researchers were not blinded to the genotype or treatment condition in the in vivo experiments. Litters containing no Hnrnph2 mutant mice (WT only litters) were excluded from the study. For the elevated plus maze, mice that fell from the maze were excluded from the data analysis of that test. In the juvenile treatment rescue study, all mice that experienced post-ICV seizures were excluded from subsequent behavioral analysis to minimize potential confounding effects of acute ASO toxicity on the phenotypes being evaluated. No outliers were removed from the data.

Statistics

For ASO potency and duration of action on Hnrnph2 and Hnrnph1 mRNA levels in WT C57BL/6J mice, we used two-way ANOVA (dose/age harvested, treatment, dose/age harvested x treatment interaction) followed by Tukey’s test to compare between control ASO and Hnrnph2 ASO at each dose/age harvested, between lowest control ASO dose and higher control ASO doses, and between lowest Hnrnph2 ASO dose and higher Hnrnph2 ASO doses. For experiments comparing the effect of control ASO, Hnrnph2 ASO, and vehicle on mRNA or protein expression, we used a one-way ANOVA followed by Tukey’s multiple comparison test. For body weight, we used a used two-way ANOVA (sex, treatment, sex x treatment interaction) followed by Tukey’s test to compare males vs. females for each treatment, and control ASO vs. Hnrnph2 ASO, control ASO vs. vehicle, and Hnrnph2 ASO vs. vehicle in males and females separately.

The log-rank (Mantel-Cox) test was used to determine significant differences between Kaplan-Meier survival curves of all groups. If the test showed a significant difference, pairwise survival curve comparison was performed for control ASO-treated WT vs. control ASO-treated mutant, Hnrnph2 ASO-treated WT vs Hnrnph2 ASO-treated mutant, WT control ASO-treated vs. WT Hnrnph2 ASO-treated, mutant control ASO-treated vs. mutant Hnrnph2 ASO-treated mice and resulting P values were adjusted for multiple comparison by the Holm-Šídák method.

For experiments comparing control ASO vs. Hnrnph2 ASO-treated mutants and their WT littermates where a single measurement was taken, we used two-way ANOVA (genotype, treatment, genotype x treatment interaction) followed by Tukey’s test or Sidak’s test to compare control ASO-treated WT vs. control ASO-treated mutant, Hnrnph2 ASO-treated WT vs. Hnrnph2 ASO-treated mutant, WT control ASO-treated vs. WT Hnrnph2 ASO-treated, mutant control ASO-treated vs. mutant Hnrnph2 ASO-treated mice. Only seizure score was analyzed by non-parametric Scheirer–Ray–Hare test, followed by Mood’s median test was used to perform group wise comparisons. For experiments comparing control ASO vs Hnrnph2 ASO-treated mutants and their WT littermates where multiple measurements were taken over time, we used a linear mixed effects model with random intercept to explore how the measurement changed over time and how it was affected by genotype and treatment. The post hoc comparisons were adjusted by the false discovery rate method within each time period.

For the incidence of hydrocephalus and post ICV seizures, we used a 4-by-2 way and 4-by-3 way contingency table, respectively, followed by Fisher’s exact test. If the P value was significant, we performed pairwise comparisons of control ASO-treated WT vs. control ASO-treated mutant, Hnrnph2 ASO-treated WT vs. Hnrnph2 ASO-treated mutant, WT control ASO-treated vs. WT Hnrnph2 ASO-treated, mutant control ASO-treated vs. mutant Hnrnph2 ASO-treated groups, followed by Fisher’s exact test. P values were adjusted for multiple comparison by the Holm-Šídák method.

For experiments in HEK293T cells and iPSC-derived neurons, we used student t-test, or ordinary one-way ANOVA with Dunnett's multiple comparison.