3.1. Generation of a Parp6 Knockout Mouse Line Using CRISPR-Cas9 Mutagenesis
Parp6 is expressed preferentially in the mouse nervous system (
www.informatics.jax.org and
www.Biogps.org databases, accessed on 2015–2021). In the mouse brain, Parp6 is expressed almost exclusively in neurons (
http://dropviz.org/ and
https://amckenz.shinyapps.io/brain_gene_expression/, accessed on 2015–2021). shRNA- and miRNA-based knockdown experiments in rat hippocampal neurons in vitro and in vivo (using in utero electroporation of shRNA-based plasmids) demonstrate that PARP6 regulates dendritic arborization [
7]. To further study the role of Parp6 in the nervous system in vivo, we generated a Parp6 knockout mouse line using the CRISPR-Cas9 mutagenesis system, using a single-guide RNA (sgRNA) sequence corresponding to exon 18 in the C-terminal catalytic domain of Parp6 (
Figure 1A). The mutation resulted in a frameshift mutation 91 residues into the catalytic domain (aa483), and a subsequent early termination codon at amino acid 507 (mouse numbering), which resulted in a truncated Parp6 protein (Parp6
TR) that was devoid of the catalytic domain (
Figure 1B,
Table S4). RNAseq analysis from mouse hippocampal tissue showed a 3-fold decrease in
Parp6 RNA levels in Parp6
TR compared to Parp6
WT samples (
Figure S1). Importantly, the expression levels of all other known
Parp transcripts were not affected in the
Parp6 mutant line created (
Figure S1).
Once the genotyping protocol was established, the mouse pups were weaned and genotyped. In general, the crossing of heterozygous Parp6TR males and females originated a genetic ratio of 139:239:8 (n = 386; 36.01:61.92:2.07%), with only six mutants surviving longer than 24 h: three were P23 days old when they suddenly died, two were sacrificed at P21 days old, one male lived to be 100 days old and was sterile, and two P0 pups lived less than one hour (see below). This Mendelian ratio differed significantly from the expected 1:2:1 ratio (Chi-square (2, n = 386) p < 0.0001), indicating that mice expressing Parp6TR generally do not survive to adulthood.
To rule out that the death of homozygous Parp6TR was due to an off-target effect of the sgRNAs, we used an online CRISPR/Cas9 target online predictor tool to identify potential off-target sequences for our Parp6 sgRNA sequence, then PCR-amplified and sequenced the top five hits for possible off-target genes: Cacna2da, Pstpip1, Rims2, Atb2b2, and Zfp708. No mutations were found in any of these genes (data not shown). In addition, we backcrossed our heterozygote Parp6HT mice to new C57BL/6N Parp6WT mice and obtained the same genetic ratios after following this line for eight generations.
To further validate our Parp6 knockdown mouse line, we performed RT-PCR reactions to look at the reduction in
Parp6 transcript levels in the Parp6
TR embryos compared to the Parp6
WT embryos. Because the indel was in the catalytic domain, we designed two sets of primers that recognized the 5′ end of the
Parp6 transcript to look at
Parp6 transcript stability and used these specific primer sets on total RNA obtained from whole brains of Parp6
WT, Parp6
HT, and Parp6
TR animals. The overall reduction in the
Parp6 transcript in Parp6
TR relative to Parp6
WT animals was 86.2% (n = 4), while the Parp6
HT showed almost no reduction compared to Parp6
WT (
Figure 2A). These results suggest that the
Parp6TR mRNA is not stable and is likely rapidly degraded. Similar results have been obtained in other CRISPR/Cas9 knockout studies [
11].
We next evaluated Parp6 protein levels in Parp6
WT, Parp6
HT, and Parp6
TR mouse brains by Western blotting. We used a Parp6 antibody generated by the Chang lab, which recognizes a region within the C-terminal catalytic domain of Parp6 [
3]. We used this antibody because we found that all commercially available Parp6 antibodies were unreliable (data not shown). We first validated this antibody in rat cortical neurons transduced with a validated shRNA-based knockdown lentivirus that targeted Parp6 [
7] (
Figure 2B). We observed a complete loss in full-length Parp6 protein levels in Parp6
TR compared to Parp6
WT and Parp6
HT brain lysates (
Figure 2C). Because we do not have a Parp6 antibody that recognizes the N-terminal region, we cannot determine if Parp6
TR is actually expressed in the Parp6
TR mice.
3.2. Characterization of Parp6TR Embryos and Few Surviving Adults
In our previous studies, we found that Parp6 regulates dendrite morphogenesis in embryonic rat hippocampal neurons [
7]; however, we did not examine axon outgrowth. Parp6 is not only expressed in central nervous system (CNS) neurons, but also in peripheral nervous system (PNS) neurons (e.g., dorsal root ganglion and spinal cord cells) (
www.Biogps.org,
https://gp3.mpg.de and
https://mouse.brain-map.org databases, accessed on 2015–2021). To examine peripheral nerve development, we performed whole mount immunolabeling for neurofilament medium chain (NF165) on embryonic day 12.5 (E12.5) Parp6
TR and Parp6
HT embryos (
Figure S2). Immunolabeled and cleared embryos were imaged and overall embryo development was quantified through eye and paw area measurements. Peripheral nerve development was quantified through outgrowth measurements of the forepaw axon bundle and intercostal axon bundle outgrowth. No significant difference in overall development or axonal outgrowth was seen between Parp6
HT and Parp6
TR animals from multiple litters (
Table S5). We also compared Parp6
WT and Parp6
HT animals from a single litter and saw no significant difference (
t-test for same measurements shown in
Table S5:
p = 0.2423,
p = 0.3165,
p = 0.2141,
p = 0.4639, and
p = 0.5953, respectively, n = 3–4). These results demonstrate that catalytic activity of Parp6 is not required for axon outgrowth during development in mice.
Because Parp6
TR newborn pups did not survive more than 24 h, we wanted to identify when Parp6
TR animals were dying. We genotyped E18 embryos and observed a genetic ratio of 83:114:76 (n = 273; 30.4:41.8:27.8%; Chi-square (2, n = 273)
p = 0.1445) indicating that the PARP6
TR embryos survived at least until E18 based on Mendelian genetic ratios. In addition, based on their gross morphology, Parp6
TR E18 embryos were indistinguishable from Parp6
WT and Parp6
HT (
Figure 3A). The weights of E18 embryos obtained from two pregnant females, even though one litter was larger in overall size than the other, showed no significant differences between the three genotypes (
Figure 3B; one-way ANOVA
p = 0.4087 and
p = 0.1027), and when weights were normalized to Parp6
WT, Parp6
HT and Parp6
TR were 104 and 97% of the Parp6
WT, respectively.
The only two Parp6
TR mice that survived at least to 21 days old were significantly smaller than their Parp6
WT and Parp6
HT siblings (WT 11.818 ± 3.368 g, HT 9.561 ± 1.792 g, TR 3.675 ± 0.035 g) (
Figure 3C, one-way ANOVA
p = 0.0066). When weights were normalized to Parp6
WT, Parp6
HT and Parp6
TR were 80 and 31% of the Parp6
WT, respectively. In addition to the significantly smaller size at P21 days, these animals showed a marked difference in their motor activity (
Video S1). Their behavior was erratic and jumpy, and the twitching behavior was lost over time in the male that lived 100 days, who eventually caught up in size and weight to his Parp6
WT and Parp6
HT siblings, making him indistinguishable from his siblings. The only clear phenotype in the long-term surviving Parp6
TR male was that he was sterile. Interesting to note is that Parp6 is also highly expressed in the testis (
www.biogps.org, accessed on 2015–2021), suggesting a possible role for Parp6 in male fertility.
3.3. Loss of Full-Length Parp6 Leads to Perinatal Lethality
At birth, the Parp6
TR pups were indistinguishable from their siblings, and were able to feed milk from their mothers (
Figure 4A). However, in less than 30 min Parp6
TR mice slowly became cyanotic and died (
Figure 4B). We sought to understand why Parp6
TR mice died soon after birth. We hypothesized that the breathing problems in Parp6
TR mice could be due to defects in nerve innervation of the diaphragm and/or rib cage muscles at later stages of embryonic development. To test this hypothesis, we looked at nerve innervation of the rib cage muscles as well as the diaphragm in embryos at embryonic day E18 by immunofluorescence. We found that nerve innervation of the rib cage muscles and the diaphragm was also normal in Parp6
TR mice (
Figure 5A—top and bottom, respectively).
Another possible explanation for the abrupt cessation in breathing is the improper innervation, formation, or connections in the respiratory centers located in the medulla oblongata and pons, which are parts of the brainstem. These centers control the rate and depth of respiratory movements of the diaphragm and other muscles, and the arrhythmic firing of neurons in a few specific brain areas (pre-Bötzinger and Bötzinger) can lead to abnormal breathing and death by asphyxia. We then looked at the expression of
Parp6WT transcripts in different brain regions of 4-week-old mice. Interestingly, the highest expression was seen in the dorsal medulla, which has an important role in processing sensory information from the upper and lower airways for the generation and control of airway protective behaviors (
Figure S3). Based on this finding, we asked if there was enough air in the lungs of P0 Parp6
TR mice that turned cyanotic and died just after birth. We prepared cross-sections of P0 lungs from Parp6
WT and Parp6
TR mice and measured the number and area of clear spaces (air). Overall, there were more (389 vs. 288) and larger clear spaces (2403 vs. 1683 µm
2) in the Parp6
WT lungs than in the Parp6
TR ones (
t-test
p = 0.0029, n = 4,
Figure 5B,C). When the clear spaces potentially filled with air are expressed as a percentage of the total area, the Parp6
TR showed significantly less air in their lungs compared to Parp6
WT (46.08 vs. 24.05%,
t-test
p < 0.0001). This could indicate that the breathing control system was not developed correctly or completely.
3.4. Human PARP6 Mutations: Clinical Findings
While pursuing studies of Parp6 function in mice, we learned about six patients between the ages of 4 to 28 years old that had mutations in their
PARP6 gene. Of these mutations, four were de novo mutations and the two remaining, a pair of siblings, were congenital with the healthy parents being heterozygous carriers. Whole exome sequencing of the patient DNA samples followed by data analysis, including read alignment and variant calling, identified two mutations in the catalytic domain and two in the non-catalytic domains of the PARP6 protein (
Table 1). In general, these patients showed developmental delay, learning disabilities, and epilepsy.
The twenty-eight-year-old male was shown to have a mutation (P111L) in the non-catalytic domain of PARP6. He was the second of three siblings born to healthy non-consanguineous parents, following a normal pregnancy and birth weight. He struggled with a persistent food disorder (gastro-esophageal reflux disease), had a laparoscopic Nissen fundoplication at the age of 20 years old, and currently still has problems with solid food. Seizures evolved from the age of three years on, and he is currently on antiepileptic medications. He had delayed motor (started to walk at the age of 4 years, with flexion of knees and prone to falling, showing global and fine motor skills impairment) and language (no words but cries, very limited understanding) development. A micro-array-based comparative genomic hybridization test (array CGH test) was shown to be normal, a screen of an intellectual disability panel of 275 genes was normal, and the FMR1 gene was also normal. The patient does not have hearing or visual problems, and the brain MRI and heart ultrasound were normal.
The four-year-old girl with the second mutation in the non-catalytic domain (H256Y) showed intrauterine growth restriction (IUGR) followed by normal psychomotor development, and partial agenesis of the corpus callosum. The in silico combined annotation-dependent depletion (CADD) score was 25.
With the exception of the P111L mutation, patients with the mutations in the catalytic domain showed more severe symptoms. All four patients with mutations in the catalytic domain showed mild to severe global developmental delays since early infancy, epilepsy, and two of them (R485 mutations) showed speech delay and learning disabilities. Interestingly, the amino acid position 485 was in close proximity to the catalytic tyrosine at position 487 [
7]. The de novo R485H mutation was discovered by whole exome sequencing in a 10-year-old boy who was suffering from recurrent focal motor seizures and mild to moderate intellectual disability. Chromosomal microarray analysis was initially performed and was normal. His seizures resolved by the age of 13 years and he no longer requires anti-epileptic medications. He continues to have mild to moderate intellectual disability (ID) and performs approximately two grade levels below his age-related peers. In addition to a history of seizures and ID, he has left renal hypoplasia and hypospadias which was repaired at age 2 years. His height, weight, and head circumference have all been tracking within the normal range on standard pediatric growth charts.
Neurological examination for one of the patients (boy with C563R mutation) revealed overall axial and peripheral hypotonia, globally reduced muscle strength (3/5), and hyperactive deep tendon reflexes (+3). A brain MRI performed at three years of age revealed cortical atrophy and delayed myelination with a paucity of white matter with both supra- and subtentorial involvement, which included the cerebellar hemispheres, vermis, and pons. Other systems were not involved as indicated by the normal abdominal ultrasound, echocardiogram, and the results of general blood hematology and biochemistry. Specifically, he had no unusual infections and no skin sensitivity. Hearing and ophthalmic examination were normal in general. As for his sister with the same C563R mutation, MRI showed macrocephalus with frontotemporal enlargement of the outer cerebrospinal fluid space, and symmetrical hyperintense signal alteration in the dentate nucleus on both sides. Taken together, these clinical phenotypes support a critical role for PARP6 in neurodevelopment in humans.
We next examined the impact of two human mutations (R485H and C563R, both in the catalytic domain) on PARP6 activity and function. We generated GFP-Parp6 expression constructs of these mutants, as well as the Parp6
TR mutant (CRISPR Parp6 mutant) (
Figure 6A). To examine the catalytic activity of Parp6, we used a GFP-immunoprecipitation (IP)-auto-MARylation assay developed previously in our lab [
7]. This assay uses 6-alkyne-NAD
+, an NAD
+ analogue that can be coupled to biotin-azide via the “click” reaction. As expected, GFP-Parp6
WT exhibited robust auto-MARylation activity whereas GFP-Parp6
C563R was completely inactive, while GFP-Parp6
R485H showed a slight reduction in auto-MARylation activity (
Figure 6B,C). As expected, GFP-Parp6
TR was completely inactive due to the absence of the catalytic domain.
Next, we then examined the effects of two of the human PARP6 mutants on dendritic complexity in primary rat hippocampal neurons. In previous studies we found that overexpression of Parp6
WT in primary rat hippocampal neurons increased dendritic complexity, whereas overexpression of a catalytically inactive variant decreased dendritic complexity [
7]. Primary neuronal cultures were transfected on DIV7 with GFP-Parp6
WT (control), Parp6
C563R, GFP-Parp6
R485H, and GFP-Parp6
TR, together with an mCherry fill, and were fixed on DIV12. We found that overexpression of Parp6
C563R and GFP-Parp6
TR significantly decreased proximal (50 μm from cell soma) dendritic complexity compared to the GFP-Parp6
WT control (
Figure 6D). Conversely, overexpression of GFP-Parp6
R485H significantly increased distal dendritic complexity (100 μm from cell soma) compared to the Parp6
WT control. These overall results show that the Parp6
C563R mutant phenocopies the catalytically inactive PARP6 mutants in neurons, consistent with the loss of its catalytic activity. The fact that the Parp6
R485H mutant increased dendritic complexity relative to Parp6
WT suggests that it acts as a gain-of-function mutant.
The mechanism by which Parp6 regulates dendritic complexity is currently unknown. As a starting point to investigate how Parp6 regulates dendritic complexity, we determined the Parp6 interactome in neurons using BioID proximity labeling [
8]. We generated a Parp6 variant in which a promiscuous biotin ligase (Myc-BirA*) was fused to the N-terminus of Parp6
WT. When cells are treated with media containing exogenous biotin, BirA* will biotinylate nearby proteins in a diffusion-limited process allowing for identification of proximal proteins using biotin affinity capture. We expressed Myc-BirA*-P2A-GFP (control) and Myc-BirA*−Parp6
WT-P2A-GFP in cortical neurons using lentiviruses, treated with biotin, and observed an increase in biotinylated proteins (
Figure S4). To identify potential Parp6 interactors, we captured biotinylated proteins using Neutravidin agarose, performed on-bead trypsin digestion, and subjected resultant peptides to LC-MS/MS analysis (in duplicate). We obtained a list of 89 proteins that were present in either or both runs, which was then reduced to five possible Parp6 interactors after calculating enrichment over control (
Table S6). Interestingly, the top four interactors were microtubule-associated proteins (microtubule-associated proteins 2, 1B, protein tau, and RP/EB family member 2), which play important roles in regulating the microtubule cytoskeleton in neurons, especially during development. These results are consistent with our functional studies and suggest that Parp6 may play a role in regulating the neuronal microtubule cytoskeleton.