TDP‐43 pathology and functional deficits in wild‐type and ALS/FTD mutant cyclin F mouse models

Abstract Aims Amyotrophic lateral sclerosis (ALS) is characterised by a progressive loss of upper and lower motor neurons leading to muscle weakness and eventually death. Frontotemporal dementia (FTD) presents clinically with significant behavioural decline. Approximately 10% of cases have a known family history, and disease‐linked mutations in multiple genes have been identified in FTD and ALS. More recently, ALS and FTD‐linked variants have been identified in the CCNF gene, which accounts for an estimated 0.6% to over 3% of familial ALS cases. Methods In this study, we developed the first mouse models expressing either wild‐type (WT) human CCNF or its mutant pathogenic variant S621G to recapitulate key clinical and neuropathological features of ALS and FTD linked to CCNF disease variants. We expressed human CCNF WT or CCNFS621G throughout the murine brain by intracranial delivery of adeno‐associated virus (AAV) to achieve widespread delivery via somatic brain transgenesis. Results These mice developed behavioural abnormalities, similar to the clinical symptoms of FTD patients, as early as 3 months of age, including hyperactivity and disinhibition, which progressively deteriorated to include memory deficits by 8 months of age. Brains of mutant CCNF_S621G mice displayed an accumulation of ubiquitinated proteins with elevated levels of phosphorylated TDP‐43 present in both CCNF_WT and mutant CCNF_S621G mice. We also investigated the effects of CCNF expression on interaction targets of CCNF and found elevated levels of insoluble splicing factor proline and glutamine‐rich (SFPQ). Furthermore, cytoplasmic TDP‐43 inclusions were found in both CCNF_WT and mutant CCNF_S621G mice, recapitulating the key hallmark of FTD/ALS pathology. Conclusions In summary, CCNF expression in mice reproduces clinical presentations of ALS, including functional deficits and TDP‐43 neuropathology with altered CCNF‐mediated pathways contributing to the pathology observed.

CCNF encodes the 786 amino acid protein Cyclin F (CCNF). CCNF is a member of the F-box protein family, characterised by the presence of an F-box motif, which acts as the substrate recognition component of the Skp1-Cul1-F-box (SCF) E3 ubiquitin ligase complex responsible for mediating ubiquitination of target proteins. CCNF plays an important role in maintaining homeostasis through the timely degradation of damaged and unwanted proteins [15]. Cellular models utilising CCNF expression had been extensively used to assess key mechanisms involved in understanding the pathobiology of disease-linked CCNF variants. Expression of S621G mutant CCNF (CCNF S621G ) in cells resulted in excessive CCNF-mediated ubiquitination [16]. Proteomic analysis identified the presence of disrupted caspase 3-mediated cell death and viability pathways, suggesting a toxic gain-of-function associated with CCNF S621G expression [17]. Expression of CCNF S621G in Neuro-2A cells resulted in a specific increase in levels of Lys48-ubiquitylated proteins, but not Lys63-ubiquitylated proteins [18]. Proteomic analysis identified the clustering of these Lys48-ubiquitylated proteins to the autophagy pathway [18].
Zebrafish were used to study disease-linked CCNF functions for the first time in vivo [17]. CCNF S621G expression led to increased cleaved caspase-3-mediated cell death as well as abnormal axonal outgrowth with concomitantly reduced motor function in zebrafish [17]. However, no accumulation of ubiquitinated proteins was observed in this model, which was attributed to the brief period of mutant CCNF S621G expression.
In this study, we developed the first mouse models expressing either wild-type (WT) human CCNF or its mutant pathogenic variant S621G, which recapitulate key clinical presentations of FTD and neuropathological features of both ALS and FTD linked to CCNF disease variants. We expressed human CCNF WT or CCNF S621G throughout the murine central nervous system by intracranial delivery of adeno-associated virus (AAV) to achieve widespread brain delivery via somatic brain transgenesis. These mice developed behavioural abnormalities similar to the clinical symptoms of FTD patients. Brains of CCNF mice displayed accumulation of ubiquitinated proteins, elevated levels of phosphorylated TDP-43 and cytoplasmic TDP-43 inclusions, recapitulating the key hallmark of FTD/ALS pathology.
Behavioural and neuropathological changes were more pronounced in CCNF_S621G than CCNF_WT mice.

AAV plasmid and production
N-terminally V5-tagged human WT and mutant CCNF with the point mutation S621G were separately cloned into the multiple cloning site of an AAV vector (pAM-CAG) under the CAG-promoter for neuronal expression [19]. All plasmids were amplified and propagated in VB UltraStable competent cells (VectorBuilder).
Packaging of AAV vectors was performed as previously described using the AAV.PHP.B capsid [20,21]. Briefly, 293 T cells were seeded

Key points
• CCNF WT and CCNF S621G mice display behavioural and functional deficits reminiscent of frontotemporal dementia (FTD) and amyotrophic lateral sclerosis (ALS) including hyperactivity, disinhibition and progressive memory deterioration.
• CCNF WT and CCNF S621G mice recapitulate classical neuropathological features of the disease including the presence of ubiquitinated proteins such as the splicing factor proline and glutamine-rich (SFPQ) protein, valosincontaining protein (VCP) and elevated levels of phosphorylated TDP-43.
• Key proteins implicated in FTD and ALS such as CCNF, VCP, SFPQ and TDP-43 all mechanistically converge on common molecular pathways in relation to FTD/ALS pathogenesis.
in complete DMEM (Sigma) with 10% FBS, and the medium was changed to IMDM (Sigma) with 5% FBS 3 h prior to transfection. Cells

Mice and AAV injections
Time-mated C57Bl/6 mice were obtained from ARC Perth and allowed to give birth. AAV injections were carried out on pups at postnatal day 0 to 1 (P0-P1) whereby 1.2 μl/site (6 Â 10 10 viral particles) of AAV particles was injected at five sites bilaterally into the brains (total 30 Â 10 10 viral particles) of cryo-anaesthetized neonatal mice as described [20].
All mice were housed on a 12-h light/dark cycle with access to standard chow and water ad libitum. Both male and female mice were used throughout this study. All experiments were approved by the Macquarie University animal care and ethics committee.

Histological analysis
Mice were anaesthetized and transcardially perfused at either 3, 8 or 12 months of age with PBS. Brains were removed and the hemispheres separated. One hemisphere of the brain was immersion-fixed in 4% paraformaldehyde for immunohistochemical analysis as previously described [22,23], and the other hemisphere was sub-dissected and snap-frozen in liquid nitrogen for biochemical analysis. Fixed brains were processed using an automated system (Excelsior, Thermo, USA), embedded in paraffin and sagittally sectioned at the level of the mid-hippocampus into 3-μm-thick sections using a microtome (Thermo, USA). All staining was done in Sequenza staining racks for standardisation using previously reported protocols [24]. The

Behavioural, memory and motor testing
C57Bl/6 mice injected with WT CCNF (n = 19), CCNF S621G (n = 14) or EGFP (n = 11) AAV underwent motor (rotarod and hanging wire test) and behavioural testing (elevated plus maze and open field) at 3 and 6 months of age followed by cognitive testing (Morris Water Maze) at 8 months of age.

Rotarod
The motor performance of mice was determined using a Rota-Rod

Open field
Mice were placed at the periphery of a 40 cm Â 40 cm Perspex box in an enclosed cupboard and videoed for 10 min. Videos were analysed using the AnyMaze software, and the box was divided into an outer and inner zone (the inner zone was a 17.5 cm Â 17.5 cm 2 in the centre of the box).

Hanging wire test
Mice were placed on a wire mesh and allowed to hang upside down for a maximum of 3 min; latency to fall off was recorded (longest time out of two attempts).

Morris water maze
The apparatus consisted of a 1.5-m-diameter tank with a 40-cm-high Perspex platform (diameter 10 cm), which was placed roughly 20 cm TDP-43 PATHOLOGY AND FUNCTIONAL DEFICITS IN WILD-TYPE AND ALS/FTD MUTANT CYCLIN F MOUSE MODELS from the edge of the wall. The tank was filled to 0.5-1 cm above the surface of the platform, and a non-toxic, acrylic-based paint was added to the water to obscure the platform. Four signposts with different shapes were placed equidistant around the pool as visual cues. Mice were acclimatised to the room for 1 h prior to testing each day. Days 1-6 consisted of an acquisition phase, in which mice were placed in the quadrant opposite the platform at one of four starting positions and given 60 s to locate the hidden platform. Mice that failed to find the hidden platform were guided to the escape platform, and all mice remained on the platform for an additional 60 s before being removed from the maze. Mice had four trials per day, each starting from a different position, and the order of starting positions was altered each day. On the seventh day, the platform was removed, and the mice were given 30 s to explore the pool (probe trial). On the eighth day, the platform was placed back in the pool with a flag attached, and visual cues were removed from outside of the pool, to ensure that all mice had normal vision. Videos were analysed using the AnyMaze software. Accordingly, trace plots obtained for each swim were classified visually based on previously published paradigm [25]. Briefly, swim patterns were scored as follows: 1, thigmotaxis; 2, random swim; 3, scanning; 4, chaining; 5, directed search; 6, focal search; and 7, direct swim. Based on this scoring scheme, 1-3 reflect non-spatial hippocampal-dependent, while search strategies 4-7 are categorised to reflect spatial hippocampal learning.

RIPA/UREA solubility extraction
Sequential extraction to determine the solubility of TDP-43 has been previously described [26]. Briefly, 50-100 mg of brain tissue was homogenised and sonicated in a modified RIPA buffer (50 mM tris [pH 8], 150 mM NaCl, 0.1% Na-dodecyl sulphate, 0.5% Na-deoxycholate, 1% Nonidet™ P 40 substitute, 5mM EDTA and protease inhibitor cocktail [cOmplete™, Roche]). Lysates were incubated for 30 min, rotating at 4 C before they were subjected to a 1 h, 50,000g centrifugation step. The supernatant was collected as the RIPA-soluble fraction. Two further homogenisation/centrifugation washes were carried out to remove any remnants of RIPA-soluble proteins before the RIPA-insoluble pellet was further extracted and sonicated in Urea buffer (7 M urea, 2 M thiourea, 4% CHAPS and 30 mM tris [pH 8.5]). Protein extracts were analysed by immunoblotting as previously described [27]. The following antibodies were used: mouse anti-

Statistical analysis
All statistical analyses were performed using Graphpad Prism 9 software using ANOVA (Tukey's multiple comparison test) or binomial distribution (for MWM swim path analysis only). P values below 0.05 were considered significant. All values are presented as the mean and standard error of the mean.

RESULTS
Generation of WT CCNF and mutant CCNF (S621G) mouse models by AAV-mediated gene expression  Figure 2C).
Despite the altered exploration patterns in 6-month-old CCNF_WT and CCNF_S621G mice, their total distance travelled during testing was comparable ( Figure S1C).
To determine if memory formation is compromised in CCNF_WT and CCNF_S621G mice, spatial memory was assessed using the Morris water maze (MWM) paradigm. Eight-month-old CCNF_WT and CCNF_S621G mice showed delayed learning during the memory acquisition phase, taking significantly longer to find the hidden platform at day 3 than EGFP controls ( Figure 2D). At the same time, there were no differences in time spent in the platform quadrant or distance travelled during the probe trials between test cohorts ( Figure S1D). A detailed analysis of the learning phase by categorising swim strategies into the hippocampus-or non-hippocampus-dependent patterns showed that CCNF_WT mice used significantly fewer hippocampusdependent swim strategies on days 1-6 and CCNF_S621G mice on days 1, 2, 3 and 5 compared to EGFP controls ( Figure 2E).
Finally, motor performance was assessed in EGFP, CCNF_WT and CCNF_S621G mice by subjecting them to the accelerating rotarod and hanging wire tests. No differences between groups were observed in either rotarod ( Figure S1E) or hanging wire tests ( Figure S1F) at 3 or 6 months of age. Similarly, no differences in weight were observed in males or females between the groups at either 3 or 6 months of age ( Figure S1G).
Taken together, CCNF_WT and CCNF_S621G mice presented with hyperactivity, reduced anxiety and impaired spatial memory acquisition, reminiscent of clinical symptoms of FTD [1] in the absence of any overt motor impairments.
Loss of POU3F2 immunoreactivity in the cortex of aged mutant CCNF_S621G mice  Figure 3A). Astrogliosis is a common feature of neurodegenerative conditions and is frequently observed in mouse models of ALS [26,28,29]. Therefore, astroglial states in EGFP, CCNF_WT and CCNF_S621G mice were next assessed by the presence of GFAPpositive cells in the cortex. Surprisingly, there was a significant reduction of GFAP levels in cortical RIPA-fractions of 12-month-old CCNF_S621G mice compared to EGFP controls. The observed reduction was progressive, as GFAP levels were comparable at 3 months of age across the three lines, with only a trend towards reduction in CCNF_S621G mice ( Figure 3B). Immunofluorescent staining of 12-month-old CCNF_S621G cortex further confirmed the reduction in GFAP immunoreactivity compared to EGFP controls ( Figure 3C).
This reduction in astroglial activation occurred independently of any change to the overall levels of the astroglial marker S100beta, which were comparable in EGFP, CCNF_WT and CCNF_S621G mice at both 3 and 12 months of age ( Figure 3B).
In summary, we found a progressive reduction in astroglial activation in 12-month-old CCNF_S621G mice, concomitant with a reduction in POU3F2-immunoreactivity in cortical layer V of these mice.

Ubiquitination and insolubility of CCNF's interaction targets in CCNF mice
Aberrant protein phosphorylation and ubiquitination are hallmark features of ALS [30]. Using sequential protein extraction with buffers of increasing stringency (=RIPA followed by UREA buffer), we found elevated levels of ubiquitinated proteins only in cortices of CCNF_S621G mice but not CCNF_WT mice compared to EGFP controls ( Figure 4A).
In addition, we probed the effects of WT and mutant CCNF expression on their known interaction partners, TDP-43 and the splicing factor proline and glutamine-rich (SFPQ). Phosphorylation of TDP-43 at serine 409/410 is a pathological feature present in all forms of TDP-43 proteinopathies [31]. Consistent with this, we found an increase in the levels of TDP-43 phosphorylated at serines 409/410 in both CCNF_WT and CCNF_S621G mice ( Figure 4B). Furthermore, we analysed levels of SFPQ in these mice and found increased SFPQ in the UREA fraction of 12-month-old CCNF_WT and CCNF_S621G mice ( Figure 4B).
In essence, elevated levels of TDP-43 phosphorylation at pathological sites and accumulation of insoluble SFPQ were observed in both CCNF_WT and CCNF_S621G mice, while aberrant protein ubiquitination was only observed in CCNF_S621G mice.

Cytoplasmic inclusions in CCNF_WT and CCNF_S621G mice
Cytoplasmic localisation of CCNF has previously been reported when overexpressed in cells [32]. Accordingly, some cytoplasmic localisation of CCNF was observed in the cortices of 3-month-old CCNF_WT mice in addition to its nuclear presence. This cytoplasmic CCNF was more pronounced in CCNF_S621G mice ( Figure S2).
Mutations in the valosin-containing protein (VCP) have previously been reported in ALS [33], and CCNF has been shown to directly interact with VCP, enhancing its activity [34]. Accordingly, we found CCNF colocalised with VCP in the nucleus and cytoplasm of neurons in CCNF_WT mice ( Figure 5A). Interestingly, we observed the presence of cytoplasmic inclusions of CCNF that colocalised with VCP in CCNF_S621G mice ( Figures 5A and S2). Co-immunofluorescence labelling of NeuN, DAPI with either GFP or V5 confirmed the cytoplasmic localization of these CCNF inclusions in CCNF_WT and CCNF_S621G cortices ( Figure 5B). This led us to next investigate whether expression of CCNF in mice leads to the altered subcellular localisation of TDP-43. In line with ALS-like pathology, we found cytoplasmic inclusions of TDP-43 in the cortices of 12-month-old CCNF_WT and CCNF_S621G mice ( Figure 5C). Interestingly, the colabelling of sections showed that the TDP-43 inclusions did not colocalise with CCNF ( Figure 5C).
Taken together, cytoplasmic inclusions, containing both CCNF and VCP, were observed in 3-month-old CCNF_WT and CCNF_S621G, which eventually led to cytoplasmic TDP-43 aggregates at 12 months of age.

DISCUSSION
To the best of our knowledge, our study is the first to generate mouse models of ALS with transgenic expression of CCNF in either its non-

CONFLICT OF INTEREST STATEMENT
The authors declare that they have no competing interests.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.

ETHICS STATEMENT
All experiments were approved by the Macquarie University Animal Care and Ethics Committee.