Microglia deficiency accelerates prion disease but does not enhance prion accumulation in the brain

Abstract Prion diseases are transmissible, neurodegenerative disorders associated with misfolding of the prion protein. Previous studies show that reduction of microglia accelerates central nervous system (CNS) prion disease and increases the accumulation of prions in the brain, suggesting that microglia provide neuroprotection by phagocytosing and destroying prions. In Csf1r ΔFIRE mice, the deletion of an enhancer within Csf1r specifically blocks microglia development, however, their brains develop normally and show none of the deficits reported in other microglia‐deficient models. Csf1r ΔFIRE mice were used as a refined model in which to study the impact of microglia‐deficiency on CNS prion disease. Although Csf1r ΔFIRE mice succumbed to CNS prion disease much earlier than wild‐type mice, the accumulation of prions in their brains was reduced. Instead, astrocytes displayed earlier, non‐polarized reactive activation with enhanced phagocytosis of neuronal contents and unfolded protein responses. Our data suggest that rather than simply phagocytosing and destroying prions, the microglia instead provide host‐protection during CNS prion disease and restrict the harmful activities of reactive astrocytes.


| INTRODUCTION
The parenchymal macrophages of the central nervous system (CNS) are known as microglia (Hortega, 1919) and their proliferation and survival is dependent upon signaling via the colony stimulating factor 1 receptor (CSF1R) (Hume et al., 2020). Microglia have been attributed essential functions in the development and homeostasis of the CNS including synaptogenesis, neurogenesis, and maturation of neuronal circuits (Prinz et al., 2019). However, mice with a Csf1r hypomorphic mutation (Csf1r ΔFIRE ) (Rojo et al., 2019), with conditional Csf1r deletion (using Iba1-cre) (Nakayama et al., 2018) and rats with a Csf1r null mutation (Pridans et al., 2018) each lack microglia entirely but have normal CNS development. These findings indicate that developmental roles of microglia are redundant as studies reveal their functions can be carried out by other cells when microglia are absent (Damisah et al., 2020;Guo et al., 2019;Patkar et al., 2021). There is much greater evidence that microglia contribute to neuropathology (Prinz et al., 2019). Neurodegenerative diseases associated with mutations in microglia-expressed genes such as CSF1R in humans have been referred to as microgliopathies (Hume et al., 2020).
Prion diseases, or transmissible spongiform encephalopathies, are fatal progressive neurodegenerative diseases to which there are no cures. Infectious prions are considered to result from the misfolding of the host's cellular prion protein (PrP C ) into an abnormal disease-associated isoform (PrP Sc ) (Prusiner, 1982). The accumulation of PrP Sc within the brain is accompanied by the impairment of neuronal dendritic spines and synapse structures, glial cell activation, vacuolar (spongiform) degeneration and ultimately neurodegeneration. Inhibiting the proliferation and pro-inflammatory responses of microglia via CSF1R inhibition decelerated CNS prion disease (G omez-Nicola et al., 2013). Conversely, the partial depletion or deficiency in microglia was reported to enhance the accumulation of prions in the brain and accelerate the onset of clinical disease (Carroll et al., 2018;Zhu et al., 2016). However, none of these studies resulted in 100% microglial ablation nor addressed the potential confounding effects of ablative cell death or bystander effects, such as impact upon other non-microglial CSF1R-sensitive mononuclear phagocyte populations. For example although the CSF1R-targeting kinase inhibitor PLX5622 has been widely used to ablate the microglia in the brain, such kinase inhibitors also impact peripheral CSF1R-dependent macrophages (Hume & Macdonald, 2012). Since the ablation of peripheral macrophages enhances prion accumulation in the secondary lymphoid tissues (Beringue et al., 2000;Maignien et al., 2005), effects on peripheral macrophage populations in the above studies also cannot be excluded.
To address the above concerns we investigated CNS prion disease in Csf1r ΔFIRE mice which have a complete and specific lack of microglia in the brain but retain brain-associated macrophages (Rojo et al., 2019). We show that microglial-deficiency in Csf1r ΔFIRE mice was associated with accelerated prion disease in the absence of increased PrP Sc accumulation or prion-seeding activity. Instead, earlier astrocyte activation was associated with increased engulfment of neuronal contents and unfolded protein responses without induction of genes associated with neurotoxic (A1) or neuroprotective (A2) reactive astrocyte polarization (Liddelow et al., 2017). These data indicate that microglia provide neuroprotection during CNS prion disease independently of PrP Sc clearance, and restrict the harmful effects of reactive astrocyte activation. Identification of the mechanisms by which the microglia provide neuroprotection during CNS prion disease may reveal novel targets for therapeutic intervention in these and other neurodegenerative disorders.

| Ethics statement
Ethical approvals for the in vivo mouse experiments were obtained from The Roslin Institute's and University of Edinburgh's ethics committees. These experiments were also performed under the authority of a UK Home Office Project License and in accordance with the guidelines and regulations of the UK Home Office "Animals (scientific procedures) Act 1986." Appropriate care was provided to minimize harm and suffering, and anesthesia was administered where necessary. Mice were humanely culled at the end of the experiments by cervical dislocation.

| Prion infection
Mice were infected at 10 weeks old via intracerebral injection with 20 μl of a 1.0% (wt/vol) brain homogenate prepared from mice terminally infected with ME7 scrapie prions. Mice were culled at the intervals indicated after exposure, or observed for signs of clinical prion disease as described elsewhere (Brown & Mabbott, 2014) and culled at a standard clinical end-point. Survival times were calculated as the interval between injection and positive clinical assessment of terminal prion disease. Groups of age-matched Csf1r ΔFIRE mice and Csf1r WT mice were used throughout the study.

| Gait analysis
Gait analysis was performed weekly using the CatWalkXT (Noldus) from 8 weeks of age until positive clinical assessment of prion disease.
Uninfected mice of both genotype were monitored weekly from 8 to 30 weeks of age as controls.

| Immunohistochemistry
Paraffin-embedded sections (thickness 6 μm) were deparaffinized, pretreated by autoclaving in distilled water at 121 C for 15 min, and for PrP-immunostaining immersed in 98% formic acid for 10 min, endogenous peroxidases were quenched by immersion in 4% H 2 O 2 in methanol for 5 min. Sections were incubated overnight with primary antibodies (see Table 1). Primary antibody binding was detected using biotinylated goat anti-species specific antibodies (Jackson Immunoresearch, West Grove, PA) where necessary and visualized using the Elite ABC/HRP kit (Vector Laboratories, Peterborough, UK) and diaminobenzidine (DAB) between stringent washing steps. Sections were lightly counterstained with hematoxylin and imaged on a Nikon Ni.

| Western blot analysis
Brain homogenates (10% wt/vol) were prepared in NP40 lysis buffer (1% NP40, 0.5% sodium deoxycholate, 150 mM NaCl, 50 mM TrisHCl [pH 7.5]). For the detection of PrP Sc a sample of homogenate was incubated at 37 C for 1 h with 20 μg/ml proteinase K (PK) and digestion halted by addition of 1 mM phenylmethylsulfonyl fluoride. Samples were denatured at 98 C for 15 min in 1Â SDS sample buffer (Life Technologies) and separated via electrophoresis through 12% Tris-glycine polyacrylamide gels (Nupage, Life Technologies) and transferred to polyvinylidene difluoride PVDF membranes by semi-dry electroblotting. Primary antibodies (Table 1) were detected by horseradish peroxidase-conjugated goat antispecies specific antibody (Jackson Immunoresearch) and visualized via chemiluminescence (BM Chemiluminescent substrate kit, Roche, Burgess Hill, UK) as described previously (Bradford et al., 2017).

| Image analyses
Image analysis was performed using ImageJ software (http://imagej/ nih.gov/ij) (Schneider et al., 2012). The magnitude of PrP d , GFAP, and CD44 immunostaining on DAB stained sections was compared as previously described . Briefly, the optical density (OD) values for immunostaining were calculated using ImageJ software following H-DAB deconvolution. Mean gray OD values were measured from DAB grayscale images (scaled 0-255) and expressed as a % relative intensity by dividing by the maximum value (255). Immunofluorescent images were analyzed using ImageJ as previously described (McCulloch et al., 2011). Briefly intensity thresholds were applied and then the number of pixels of each color counted and presented as a proportion of the total pixel area under analysis (% area coverage). The preferential co-localization of fluorochromes was determined as previously described (McCulloch et al., 2011) by comparing the observed distribution of colors with those predicted by the null hypothesis that each element of positive staining was randomly and independently distributed. Values significantly greater (P < .05) than the null hypothesis confirm significant co-localization of fluorochromes. The assessment of relative synaptic protein phagocytosis was calculated as the % of PSD95 or gephyrin staining co-localized with GFAP relative to total of each synpaptic protein. Western blot images were subjected to densitometric analyzed by ImageJ. Briefly lanes and bands were identified, threshold levels set and area under the curve measurements taken (pixels). For PrP C and PrP Sc relative expression levels were calculated as a percentage relative to a control normal brain PrP C measurement.   (Table 2). Gene expression relative to naïve Csf1r WT mice was calculated using the ΔΔCT method (Livak & Schmittgen, 2001) using Rpl19 as a reference gene.

| Statistical analyses
Statistical analyses were performed in GraphPad Prism 6.01 (Graph-Pad Software Inc. Csf1r ΔFIRE mice rapidly succumb to prion disease. (a) Survival curve following intracerebral injection of ME7 prions into Csf1r WT or Csf1r ΔFIRE mice (N = 5-6 mice/group). Log-rank Mantel Cox test, P = .0018. (b) Catwalk XT automated gait analysis weekly assessment of hind base of stance in age-matched uninfected Csf1r WT or Csf1r ΔFIRE mice. Points represent group mean and error bars 95% confidence interval. (c) Weekly assessment of hind base of stance in prion-infected Csf1r WT or Csf1r ΔFIRE mice. (d) Weekly assessment of right hind (RH) paw print area in age-matched uninfected mice. Two-way ANOVA. (e) Weekly assessment of right hind (RH) paw print area in prion-infected mice. (f) Weekly assessment of right front (RF) paw intensity in age-matched uninfected mice.

| Longitudinal gait analysis during prion infection
CNS prion disease in mice is associated with profound motorcoordination disturbances (Heitzman & Corp, 1968). We therefore used longitudinal gait analysis to determine whether microglia-deficiency affected the onset of motor disturbances during CNS prion disease Notably, the Csf1r ΔFIRE mice are not monocyte-deficient but their monocytes lack CSF1R expression (Rojo et al., 2019). The IHC and expression profiling indicates that the Csf1r ΔFIRE mutation also prevents monocyte recruitment into the injured brain. Why monocytes aren't recruited into the brains of Csf1r ΔFIRE mice is uncertain. Studies by G omez-Nicola and colleagues have similarly shown that CNS prion disease was not associated with significant monocytic recruitment in wild-type mice, and the absence circulating monocytes in  (Figure 3f). Whole brain PrP C protein expression (Figure 3g,h) was also similar between naïve Csf1r ΔFIRE mice and Csf1r WT mice. Partial-deficiency or temporary ablation of microglia during CNS prion infection was reported to accelerate the accumulation of prion-disease-specific PrP Sc in the brain (Carroll et al., 2018;Zhu et al., 2016). By contrast, PrP Sc accumulation was reduced in the brains of terminally prion-infected Csf1r ΔFIRE compared to Csf1r WT mice (Figure 3i,j).
F I G U R E 3 Legend on next page.
3.5 | Altered neuropathology in the absence of microglia during CNS prion disease Consistent with data presented in Figure 3i,j, immunostaining for prion disease-associated PrP (PrP d ) in the brains of Csf1r ΔFIRE mice at the terminal stage was approximately 50% of the intensity detected in Csf1r WT mice (Figure 4a,b). Since the accumulation of PrP Sc within the brain increases as the infection proceeds (Tatzelt et al., 1999), this finding is most likely a consequence of their significantly shortened survival times, and implies that microglia deficiency produces hypersensitivity to the accumulation of PrP Sc .
CNS prion disease is accompanied by extensive reactive astrocytosis brains of prion-infected Csf1r ΔFIRE mice but the level of GFAP + and CD44 + immunostaining was lower than in infected Csf1r WT mice. As astrocyte activation also increases temporally during CNS prion infection Hwang et al., 2009), this again is most likely a consequence of the Csf1r ΔFIRE mice succumbing to terminal prion disease significantly earlier than infected Csf1r WT mice. In summary, these data reveal that although CNS prion disease duration is shorter in microgliadeficient Csf1r ΔFIRE mice, this is not accompanied by increased neuronal vacuolation, prion accumulation, or upregulation of GFAP or CD44 at the terminal stage, when compared to infected Csf1r WT mice.
3.6 | Absence of induction of neurotoxic "A1" or neuroprotective "A2" reactive astrocyte-associated genes in the brains of prion-infected microgliadeficient mice Reactive astrocytes may be classified into distinct functional subclasses; an A1 subclass with neurotoxic activity and A2 astrocytes considered neurotrophic (Liddelow et al., 2017). Microglia-derived factors have been implicated in the induction of pan-and A1-reactive astrocyte-associated genes (Liddelow et al., 2017). Consistent with the immunohistochemistry data presented in Figure 4, high levels of mRNA encoding the panreactive astrocyte-associated genes Gfap, Cd44, and Cd44v6 were detected in the brains of prion-infected Csf1r WT mice (Figure 5a-c, respectively). The LPS-mediated induction of expression of pan-reactive astrocyte-associated genes including Gfap and Cd44 was reported to be blocked in microglia-deficient Csf1r À/À mice (Liddelow et al., 2017).
However, because of the limited viability of Csf1r À/À mice, these studies were performed at postnatal day 8, and these mice are also deficient in peripheral macrophage populations. In the Csf1r ΔFIRE mice, the expres-

| Csf1r ΔFIRE mice display accelerated onset of vacuolation but unaltered kinetics of prion accumulation
To determine how disease progression was affected by the absence of microglia, brains were collected from groups of Csf1r WT and   (Figure 6c, middle and lower panels, respectively).
The levels of PrP Sc in the brains of Csf1r ΔFIRE or Csf1r WT mice at 98 dpi were indistinguishable (Figure 6d,e). In parallel, the highly sensitive real-time quaking-induced conversion (RT-QuIC) assay was used to quantify the relative prion seeding activity present within the brains of each group (Atarashi et al., 2011). Consistent with data presented in Figure 6c-e, the relative levels of prion seeding activity were also similar in the brains of infected Csf1r ΔFIRE mice and Csf1r WT mice (Figure 6f).
Since certain peripheral macrophages will also have been ablated in the previous studies (Carroll et al., 2018;Lei et al., 2020;Zhu et al., 2016) it is plausible that this may have increased the burden of prions in the spleen and other secondary lymphoid organs, and by F I G U R E 6 Microglial deficiency accelerates prion vacuolation but not brain or peripheral prion accumulation. (a) Lesion profile analysis of prion-infected brains at 98 dpi (N = 4 mice/group). Points represent the mean vacuolation score, error bars = SEM. *P < .05; **P < .01; ****P < . doing so, enhanced their rate of spread to the brain. However, such an effect was unlikely to responsible for the accelerated prion disease in Csf1r ΔFIRE mice, as a similar abundance of prion-specific PrP d was detected on FDC in the spleens of Csf1r ΔFIRE mice and Csf1r WT mice ( Figure 6g). This is consistent with the demonstration that spleen macrophage populations are not affected in Csf1r ΔFIRE mice (Rojo et al., 2019).

| Accelerated onset of reactive astrocyte activation in the absence of microglia
The increased prion-specific vacuolation in multiple brain regions by 98 dpi (Figure 6a), for example within the superior colliculus and hypothalamus (vacuolation scoring areas G3 and G4, respectively) in the Csf1r ΔFIRE mice, was not accompanied by evidence of loss of neuronal nuclear antigen NeuN + neurons within these regions in either Csf1r ΔFIRE or Csf1r WT mice at this time (Figure 7a,b). Instead, the increased vacuolation (Figure 7c) in the intermediate gray layer (motor associated area) of the superior colliculus of prion-infected Csf1r ΔFIRE mice compared to Csf1r WT mice at 98 dpi was accompanied by increased expression of the pan-astrocytic activation marker CD44 ( Figure 7d,e)  and increased frequency of GFAP + morphologically reactive astrocytes (Figure 8a,b).
Astrocytes in the steady state prune synapses to help maintain neural circuitry (Chung et al., 2013). However, abnormal astrocyte synaptic engulfment has been implicated in the pathogenesis of some neurodegenerative disorders (reviewed in Lee & Chung, 2019), and synaptic alterations are considered to contribute to the early behavioral changes observed during CNS prion disease (Cunningham et al., 2003). We therefore assessed the localization of the postsynaptic proteins gephyrin and post-synaptic density protein 95 (PSD95) in relation to GFAP + astrocytes (Figure 8d). The colocalization of both post-synaptic marker proteins in punctate inclusions within GFAP + morphologically reactive astrocytes was increased in the superior colliculus of Csf1r ΔFIRE compared to Csf1r WT mice at 98 dpi (Figure 8d). Morphometric analyses suggested over half of the total amount of these synaptic proteins detected in Csf1r ΔFIRE mice were within astrocytes (Figure 8e,f). Furthermore, additional analyses suggested that the preferential co-localization of PSD95 and gephyrin within the GFAP+ reactive astrocytes of prion-infected Csf1r ΔFIRE F I G U R E 7 Accelerated astrocyte activation in the absence of microglia. (a) Superior colliculus (G3)  Gephyrin/GFAP, P < 7 Â 10 À5 ). Together, these data reveal a statistically significant increase in synaptic engulfment by reactive astrocytes in the brains of prion-infected Csf1r ΔFIRE mice compared to Csf1r WT mice at 98 dpi within this region. (c) Quantitation of relative expression levels of PERK uninfected Csf1r WT and Cs1fr ΔFIRE mouse brain. Not significantly different, Student's t-test.
(d) Western blot analysis of 98 dpi prion-infected Csf1r WT and Cs1fr ΔFIRE mouse brain for unfolded protein response components as indicated.
(g) Immunohistochemical analysis of phosphorylated PERK (PERK-P; red) and GFAP (green) in 98 dpi prion infected, terminal prion infected and age-matched uninfected Csf1r WT and Cs1fr ΔFIRE superior colliculus (G3). Scale bars = 100 μm or 20 μm as indicated. (h) Western blot analysis of terminal prion-infected brain homogenates probed for unfolded protein response components as indicated, β-Actin displayed as a loading control.
(i) Quantitation of the percentage of total phosphorylated eIF2α in terminal prion-infected Csf1r WT and Cs1fr ΔFIRE mouse brains. Not significantly different, Student's t-test. Points show individual mice. Panels A-G, N = 4 mice/group. Horizontal bar = median. Panels H&I, N = 5-6 mice/group 3.9 | Accelerated onset of unfolded protein response in the absence of microglia Accumulation of misfolded PrP Sc in the brain triggers the unfolded protein response in reactive astrocytes (Smith et al., 2020). Specifically, phosphorylation of protein kinase-like endoplasmic reticulum kinase (PERK) causes the transient shutdown of protein synthesis via phosphorylation of eukaryotic translation initiation factor 2A (eIF2α).

| DISCUSSION
In this study, we investigated prion neuropathogenesis in microgliadeficient Csf1r ΔFIRE mice. Spongiform vacuolation and neuronal loss at the terminal stage were indistinguishable in Csf1r WT and Csf1r ΔFIRE mice and the onset of pathology was not correlated with the accumulation of misfolded prions, which are in any case not directly neurotoxic (Benilova et al., 2020). Microglia deficiency did not lead to the increased accumulation of prions in the brain, suggesting that microglial degradation of prions (if it occurs) can be compensated by other cells such as reactive astrocytes. We conclude that the non-redundant function of microglia is to moderate the harmful effects of dysregulated reactive astrocytes and/or to provide supportive factors to neurons (Sariol et al., 2020). Consistent with that interpretation, microglia can suppress astrocyte phagocytic activity and astrocytes are capable of complete, though slower, clearance of neurons in the absence of microglia (Damisah et al., 2020). Previous studies have used a CSF1R kinase inhibitor to infer the role of microglia in CNS prion disease and reported that overall expression of A1-and A2-reactive astrocyteassociated transcripts in the brain was enhanced upon microglial depletion (Carroll et al., 2018;Carroll et al., 2020). However, use of CSF1R inhibitors can lead to partial depletion of microglia, impact other kinases (e.g. KIT, FLT3), cause localized microglial cell death and impact monocytes and macrophages outside the brain. So, the impacts on pathology should be interpreted with caution (Hume et al., 2020).
During the early stage of prion infection, the reactive astrocytes were more abundant in the brains of Csf1r ΔFIRE mice. Although there was no induction of A1 neurotoxic astrocyte-associated genes, the reactive astrocytes displayed signs of enhanced engulfment of neuronal synapses. The observation of activated astrocytes engulfing synapses in the superior colliculus (G3) region of the brains of Csf1r ΔFIRE mice at 98 dpi with prions was coincident with the commencement of overt clinical signs in these mice at this time. These observations strengthen the hypothesis that loss of neuronal connectivity underlies neurological symptoms and precedes complete loss of neurons (Brown et al., 2001;Cunningham et al., 2003;Jeffrey et al., 2000). The engulfment of damaged synapses and neurons by reactive astrocytes could provide a clearance mechanism to protect surrounding undamaged neurons and synapses, as neuronal damage is required for astrocyte-mediated toxicity (Guttenplan et al., 2020).
Independent studies have shown that the reactive astrocytes in the prion infected brain express complement component C3 and LCN2 highly (Hartmann et al., 2019;Kushwaha et al., 2021;Smith et al., 2020). In the current study the onset of the expression of C3  Lim et al., 2021).
The expression of LCN2 may also play a role in the phagocytosis of neuronal material by the reactive astrocytes (Wan et al., 2022).
The phosphorylated activation of PERK and eIF2α in the unfolded protein response pathway is also upregulated in reactive astrocytes during CNS prion disease (Smith et al., 2020), and the onset of this activation was similarly accelerated in the brains of microglia-deficient prion-infected Csf1r ΔFIRE mice. Targeted blockade of this pathway specifically in astrocytes has proved beneficial during prion disease (Smith et al., 2020). Our data from microglia-deficient Csf1r ΔFIRE mice indicate that the microglia employ mechanisms to protect the neurons in the brain against prion infection by restricting both phagocytosis and unfolded protein response in astrocytes. A similar role for microglia has recently been described in the suppression of ATP-mediated excitoxicity in neurons (Badimon et al., 2020).
In conclusion, our data indicate that the microglia provide neuroprotection independently of PrP Sc clearance during prion disease and restrict the harmful activities of reactive astrocytes. Since astrocytes can contribute to both prion replication (Krejciova et al., 2017;Raeber et al., 1997) and synaptic loss in infected brains, preventing these activities would have therapeutic potential (Smith et al., 2020). Of course, since microglia have been attributed essential functions in CNS development and homeostasis (reviewed in Prinz et al., 2019) we cannot entirely exclude the possibility that the absence of microglia in Csf1r ΔFIRE mice may have rendered their neurons more vulnerable to prion-mediated damage.
However, CNS development appears normal in Csf1r ΔFIRE mice despite the complete absence of microglia (Rojo et al., 2019). We also cannot exclude the possibility that the microglia play a role in modulating prion particle toxicity. Abnormal prion accumulations within the brain may comprise a mixture of fibrillar and smaller oligomers of PrP Sc . However, since the smaller, non-fibrillar, PrP Sc particles are more pathological than larger fibrillary aggregates (Silviera et al., 2005), the engulfment and partial digestion of fibrillary PrP Sc aggregates by the microglia may instead enhance their toxicity in the brain.
Further studies are now required to identify the molecular mechanisms by which the microglia provide neuroprotection during CNS prion disease. The previous characterization of the Csf1r ΔFIRE mice included mRNA expression profiling of the hippocampus which identified 85 transcripts that were significantly depleted when compare to wild-type mice, and were presumably not compensated by astrocytes or other cells (Rojo et al., 2019). That list does not include most endosomal and lysosome-associated genes that are more highly expressed by microglia and by inference must be upregulated by other cells in Csf1r ΔFIRE mice. An overlapping gene list was generated by expression profiling multiple brain regions in the Csf1rko rat (Pridans et al., 2018).
Amongst the most down-regulated transcripts are the three subunits of C1q, which have been implicated in regulating astrocyte function (Clarke et al., 2018;Liddelow et al., 2017) and neurodegeneration (Cho, 2019) and have complex roles in neuronal development and homeostasis (Vukojicic et al., 2019). These Csf1r-dependent genes provide a short list of non-redundant pathways that may be used by microglia to provide this neuroprotection and restrict the reactive astrocyte activation in prion disease. Paradoxically, given the focus of the literature on harmful functions of microglia, enhancing their functions may provide novel intervention treatments against these devastating neurodegenerative disorders.

CONFLICT OF INTEREST
The authors declare no conflicts of interest.

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