FGF-23 Deficiency Impairs Hippocampal-Dependent Cognitive Function

Abstract Fibroblast growth factor receptor (FGFR) and α-Klotho transduce FGF-23 signaling in renal tubules to maintain systemic phosphate/vitamin D homeostasis. Mice deficient for either the ligand, FGF-23, or the co-receptor, Klotho, are phenocopies with both showing rapid and premature development of multiple aging-like abnormalities. Such similarity in phenotype, suggests that FGF-23 and Klotho have co-dependent systemic functions. Recent reports revealed inverse central nervous system (CNS) effects of Klotho deficiency or Klotho overexpression on hippocampal synaptic, neurogenic, and cognitive functions. However, it is unknown whether FGF-23 deficiency effects function of the hippocampus. We report that, similar to Klotho-deficient mice, FGF-23-deficient mice develop dose-dependent, hippocampal-dependent cognitive impairment. However, FGF-23-deficient brains had no gross structural or developmental defects, no change in hippocampal synaptic plasticity, and only minor impairment to postnatal hippocampal neurogenesis. Together, these data provide evidence that FGF-23 deficiency impairs hippocampal-dependent cognition but otherwise results in a brain phenotype that is distinct from the KL-deficient mouse.

Although the ventrolateral thalamic nucleus of the nucleus of the central nervous system (CNS) expresses FGF-23 mRNA (Yamashita et al., 2000) and cerebrospinal fluid (CSF) carries FGF-23 protein (Kunert et al., 2017), the functions of FGF-23 in the brain are poorly understood. Hepatic overexpression of FGF-23 causes spatial memory and Schaffer collateral long-term potentiation (LTP) deficits that can be corrected by feeding mice a high phosphate diet (Liu et al., 2011). It is unclear whether overexpression of FGF-23 induces cognitive impairment because of its role in learning and memory, or as an indirect consequence of the peripheral illness caused by high levels of FGF-23. When hippocampal neurons are cultured with FGF-23 supplemented media, less complex neuronal morphology is measured that appears to be FGFR but not Klotho-dependent (Hensel et al., 2016). These in vitro results open the possibility of direct and possibly Klotho-independent effects of FGF-23 on hippocampal neurons.
To compare and contrast the in vivo effects of FGF-23 and Klotho deficiency on the brain, we measured hip-pocampal synaptic plasticity, postnatal neurogenesis, and cognitive function of FGF-23-deficient mice. Impairment of hippocampal-dependent cognitive function was not accompanied by synaptic plasticity change and we measured only minor postnatal hippocampal neurogenic niche change. Distinct from the phenotype of the Klothodeficient and Klotho-overexpressing mice, these results suggest that FGF-23 might function independent of KL within the CNS to impair hippocampal-dependent cognitive functions or else might modulate cognition indirectly, secondary to peripheral illness.

Animals
Procedures were approved by the University of Alabama at Birmingham Institutional Animal Care and Use Committee in accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals. FGF-23-deficient mice (C57BL/6J) lines were obtained from L. Darryl Quarles (University of Tennessee Health Science Center; Liu et al., 2006). Mice were generated by breeding heterozygotes. All mice were housed with free access to food and water at 26.6°C and humidity maintained above 40%. FGF-23-deficient mice die prematurely from the confluence of dysfunction body-wide caused by hyperphosphatemia (Liu et al., 2006;Kovesdy and Quarles, 2013). Throughout, male and female mice were used. The only exception is in electrophysiology experiments where only male mice were measured. To minimize support FGF-23-deficient mice, Bacon Softies or Nutra-gel (BioServ) was supplied. For lifespan studies, all mice were weighed weekly and monitored at least every 2 d until their natural death or a moribund phenotype was observed. Mice would be declared moribund if found to show combination of physical characteristics including kyphosis, muscle wasting, thinning hair, labored breathing, and decreased movement. When animals were identified as terminal, they were deeply anesthetized and then terminally perfused and fixed. For perfusion, throughout, tissue was collected after transcardial perfusion with Tyrode's solution (137 mM NaCl, 2.7 mM KCl, 1 mM MgCl 2 , 1.8 mM CaCl 2 , 0.2 mM Na 2 HPO 4 , 12 mM NaHCO 3 , and 5.5 mM glucose) and 4% paraformaldehyde. For slice culture and biochemical applications where fresh tissue was required, mice were anesthetized using isoflurane, once breathing stopped, mice were rapidly decapitated.

qPCR
RNA was extracted from flash frozen brain or thymus using RNA STAT-60 (Tel-Test Inc.) and cDNA was generated using iScript RT Supermix (Bio-Rad) per manufacturer's protocols. MRNA was measured by primer/probe duplex qPCR with SsoFast Probes Supermix (Bio-Rad) and Prime time qPCR assays to the mouse 18s ribosomal subunit (Rn18s: Assay ID Mm.PT.49.3175696.g; IDT) and mouse FGF-23 (assay ID Mm.PT.58.14071003; IDT) on a StepOne qPCR system (Applied Biosystems). Fold change relative to adult brain was calculated using the ⌬⌬Ct method (Clinton et al., 2013).

Field potential recording
Field EPSPs (fEPSPs) were recorded from stratum radiatum of CA1 using glass micropipettes (2-5 M⍀) filled with ACSF in response to extracellular stimulation of Schaffer Collateral axons by a bipolar tungsten microelectrode (FHC). Stimulation was generated by a Master-8-cp stimulator (A.P.I.) and applied with a BSI-2 biphasic stimulus isolator (BAK Electronics). The initial slope of the fEPSP was used as a measure of synaptic response.

Pentylenetetrazole (PTZ) challenge
PTZ (Sigma) was dissolved in PBS and used at 4 mg/ml. A dose of 60 mg/kg was administered by a single intraperitoneal injection. Immediately thereafter, each mouse was placed in an empty cage and observed for 20 min with simultaneous video recording. Seizure severity was scored by an observer blind to either the mouse line or individual mouse genotype. Scoring is reported as the time following PTZ injection to reach a given stage (latency): (1) immobility/freezing; (2) generalized spasm, tremble or twitch; (3) tail extension to 90°; (4) forelimb clonus; (5) generalized clonic activity; (6) bouncing/running seizure; (7) full tonic extension; (8) death (Roberson et al., 2007(Roberson et al., , 2011. After 20 min elapsed, all mice were anesthetized with isoflurane and killed by cervical dislocation

Histologic stains
Fresh dissected tissue was cryoprotected in sucrose, frozen at -80°C, and 10-m cryostat sections analyzed. Following incubation in 2% Alizarin red solution (pH 4.2), sections were dehydrated with sequential washes through acetone, acetone:xylene, and xylene before mounting in Permount mounting media (Fisher Scientific).
Following perfusion/fixation, as above, serial 30 m, free-floating coronal sections representing 1/6th of each brain were mounted on gelatinized slides. Following cresyl violet acetate incubation, fixed sections were dH 2 O washed and dehydrated with graded ethanol and xylenes. Volume was estimated in 1/6th of each brain used the Cavalieri estimator and Stereo-Investigator software (-1.22 to -3.88 mm from bregma). When dentate was estimated, total volume, dorsal (-1.22 to -2.18 mm from bregma) and ventral (-2.46 to -3.88 mm from bregma), was measured (Laszczyk et al., 2017).

Quantification
All quantification was performed by genotype blind researchers. Stereology with optical fractionator software (Stereo Investigator version 9, MicroBrightField Inc.) and a Zeiss Axio Imager (Zeiss) microscope fitted with a motorized stage and video camera (AxioCam MRc5, Zeiss; West et al., 1991) was used to estimate sub-granular zone (SGZ) proliferating cells (Ki67 quantifications) every 6th section from -1.34 to -2.10 mm from bregma (dissector height:15-m with a 5-m guard zone; fixed counting frame of 80 ϫ 80 m with a sampling grid size of 75 ϫ 75 m resulted in 100 -200 sites/brain). For all other quanti-fication, the average total number of cells in three sections of dorsal hippocampus is reported (1/6th of each brain processed, -1.34 to -2.10 mm from bregma counted; Gilley et al., 2011;Okamoto et al., 2011). Within this range, maturation stage was defined similar to Plümpe et al., counting 100 cells/brain (Plümpe et al., 2006;Laszczyk et al., 2017).

Design
Researchers conducting behavior assays were blind to genotype with mice coded to prevent identification of possible phenotypic differences. FGF-23-deficient mice may be physically distinguished and thus researchers were trained to attend to behavior not mouse body size. Statistical method and p value for all comparisons is detailed by figure, data structure, and statistical test applied. CI indicates confidence interval. When a p value was statistically significant (p Ͻ 0.05), the exact p value is reported. Non-significant data are noted as n.s. ‫ء‬ denotes that when multiple t tests occurred on a single graph, the relevant condition or age was compared between WT and KO mice tested; ‫ءء‬ denotes that the KO was compared to the WT at 0, 60, and 90 min.
Additionally, mice were behavior videos were randomized for before scoring. Mice were habituated to the researcher for three consecutive days. Groups of mice were processed sequentially through open field, spatial novelty, and context-dependent fear conditioning tasks.

Open field
On test day mice were placed in the center of the open field apparatus (43 ϫ 43 ϫ 30-cm Plexiglas box). Photo beam detectors quantified all activity for 5 min (ENV-515 software, Med Associates).

Object location memory
Testing was conducted consistent with previous protocols (Stefanko et al., 2009;Haettig et al., 2011;Vogel-Ciernia and Wood, 2014;Laszczyk et al., 2017). Objects of the same height: star shaped plastic bath toys and Lego towers were mounted on metal washers to prevent tipping. Objects were compared using naïve mice to ensure objects elicited a similar level of exploratory interest. Experimental mice were habituated to a white Plexiglas testing chamber (39 ϫ 19 ϫ 21 cm); containing only black tape marking the north facing wall and fresh bedding material for two consecutive days. On training day, two identical objects were placed on the same side of the box as the black tape and mice were allowed to freely explore for 10 min (Haettig et al., 2011). Twenty-four hours later, mice were returned to the box for 5 min where one object was displaced to the center, back of the box. Mouse behavior was recorded (TopScan, Clever Sys 2.0). Videos of task performance were manually scored. Interaction was judged to occur if mice were observed to be facing and sniffing within 2 cm of a given object. Mice had to explore each object for at least 5 s during the training phase and 15 s during testing to be included. During training, mice would be excluded if they displayed object/ side preference Ͼ20%. Three mice were excluded from the final analysis for failure to reach these metrics. For each session, discrimination index was calculated as time spent with the moved object -time spent with the nonmoved object/total time ϫ 100 (Vogel-Ciernia and Wood, 2014).

Context-dependent fear conditioning
Mice were habituated to the testing room for 2 d. On training day, mice were placed in an operant chamber inside an isolation box (Med Associates) for 5 min. During that time, mice freely explored for the first 2 min. Subsequently, a series of 3, 1 s, 0.5-mA shocks were delivered 1x/min. Mice remained in the chamber for 2 min after the last shock. Twenty-four hours after training, mice were tested by return to the same chamber for 5 min. All training and testing were recorded by automated video tracking system (Med Associates). Percentage of time spent freezing was manually scored by measuring freezing behavior in 5-s intervals. KO mice fail to gain weight after approximately five weeks of age. C, Representative Alizarin red histology of aorta from WT and KO mice. Scale bar represents 100 m. D, FGF-23 mRNA detected in WT and FGF-23-deficient (heterozygous tissue, HET) adult brain and thymus by ⌬⌬ C t qPCR (n ϭ 5; mean Ϯ SEM; ANOVA, ‫‪p‬ءءء‬ Ͻ 0.0001). E, Alizarin red staining of WT and KO brain (cerebellum shown; KO lesions indicated by arrows). Graph shows the average number of calcium lesions (five brains/genotype, average number of lesions counted in three bregma-matched sections; t test, ‫‪p‬ءء‬ Ͻ 0.005).

Statistics
All data are reported as mean Ϯ SEM. Kaplan-Meier curves were analyzed by log-rank test. Statistical significance was determined using GraphPad Prism (GraphPad version 6) software to perform Student's t test or one-way ANOVA results were determined to be significant if p Ͻ 0.05 between groups (Table 1).

FGF-23 deficiency impairs spatial memory
FGF-23 mRNA was originally described in the brain's ventrolateral thalamic nucleus (Yamashita et al., 2000). Using multiple commercially available antibodies, we were unable to detect WT FGF-23 protein by either IHC or Western blotting of brain tissue. However, quantitative polymerase chain reaction (qPCR) detected low levels of FGF-23 mRNA in WT brain (Fig. 1D). Since brain FGF-23 mRNA levels are low, we confirmed dose-dependent FGF-23 deficiency (Yamashita et al., 2000) using a representative FGF-23-expressing peripheral organ, the thymus (Fig. 1D). Although FGF-23 is not highly expressed within the brain, FGF-23 mRNA is evident. Prominent Figure 2. FGF-23 deficiency impairs spatial discrimination. Separate cohorts of WT, HET, and KO mice were measured at three and five weeks of age. A, B, Open field performance % of time spent along the walls or in the middle of the field and velocity (m/s). C, Context-dependent fear conditioning quantified as % of time spent freezing on training day before shock, training day after foot shock, and testing day, 24 h later when returned to the same context. D, Spatial novelty schematic. Identical objects were explored at a set position (positions a and b) and the following day 1 of the objects was moved to a new position (position c). E, F, Spatial novelty scored as the discrimination index [((time with moved -time with non-moved/total time) ϫ 100); three and five weeks] or % of object interaction time spent with each object, each day (five weeks shown; males and females used; n ϭ 9 -11/genotype; mean Ϯ SEM; ANOVA ‫ء‬p Ͻ 0.01, ‫‪p‬ءء‬ Ͻ 0.005, ‫‪p‬ءءءء‬ Ͻ 0.0001).
ectopic Alizarin red positive, calcium deposition is detected although deposits are smaller and stochastically distributed, relative to peripheral calcification (Fig. 1E, cerebellum shown).
Klotho deficiency correlates with the rapid onset of hippocampal-dependent cognitive impairment (Nagai et al., 2003;Li et al., 2017). If this effect results from failure of FGF-23 signal transduction through Klotho, FGF-23deficient mice may be expected show hippocampaldependent cognitive impairment. Thus, we evaluated hippocampal-dependent spatial memory function in the same behavioral tasks where Klotho-deficient cognitive impairment was reported (Laszczyk et al., 2017). We measured sex-balanced, independent cohorts of mice at three and five weeks of age. Five weeks was selected to ensure mice completed testing before FGF-23-deficient mice naturally begin to die. We evaluated FGF-23-deficient, heterozygous (HET), and WT mice to test for FGF-23 dose-dependent effects on cognition. Inclusion of HET mice also allowed us to ensure that a cognitive phenotype was not exclusively the result of a decline in motor function caused by peripheral disease. FGF-23-deficient mice do not show anxiolytic behavior at three or five weeks of age ( Fig. 2A). Although indistinguishable at three weeks of age, by five weeks FGF-23-deficient mouse open-field velocity is significantly decreased between groups (Fig.  2B). Next, we measured hippocampal-dependent contextual fear conditioning responses. Groups of mice were not different at three weeks of age; however, at five weeks of age, 24 h after training, FGF-23-deficient groups spent less time freezing (Fig. 1C). Likewise, when we tested the ability to discriminate closely related contexts using a novel object location memory task (Fig. 1D), three-weekold mice are normal but five-week-old FGF-23-deficient mice to not show the normal preference for the moved object (Fig. 1E,F). Thus similar to what is observed with Klotho deficiency (Laszczyk et al., 2017), FGF-23deficient mice rapidly develop hippocampal-dependent cognitive impairment evidenced by impaired fear and spatial memory task performance.

FGF-23 deficiency does not affect hippocampal morphology or Schaffer collateral synaptic plasticity
FGF-23-deficient mouse death is stochastic. To ensure availability of tissue at a common time point, FGF-23deficient mice and WT controls were killed six weeks after birth. When synaptic plasticity is measured at Schaffer collaterals, Klotho-deficient mice have increased synaptic plasticity with both increased PPF and increased LTP (Li et al., 2017). Using six-week-old FGF-23-deficient slice cultures from male mice, Schaffer collateral synapses show no baseline (input:output), presynaptic (PPF), or postsynaptic (LTP) functional difference (Fig. 3A-C). Consistent with normal slice culture results, FGF-23-deficient mice show no change in time to generalized seizure following PTZ injection (Fig. 3D). Thus, unlike Klothodeficient mice (Li et al., 2017), FGF-23 deficiency has no effect on synaptic plasticity at Schaffer collateral synapses. Figure 3. FGF-23 deficiency does not affect gross morphology or synaptic plasticity. WT and KO mice measured at six weeks of age. A, Input-output curves plotted as the initial slopes of the evoked field EPSP (fEPSP; mV/ms) as a function stimulus intensity (mV). No change is measured between genotypes. B, Paired-pulse ratio at 50-to 300-ms intervals showing no differences. C, Burst induced LTP measured as the fEPSP over time before and after stimulation is not different between genotypes (male mice only for electrophysiology: n ϭ 5-6 mice/genotype, two to four slices per mouse, mean Ϯ SEM; t test). D, Mice injected with 60 mg/kg of PTZ were timed for latency to generalized seizure activity. E, Quantification of total hippocampus (HCX), dentate gyrus, and fimbria volumes (-1.22 to -3.88 mm from bregma). Representative WT and KO Nissl stain. Scale bar is 200 m (males and females used; n ϭ 8 -10 mice/genotype, mean Ϯ SEM; t test).

FGF-23 deficiency affects postnatal neurogenesis
The hippocampal dentate gyrus sub-granular zone (SGZ) is one of the few brain areas with ongoing neurogenesis throughout life (Ming and Song, 2011). Klotho-deficient mice show premature collapse of this neurogenic niche over their shortened lifespan (Laszczyk et al., 2017). Klotho-deficient mice first show decreased SGZ proliferation which proceeds loss of multiple neurogenic cell populations (Laszczyk et al., 2017). To determine whether FGF-23 deficiency impairs postnatal neurogenesis, we measured SGZ progenitor proliferation. FGF-23-deficient brains have the same number of actively proliferating SGZ cells as WT, at both three and six weeks of age (Ki-67; Fig. 4A). In postnatal SGZ, generally, radial-like glial precursors divide to produce another radial-like precursor and a cell that then goes on to commit to a neuronal or astrocytic fate (Ming and Song, 2011). Neuronal precursors mature through a series of stages as differential protein expression drives differentiation (Ming and Song, 2011). The neurogenic niche of Klotho model mice is affected at every stage of neuronal differentiation (Laszczyk et al., 2017). However, the total number and number of proliferating FGF-23-deficient SGZ radial-like glial stem-like cells is indistinguishable from WT at either age examined (brain-lipid binding protein, BLBP; Fig. 4B-D). We counted a small but significant increase in neuronal committed, transient amplifying progenitors (TAPS; Sox2ϩ/BLBP-; Fig. 4E). The number of SGZ-associated astrocytes shows a concomitant significant decrease that could indicate more cells are committing to a neuronal fate (S100␤; Fig. 4F). However, more mature immature neurons do not corroborate this possibility as the number of FGF-23-deficient immature neurons, while unchanged at three weeks (DCX) is decreased at six weeks (Fig. 4G,H). We classified the FGF-23deficient immature neurons by their differentiation stage and found no maturation difference relative to WT (Fig. 4I). Similarly, we found no FGF-23-deficient immature neuron proliferation difference (Fig. 4J). These results are consistent with FGF-23 deficiency causing an overall reduction in neuronal differentiation without affecting proliferation at any stage. To determine whether increased cell death was occurring, we measured active-caspase 3 expression but it was very low and not different between groups by hippocampal Western blotting or SGZ IHC (data not shown). Thus, unlike neurogenic niche collapse under  (Lagace et al., 2007) used; n ϭ 6 -9/genotype, mean Ϯ SEM; t test: ‫ء‬p Ͻ 0.02, ‫‪p‬ءء‬ Ͻ 0.002. conditions of Klotho deficiency, FGF-23-deficient neurogenic niches show only minor changes to neurogenesis.

Discussion
FGF-23 deficiency causes hippocampal-dependent cognitive impairment as measured by impaired fear and object location memory (Fig. 2). These are without an effect on hippocampal synaptic plasticity or network excitability (Fig. 3). Meanwhile, the FGF-23-deficient SGZ shows fewer immature neurons than WT controls (Fig. 4), but the decreased number of immature neurons by six weeks of age is not the result of reduced proliferation, reduced stem cell number, or impaired immature neuron maturation (Fig. 4). Thus, decreased immature neuron number likely represents increased cell death of newly committed neurons (Ryu et al., 2016). The dramatic loss of spatial memory measured by both context-dependent fear conditioning and spatial novelty tasks (Fig. 2) indicates cognitive impairment that is not easily explained by the modest changes we measured in the hippocampus. The FGF-23-deficient brains show ectopic calcification (Fig. 1E). Thus, it is possible that hippocampal-dependent cognitive impairment could be the result of an increasingly toxic brain micro-environment caused by FGF-23 deficiency and a resulting lack of normal ion homeostasis.
Although there are reports of CSF circulating FGF-23 (Kunert et al., 2017), we did not detect brain parenchymal FGF-23 protein which may suggest that FGF-23 indirectly affects brain function. As observed with mouse cognitive effects (Liu et al., 2011;Fig. 2), both too much or too little FGF-23 can occur with human illness (ADHR Consortium, 2000;Faul et al., 2011;Singh et al., 2016). Chronic kidney disease patients show increasing serum FGF-23 concentrations as disease worsens (Fliser et al., 2007). These patients can develop cognitive impairment secondary to their systemic kidney disease (Slickers et al., 2007;Sarnak et al., 2013;Hartmann et al., 2015;Hannan et al., 2018) establishing that peripheral renal illness can indirectly impair brain function. Mouse models show that dietary correction of phosphate levels can be sufficient to correct most peripheral symptoms of either FGF-23 signaling deficiency in Klotho-deficient mice (Morishita et al., 2001) or in mice overexpressing FGF-23 (Liu et al., 2011). If an indirect effect of FGF-23 deficiency drives mouse hippocampal-dependent cognitive impairment, correction of phosphate homeostasis may either correct or prevent development of cognitive effects and do so across species impaired by altered FGF-23 levels.
The phenotypic similarity of Klotho and FGF-23deficient mice (Kuro-o et al., 1997;Shimada et al., 2004b;Sitara et al., 2004) lead to the discovery that FGF-23 signaling is mediated through Klotho/FGFR receptor complexes (Kurosu et al., 2006). While it is clear when FGF-23 function is dependent on Klotho, it remains controversial whether FGF-23 or Klotho also function independently of each other (Fukumoto, 2014). Our studies suggest that Klotho deficiency and FGF-23 deficiency cause different effects on hippocampal synaptic plasticity and postnatal neurogenesis although deficiency of either protein is sufficient to induce cognitive impairment ( Fig. 2; Laszczyk et al., 2017). Unlike Klotho deficiency (Laszczyk et al., 2017), FGF-23-deficient mice show no change in synaptic plasticity and very minor effects on adult neurogenesis. This comparison suggests that FGF-23 and Klotho might function independently of one another in the brain as is observed in the heart (Faul et al., 2011). However, in both model systems, conclusive proof that FGF-23 and Klotho function independent of one another are confounded by peripheral illness that could be, either in whole or in part, impacting brain function. Use of cell type-restricted model systems are required for clear molecular dissection of brain-specific FGF-23 and Klotho protein function.