Murine MPDZ-Linked Hydrocephalus is Caused by Hyperpermeability of the Choroid Plexus

Though congenital hydrocephalus is heritable, it has been linked only to eight genes, one of which is MPDZ. Humans and mice that carry a truncated version of MPDZ incur severe hydrocephalus resulting in acute morbidity and lethality. We show by magnetic resonance imaging that contrast-medium penetrates into the brain ventricles of mice carrying a Mpdz loss-of-function mutation, whereas none is detected in the ventricles of normal mice, implying that the permeability of the choroid plexus epithelial cell monolayer is abnormally high. Comparative proteomic analysis of the cerebrospinal fluid of normal and hydrocephalic mice revealed up to a 53-fold increase in protein concentration, suggesting that transcytosis through the choroid plexus epithelial cells of Mpdz KO mice is substantially higher than in normal mice. These conclusions are supported by ultrastructural evidence, and by immunohistochemistry and cytology data. Our results provide a straight-forward and concise explanation for the pathophysiology of Mpdz-linked hydrocephalus.


INTRODUCTION
Despite strong evidence for the heritability of congenital hydrocephalus (Kahle et al., 2016;Munch et al., 2012), to date, only eight genes have been linked to this condition. The earliest monogenic link of hydrocephalus had been made to L1CAM (Coucke et al., 1994;Jouet et al., 1993;Rosenthal et al., 1992;Van Camp et al., 1993), which encodes the L1 neuronal cell adhesion molecule. Subsequent studies identified AP1S2, a gene of subunit 2 of clathrin-associated adaptor protein complex 1 (Cacciagli et al., 2014;Saillour et al., 2007), and CCDC88C, which encodes DAPLE, a protein involved in Wnt signaling (Drielsma et al., 2012;Ekici et al., 2010;Ruggeri et al., 2018). A more recent gene linked to congenital hydrocephalus is MPDZ, encoding a large modular scaffold protein that consists of 13 PDZ and one L27 domains (Adachi et al., 2009;Ullmer et al., 1998). Several cases of severe congenital hydrocephalus identified in 5 consanguineous families (Al-Dosari et al., 2013;Saugier-Veber et al., 2017) were linked mostly to biallelic nonsense mutations that resulted in nonsense-mediated decay and total loss of MPDZ. A milder phenotype in a non-consanguineous family was linked to missense mutations and a heterozygous splice site variant (Al-Jezawi et al., 2018). The four genes that have been linked to congenital hydrocephalus most recently, TRIM71, SMARCC1, PTCH1, and SHH (Furey et al., 2018), regulate ventricular zone neural stem cell differentiation. Their loss-of-function is thought to result in defective neurogenesis, including ventriculomegaly.
Two of the genetically-linked congenital hydrocephalus conditions in humans were phenocopied in mice. Mice carrying loss of function mutations (LOF) in either L1CAM (Dahme et al., 1997;Rolf et al., 2001) or MPDZ (Feldner et al., 2017) developed severe hydrocephalus similar to human carriers of biallelic L1CAM (Kanemura et al., 2006) and MPDZ mutants (Al-Dosari et al., 2013;Saugier-Veber et al., 2017). Hydrocephalus results from impediment of the circulation of the cerebrospinal fluid (CSF), causing its accumulation in the brain ventricles as a result of either excessive CSF inflow, attenuated flow through the ventricles, or blocked outflow (Estey, 2016;Kahle et al., 2016). The L1CAM and Mpdz mouse models afforded anatomic and histological analysis for determining the nature of the defects that interfered with CSF circulation. L1CAM mice harbor stenosis of the aqueduct of Sylvius between the 3 rd and 4 th ventricles, but it was judged to be a result of the increased intraventricular pressure and the ensuing compression of the aqueduct's walls, rather than the cause of hydrocephalus (Rolf et al., 2001). The formation of hydrocephalus in Mpdz -/mouse was attributed to stenosis of the aqueduct (Feldner et al., 2017). Postmortem pathology of brains from several individuals harboring LOF MPDZ variants detected ependymal lesions but did not reveal causative mechanisms.
MPDZ is a cytoplasmic protein localized close to the junctions of epithelial (Hamazaki et al., 2002) and endothelial cells (Ernkvist et al., 2009), as well as to neuronal synapses (Krapivinsky et al., 2004). In the former cell types, MPDZ binds at least 8 junction transmembrane proteins (Adachi et al., 2009;Coyne et al., 2004; images of Mpdz -/mice (Fig. 2C). This CP configuration is similar to the morphology of the CP in human hydrocephalic brains (Al-Dosari et al., 2013;Cardoza et al., 1988). Unlike the Mpdz +/+ brains, we were able to identify visually the contrast-medium signal in T1-weighted MR images of Mpdz -/mouse brains (Fig. 2D). The location of the signal in T1-weighted coronal images of Mpdz -/brains corresponded to each other accurately to the location of the CP in the T2-weighted images. The trend of the time course of the MR signal intensity sampled in the T1-weighted coronal images was unambiguously upward, peaking within the 10 min duration of the experiments (Fig. 2E). These images indicate that the contrast-medium leaked through the CP of Mpdz -/mice into its abnormally enlarged and merged lateral ventricles.

The Sylvian aqueduct of the Mpdz -/mouse is stenotic
The MR-imaged Mpdz -/brain (Fig. 2C) and the comparison of HE-stained sections of Mpdz +/+ (Fig. 3A) and Mpdz -/- (Fig. 3B) brains indicated that the large cranial void in the brains of Mpdz -/mice resulted from the expansion and merger of the lateral ventricles, whereas the volume of the 3 rd ventricle did not change noticeably. Fixed brains sections do not maintain the original dimensions of the organ. Though it was not evident that the Sylvian aqueduct is stenotic (Fig. 3B), we injected Evans blue into the lateral brain ventricles of Mpdz +/+ and Mpdz -/mice. The aqueduct of the Mpdz -/mouse appeared stenosed in the ex vivo images of injected brain hemispheres (Fig. 3C). Unlike the lateral ventricle or the aqueduct and fourth ventricle of the Mpdz +/+ mouse, little of the injected dye seeped into the surrounding parenchyma during the overnight incubation of the brain in fixative. This indicates that the flow through the aqueduct of the Mpdz -/mouse was slower than in the Mpdz +/+ mouse. Similar to L1 -/mice (Rolf et al., 2001), aqueduct stenosis in Mpdz -/mouse could have been a result the compression of the brain rather than a cause of hydrocephalus.

Mpdz is localized apically in CPCEs; its deficiency induces depletion of tight junction proteins
We probed the CPs of Mpdz +/+ and Mpdz -/mice with antibodies specific to Mpdz and to several junction transmembrane and membrane-associated proteins to identify potential structural differences between their respective CPEC junctions. Mpdz was detected exclusively near the apical face of CPECs from third ventricle villi (Fig. 4A). This localization encompasses the sites of the tight junctions, which are the topmost structure in intercellular junctions. However, the conspicuous abundance of Mpdz on the CPEC apical surface suggests it may play an additional role unrelated to intercellular junction maintenance. The tight junction-associated scaffold protein ZO1 appeared more abundant in the CPEC monolayer of CP villi from the lateral ventricles of Mpdz +/+ mice (Fig. 4B), whereas the abundances of the adherens junction protein epithelial (E)-cadherin were similar to each other (Fig. 4C). Since quantification of the immunofluorescence of two-dimensional (2D) sections is not a robust measure of the overall abundance of the probed protein in the 3D CP, we opted to simulate the deficiency of Mpdz by knocking down MPDZ in human (h) primary CPECs by lentiviral transduction of MPDZ-targeted shRNA. We then used immunoblotting to compare the abundances of ZO1, Jam-C, and E-cadherin, to those of hCPECs transduced by non-targeting shRNA. The protein abundances measured by densitometry of the immuno-adsorbed protein bands were similar to those suggested by the corresponding immunofluorescence images (Fig. 4D), confirming that the abundances of ZO1 and Jam-C were lower in Mpdz -/mice or in hCPEC wherein Mpdz was knocked down, whereas that of E-cadherin did not change.

Epithelial cells of Mpdz -/-CP and their intercellular junctions are structurally and functionally defective
Transmission electron microscopy (TEM) detected substantial structural differences between the CPEC monolayers of Mpdz +/+ and Mpdz -/mice. While the structures of CP villi from the lateral ventricles of Mpdz +/+ and Mpdz -/mice imaged by TEM appeared grossly similar to each other (Fig. 5A), examination at higher magnification uncovered the presence of a large number of voids of varying sizes in the CPECs of Mpdz -/mice (Fig. 5B). Furthermore, the length of the adherens junctions was shorter, whereas the length of spaces between adjoining Mpdz -/-CPECs was longer in comparison to adjoining Mpdz +/+ CPECs (Fig. 5B). CPEC tight junctions, the foremost barrier to paracellular permeability (Zihni et al., 2016), between the CPECs of lateral ventricle villi from P18-P21 Mpdz -/mice were shorter, wider, and less electron-dense than those of Mpdz +/+ mice (Fig. 5C), indicating that their protein concentration was lower than that of Mpdz +/+ CPEC junctions. TEM imaging of CPEC tight junctions of P12-P14 mice revealed similar differences, though they were subtler than at P18-P21 (Fig. EV2). A second outstanding difference between Mpdz +/+ and Mpdz -/-CPECs was the state of the mitochondria. Numerous mitochondria in CPECs from lateral ventricle villi of Mpdz -/mice lacked a major part of their cristae, and some of those contained autophagosomes (Fig. 5D). It is possible that some of the voids in the Mpdz -/-CPECs were remnants of fully-dissolved mitochondria. The structural differences between the CPEC tight junctions in Mpdz +/+ and in Mpdz -/mice suggests that the former were less impervious to leakage than those between Mpdz +/+ CPECs, likely accounting for the contrast-medium leakage from the CP into the lateral brain ventricles of Mpdz -/mice. Unlike CPECs, we did not detect structural differences between the junctions of the fenestrated endothelial cells that comprise the walls of the CP capillaries in Mpdz +/+ and Mpdz -/mice (Fig. EV3).
To determine the dependence of the barrier function of CPEC monolayers on Mpdz, we compared the time course of impedance, an established permeability surrogate (Bischoff et al., 2016), of human papilloma CPEC (hpCPEC) monolayers (Feldner et al., 2017;Ishiwata et al., 2005) transduced by MPDZ-targeting or by nontargeting shRNA. The impedance of the control group was persistently higher throughout the 70-h duration of the measurement (Fig. 5E). While the MPDZ-deficient hpCPECs (Fig. 5E, inset) reached a plateau in approximately 60 h, the impedance of the control group of cells continued to rise, reaching a 42 percent higher impedance than the MPDZ-deficient hpCPECs. The lower impedance of these cells indicates they posed lower resistance to the alternating electrical current passing through the monolayer they formed, compared to the control group, and are, therefore, more permeable, in agreement with the structural findings.

Transcytosis through the CP is higher in Mpdz -/than in Mpdz +/+ mice
To fully characterize the barrier function of the CP, we compared the transcellular permeabilities of the CP of Mpdz +/+ and Mpdz -/mice by tracking fluid phase uptake and receptor dependent transcytosis. We used horseradish peroxidase injected in vivo for ex-vivo chromogenesis by hydrogen-peroxide-induced oxidation of 3,3'diaminobenzidine (DAB) (Broadwell and Brightman, 1983). The number of internalized DAB particles per cell in CPEC sections from lateral ventricle villi of Mpdz -/mice was approximately 6-fold higher than in those of Mpdz +/+ mice (Fig. 6A), suggesting that the rate of fluid phase uptake through the CPECs of Mpdz -/mice was substantially higher than its rate in Mpdz +/+ mice. Quantification of the number of DAB particles per cell showed that the fluid phase uptake by CPECs of Mpdz -/mice was more than 6-fold higher relative to that of Mpdz +/+ mice (Fig. 6B). Many of the larger endocytosed DAB particles consisted of concentric layers (magnified fields in Fig. 6A), a phenomenon seen in previous studies on ependymal cells (Broadwell and Sofroniew, 1993) and brain endothelial cells (Broadwell et al., 1996). The endocytosed DAB formed a morphologically heterogenous population of particles in the CPECs of both mouse genotypes, most of which were located in macropinosomes, following engulfment by ruffles on the CPEC basal surface (Fig. 6C). This is similar to previous observations on DAB endocytosis in mouse CPECs (Balin and Broadwell, 1988). We noticed a marked difference between the distributions of the macropinosomes in CPECS of Mpdz +/+ and Mpdz -/mice: whereas in the former the majority was located near the basal side facing the lumen of the CP villus (Fig. 6C), most macropinosomes were located near the apical side of the CPECs of the latter, facing the ventricle (Fig.  6D). In CPECs of Mpdz +/+ mice, more than twice macropinosomes were near the basal than near the apical side; in CEPCs of Mpdz -/mice, the ratio was reversed to more than 3-fold macropinosomes near the apical side (Fig. 6D). This difference between macropinosome distributions indicates that the rate of macropinosome transcytosis from the CPEC basal to apical side was higher in lateral ventricle villi of Mpdz -/than of Mpdz +/+ mice.
To compare receptor-mediated transcytosis through the CPECs of Mpdz +/+ and Mpdz -/mice, we focused on the abundance and endocytosis of the low-density lipoprotein (LDL) receptor, because apolipoprotein E (ApoE), an LDL carrier, was the most overabundant protein in the CSF of Mpdz -/mice (see below). Since LDL traverses the blood-brain-barrier by LDLR-mediated endocytosis, followed by transcytosis (Dehouck et al., 1997), we tested the presence of LDLR in CPECs of Mpdz +/+ and Mpdz -/mice. The higher intensity of the immunofluorescence signal emanating from CP sections from lateral ventricle villi of Mpdz -/mice indicated that LDLR was more abundant than in its Mpdz +/+ counterpart (Fig. 7A). The mean fluorescence intensity per CPEC of Mpdz -/mice was higher by approximately 40 percent (Fig. 7B). The higher transcytosis through the Mpdz -/-CP could have been a response to the condition of hydrocephalus that is unrelated to the Mpdz LOF or could have been induced by the absence of functional Mpdz. To test the causal connection between MPDZ expression and LDLR abundance, we knocked down MPDZ in hCPECs by lentiviral transduction of MPDZtargeting shRNA and compared the abundances of LDLR to hCPECs transduced by non-targeting shRNA. LDLR was more abundant in the MPDZ-deficient hCPECs by more than two-fold, suggesting that the depletion of MPDZ induced an increase in the amount of LDLR through a yet unknown pathway (Fig. 7C). To test the correlation between the abundance of MPDZ and the extent of LDLR transcytosis, we tracked the constitutive endocytosis (Zou and Ting, 2011) of endogenous LDLR in hCPECs that were transduced by MPDZtargeting or by non-targeting shRNA. LDLR was close to 2-fold more abundant on the cell surface prior to the initiation of endocytosis and higher by 54 percent after 8 min of constitutive endocytosis in MPDZ-deficient hCPECs (Fig. 7D).

Protein concentration is substantially higher in the CSF of Mpdz -/mice
The total protein concentration in the CSF of Mpdz -/mice at P18-P21 was more than twice higher than that of Mpdz +/+ mice (Fig. 8A). We compared the composition of the serum and the CSF of Mpdz +/+ and Mpdz -/mice by tandem liquid chromatography and mass spectroscopy (LC-MS/MS). The serum compositions of the two genotypes were highly similar, with only one over-abundant protein in the CSF of either Mpdz +/+ or Mpdz -/mice (Fig. 8B). In contrast, the composition of the CSF differed substantially between the two genotypes ( Fig. 8C). We detected a total of 313 proteins in all the samples pooled together, after excluding hemoglobin subunits a and b, catalase, peroxiredoxin and carbonic anhydrase-1 as serum contaminants (You et al., 2005) (Dataset EV1). Out of these, all but 13 proteins had been detected in murine CSF in previous studies (Cunningham et al., 2013;Dislich et al., 2015;Smith et al., 2014). The CSF of Mpdz -/mice contained 23 proteins that were either absent in the CSF of Mpdz +/+ mice, or that were at least 2-fold significantly more abundant in the CSF of Mpdz -/mice (Table 1). Only two proteins were at least 2-fold significantly more abundant in the CSF of Mpdz +/+ mice. The similarity between the serum protein compositions of Mpdz +/+ and Mpdz -/mice shows that protein over-abundance in the CSF of Mpdz -/mice is not a direct result of the serum composition but a consequence of the functional differences between the CPs of Mpdz +/+ and Mpdz -/-Mice.
All the proteins that were at least 2-fold more abundant in the CSF of Mpdz -/mice are secreted or known to have soluble forms (fibronectin, gelsolin, and vitronectin). They can be classified into several molecular function groups: components of the blood coagulation cascade (complement C4-B, complement factor H, properdin/complement factor P, and the fibrinogen b chain), extracellular matrix (extracellular matrix protein 1 (Ecm1), fibronectin, and vitronectin), lipoproteins (apoliproteins D and E), immune response (beta-2-microglobulin and the macrophage colony-stimulating factor 1 receptor), cytokines and cytokine binding proteins (a-fetoprotein/insulin-like growth factor binding protein 1 (Afp), a-2-HS-glycoprotein, granulin (Grn), hepatocyte growth factor activator, and insulin-like growth factor (IGF) binding proteins 2 and 4), enzymes and enzyme binding proteins (a-2-macroglobulin-P, lysozyme C-2, and sulfhydryl oxidase 1), and protease inhibitors (angiotensinogen and pigment epithelium-derived factor). Out of these, Afp, Ecm1, and Grn were found only in the CSF of Mpdz -/mice.
The overabundance of proteins in the of CSF of Mpdz -/mice could have resulted from an overall increase in the rate of CSF production. To test this premise, we compared the abundances of the Na + -K + -2Clcotransporter 1 (Nkcc1), recently shown to contribute significantly to CSF production (Steffensen et al., 2018), in the CP of Mpdz +/+ and Mpdz -/mice. Quantification of confocal images of CP sections indicated that Nkcc1 was approximately 75 percent more abundant in the CP from lateral ventricle villi of Mpdz -/mice (Fig. 8D). Similar to LDLR, knockdown of MPDZ in hCPECs was accompanied by an increase of approximately 68 percent in NKCC1 abundance (Fig. 8E).
The presence of blood coagulation cascade proteins in the CSF has been linked to neuro-inflammation (Ehling et al., 2011;Wang et al., 2011). ApoE is produced in the CP and secreted in response to neuronal injury (Lehtimaki et al., 1995). Its presence may indicate that the pressure exerted by the expanding hydrocephalus injured the brain parenchyma. The four IGF-binding proteins are inhibitors of the insulin growth factor like receptor (Kelley et al., 2002;Srinivas et al., 1993). Their over-abundance could be a negative feedback triggered by the swelling of the Mpdz -/brain, to suppress further IGF1-induced neurogenesis (Annenkov, 2009). Angiotensinogen is the precursor protein of angiotensins, generated by cleavage of its Nterminus by renin (Skeggs et al., 1957). The angiotensins maintain blood pressure homeostasis in response to fluid intake or loss (Gardes et al., 1982). Their production could have been triggered by the elevated brain blood pressure in hydrocephalic mice. Pigment epithelium-derived factor supports neuron differentiation and growth (Steele et al., 1993). The functional classifications of these proteins suggest that their over-abundance is part of a multifaceted physiological response to the stress imposed on the brain by the expanding hydrocephalus, rather than a haphazard collection of unrelated proteins. We did not detect lesions in the brain of hydrocephalus-harboring P21 mice either macro-or microscopically. Correspondingly, none of the overabundant proteins in the CSF of Mpdz -/mice is cytoskeletal; gelsolin has a secreted isoform (Yin et al., 1984). Furthermore, there was no statistically significant difference between the concentrations of serum albumin, amyloid A-4, amyloid P-component, and paraoxonase-1 in the CSF of Mpdz +/+ and Mpdz -/mice (Dataset EV1). It is unlikely, therefore, that the overabundance of these proteins reflects breakdown of the brain parenchyma.

DISCUSSION
Our results indicate that Mpdz LOF has a two-fold effect on the barrier function of murine CP: it increases both passive paracellular permeability, as indicated by the leakage of the GD-based MRI contrast-medium, and the rate of protein transcytosis, as indicated by the higher abundance of close to 10 percent of all the proteins found in the CSF of Mpdz +/+ and Mpdz -/mice. The first effect is likely caused by degradation of the integrity of CPEC intercellular junctions, because we did not find structural differences between the junctions of the fenestrated endothelial cells of CP vessels. The second effect, which is likely to be a physiological response to the stress inflicted on the brain by the formation of hydrocephalus, could involve both cell monolayers.
The first mouse model of hydrocephalus, generated by a L1cam LOF mutation, harbored neural defects, but the ependymal cells that coat the ventricle lumen appeared normal (Dahme et al., 1997). The brain malformations typical to the L1 syndrome were attributed to the inter-neuronal adhesion function of L1CAM (Miura et al., 1992), but no mechanism had been invoked to account specifically for the formation of hydrocephalus. AP1S2 is also located in the X chromosome, and like L1CAM, underlies several syndromic defects, one of which is hydrocephalus. In some cases, AP1S2-linked hydrocephalus was associated with stenosis of the Sylvian aqueduct (Saillour et al., 2007). No causal connection has been made between AP1S2 LOF mutations and the formation of hydrocephalus.
The hydrocephalus observed in carriers of CCDC88C mutations resulted presumably from a dysfunction of Wnt non-canonical signaling, because these mutations truncate the C-terminus of DAPLE, the protein it encodes, disabling its binding to the PDZ domain of DVL. Since DVL is required for the formation of Wnt-induced planar polarity of ependymal cells (Ohata et al., 2014), it is conceivable that CCDC88C-linked hydrocephalus was caused by the loss of the alignment of the cilia of the ependymal cells on the lumen of the ventricles, and the resulting slowdown of CSF flow. Furthermore, DAPLE binds to (Redwine et al., 2017) and functions as an activating adaptor of dynein (Reck-Peterson et al., 2018), a molecular motor that is required for cilium motility (Gibbons and Rowe, 1965). To date, the premise that CCDC88C mutations cause hydrocephalus by impeding ciliary function has not been tested.
The localization of MPDZ to epithelial and endothelial intercellular junctions, its binding of multiple transmembrane junction proteins, and its high abundance in the CP are collectively suggestive of a causal connection between the formation of hydrocephalus and MPDZ LOF mutations in humans and mice. Presumably, if MPDZ is required for the stabilization of transmembrane proteins at endothelial and epithelial cell junctions, its absence would impair junction integrity. Though this scenario had been entertained for both human (Al-Dosari et al., 2013;Saugier-Veber et al., 2017) and murine (Feldner et al., 2017) MPDZ-linked hydrocephalus, no direct causative connection has been established between possible MPDZ LOF deleterious effects on intercellular junction integrity and the formation of hydrocephalus in humans or mice carrying MPDZ LOF mutations. The first reported MPDZ mutation linked to severe congenital hydrocephalus in humans would have truncated 12 of the 13 PDZ domains (Al-Dosari et al., 2013). A subsequent study reported 3 new mutations that caused truncation of MPDZ within PDZ domain #3, a frame shift within PDZ domain #1, and a truncation within PDZ domain #5 (Saugier-Veber et al., 2017). All three mutations were expected to introduce a premature stop codon and result in nonsense-mediated decay of the transcript. In all three affected individuals, the aqueduct of Sylvius was stenotic, and the ependyma in the aqueduct as well as in the third and fourth ventricles linked by the aqueduct was interspersed with denuded focal lesions. Since MPDZ is abundant in aqueduct ependymal cells of normal individuals, the ependymal lesions were attributed to the formation of defective tight junctions, but there was no direct evidence for this eventuality.
The initiation of hydrocephalus in the Mpdz LOF mouse model of Feldner et al. was attributed to deterioration of ependymal cell integrity as a result of weakened tight junctions, similar to the explanation of MPDZ-linked human hydrocephalus. No evidence was provided to support this premise. In contrast, we provided direct evidence that the CP of Mpdz -/mice was leaky and structurally defective, implying that the hydrocephalus formed because of imbalance between the rate of CSF production and removal from the lateral ventricles through the relatively narrow 3 rd ventricle, aqueduct, and 4 th ventricle, resulting in the expansion and merging of the lateral ventricles into the large void we detected in the brains of Mpdz -/mice. Our pathophysiological model is not necessarily in conflict with the findings of Feldner et al. but would consider the ependymal lesions and stenosis of the Sylvian aqueduct observed by them as possible effects rather than causes.
The composition of the CSF of Mpdz -/mice is indicative of the stress posed on the brain by the swelling of the lateral ventricles. This response could have been instigated, conceivably, by the ependymal lesions observed by Feldner et al., as well as by other injuries to the brain parenchyma that became ultimately fatal at 3 weeks of age. The protein over-abundance in the CSF of Mpdz -/mice is likely the result of augmented transcytosis, a process that occurs normally in the CP (Grapp et al., 2013). The higher rate of fluid-phase uptake by the CPECs of Mpdz -/mice could have contributed to fluid accumulation in their lateral ventricles. The large imbalance between the protein contents in the CSF and the interior of the choroid plexus could have caused an osmotic pressure gradient between the two compartments, further driving fluid flow from the choroid plexus into the ventricles. At this time, we are aware of a single comparative proteomic analysis of CSF of normal and hydrocephalic subjects (Finehout et al., 2004). This study did not analyze, however, CSF from patients harboring congenital hydrocephalus, but that of a single subject with normal pressure hydrocephalus and from two normal patients. The study detected a relatively low total number of only 82 proteins, possibly because the samples were resolved by 2D electrophoresis rather than undergoing liquid chromatography prior to mass spectrometry. This may have reduced the sensitivity of the assay.
The manner in which the Mpdz loss-of-function mutation in vivo or Mpdz knockdown in vitro brings about an increase in the abundances of LDLR, and Nkcc1, and a decrease in the abundances of ZO1 and Jam-C, is unknown at this time.

MATERIALS AND METHODS Animals
Mice were housed in in ventilated cages and supervised by the Thomas Jefferson University Laboratory Animal Services Animal (Welfare Assurance Number D16-00051) and used according to protocols approved by the Institutional Animal Care & Use Committee and renewed annually. Thomas Jefferson University is accredited by the Association for Assessment and Accreditation of Laboratory Animal Care. C57BL/6J Mpdz +/mice (Milner et al., 2015) were kindly gifted by Dr. Kari Buck, Oregon Health and Science University). The Mpdz mutation was generated by Bay Genomics as gene-trap model MpdzGt(XG734)Byg. It translates to a G510Vfs*19 change in the protein sequence. Mpdz -/mice were bred by heterozygote crossing and used between P0 and P21. Within each genotype, the inclusion of gender and of individual mice was random. The reporting of in vivo experiments in this article conforms with the Animal Research: Reporting of In Vivo Experiments (ARRIVE) guidelines (Kilkenny et al., 2010).

Cell Culture, MPDZ Knockdown, and Immunoblotting
Human hCPEC (ScienCell Research Laboratories) were grown in Epithelial Cell Medium purchased from the same provider. Mpdz was knocked down by transduction with lentivirus expressing a validated shRNA (clone TRCN0000349430, Millipore-Sigma) or a non-targeting shRNA (pLKO.1-puro shRNA Control, Millipore-Sigma). The virus was produced in HEK293 cells (American Type Cell Culture) that were transfected (Fugene HD, Promega) with packaging and coat protein plasmids (pCMVR8.74 and pMD2.G, respectively, Addgene). Lentivirus particles were concentrated from the cell culture medium by centrifugation (Speedy Lentivirus Purification kit, ABM). hCPEC were lysed 3 days after lentivirus transduction in RIPA buffer (ThermoFisher Scientific (TFS)) supplemented with protease inhibitor cocktail tablets (Roche), and immunoblotted with LDLR antibody (Millipore-Sigma) diluted 1:1000. LDLR band density was quantified by an ImageJ plugin and normalized relative to the load-control band density (antibody to b-actin, Millipore-Sigma) diluted 1:2000.

PET imaging
Mice aged 18-21 days were injected through the tail vein with 0.4-0.5 mCi of [ 18 F]flourodeoxyglucose ( 18 FDG, Siemens Medical Solutions USA, Inc). and imaged as described (Seidler et al., 2006). Emission intensity was quantified by measuring the signal per unit area in two dimensional regions of interest (ROIs), after subtracting the background signal.

Magnetic Resonance Imaging
Mice aged 18-21 days anesthetized by 3 percent isoflurane inhalation were imaged with a 4.7 Tesla, 50 cm horizontal bore instrument by T1 and T2-weighted modalities at the Small Animal Imaging Facility of the University of Pennsylvania. Body core temperature was kept at 37°C by a rectal probe. Mice were injected through the tail vein with 50 µl 0.1 mmol/kg diethylene-triaminepentaacetic acid (DPTA) dimeglumine salt Gd chelate (Mr 938 Da, Magnevist®, Bayer) immediately after the completion of T2-weighted imaging, and underwent T1-weighted imaging for 10 min. Ventricle and total brain volumes were quantified by the ITK-SNAP (V3.2) application (http://www.itksnap.org) (Yushkevich and Gerig, 2017). Signal intensities in T1-weighted images were quantified at each time point by ImageJ (version 1.52d, https://imagej.nih.gov) at ROIs corresponding to the location of the ventricles in T2-weighted images of the same section. Change in total intensity relative to the initial value was normalized in each ROI by division by the difference between the maximum and minimum values.

Histology, immunohistochemistry, and optical microscopy
Anesthetized mice aged 18-21 were perfused intracardially with physiologic saline, followed by 4 percent paraformaldehyde (PFA) in PBS, pH 7.4. Brains were removed, fixed overnight in 4 percent PFA/PBS, immersed in 30 percent (w/v) sucrose in TBS, pH 7.4, at 4°C until they became submerged, and frozen to 180°C in an optimal cutting temperature medium (Fisher Scientific) Coronal sections were cut at 20 µm (HM560, Microm), mounted on glass slides (SuperFrost Plus, TFS) and air-dried overnight. Sections were blocked in 5 percent normal goat serum for 1 h at 23°C and incubated overnight at 4°C by primary antibodies to ZO1 (Invitrogen) diluted 1:200, LDLR (eBiosciences) diluted 1:100, E-cadherin (BD Biosciences) diluted 1:200, Jam-C (R&D Systems) diluted 1:200, Nkcc1 (Cell Signaling Tech.) diluted 1:200, and MPDZ (gift of Prof. Elior Peles, Weizmann Institute of Science (Poliak et al., 2002)) diluted 1:100. The primary antibodies were detected by secondary antibodies to the IgG of the appropriate host species of the primary antibodies, conjugated to Alexa Fluor® 488 or 555 (TFS) diluted 1:1000. The sections were mounted in anti-fade 4′,6diamidino-2-phenylindole (DAPI)-containing medium (ProLong, TFS). Immunofluorescence intensity was quantified with ImageJ by measuring the average pixel intensity within manually-defined contours of 2D cell images, to exclude effects of cell size. Only cells that were sectioned through the nucleus were quantified. HE-stained sections were imaged by scanning with a 10X objective (EVOS, FSI) and stitching the fields. Immunolabeled sections were imaged by laser-scanning confocal microscopy (Nikon A1R+).

Transcytosis
Mice aged 2-3 weeks were anesthetized and injected with HRP (0.5 mg/g body weight HRP type II, Sigma Aldrich) in PBS through the left ventricle. After 5 min circulation, mice were euthanized, and the brain was excised and immersed immediately in 4 percent PFA/PBS during CP dissection. The dissected CP was fixed by 5 percent glutaraldehyde and 4 percent PFA in 0.1 M sodium-cacodylate for 1 h at 23°C, followed by overnight fixation by 4 percent PFA in 0.1 M sodium-cacodylate at 4°C. After fixation, the tissue was washed three times and immersed overnight in 0.1 M sodium-cacodylate. CPs were then incubated in 0.01 percent hydrogen peroxide, and 0.5 mg/mL of DAB (Sigma Aldrich) in 0.05 M Tris-HCl pH 7.6 for 45 min at 23°C. The CPs were stored in 5 percent glutaraldehyde and 4 percent PFA in 0.1 M sodium-cacodylate until preparation for TEM. Transcytosis was quantified by counting the number of DAB-stained vesicles per cell.

LDLR endocytosis
Human papilloma CPECs grown on glass coverslips were incubated on ice with antibody to LDLR (eBiosciences) diluted 1:100, for 30 min, washed by ice-cold growth medium, and transferred to 37°C growth medium. After 8 min, the coverslips were washed by ice-cold PBS pH 2.5 for 30 s to remove remaining cell surface antibody, fixed with 4 percent PFA/PBS for 20 min @ RT, and permeabilized with 0.1 percent Triton X-100 in PBS for 5 min at 23°C. The primary antibody was detected by a 30 min incubation at 23°C with anti-chicken IgG conjugated to Alexa Flour 488 (TFS). Coverslips were mounted on slides (ProLong, TFS) before imaging by confocal microscopy (Nikon A1R+).

Evans blue intracranial injection
Mice anesthetized by ketamine/xylazine were injected through the left ventricle with five microliter Evans blue dye diluted 1:100 in PBS by a 10 ml syringe (Hamilton). The syringe was left in the injection site to prevent fluid reflux of fluid during the next 5 min, after which the mice were euthanized by decapitation. The heads were fixed immediately fixed in 4% paraformaldehyde/PBS overnight. Brains were dissected and imaged on a stereomicroscope (Leica MZ10F).

Electron microscopy
Brains removed as described were dissected under a stereomicroscope (Leica MZ10F) to isolate the choroid plexus from the lateral ventricles. Isolated choroid plexi were fixed overnight in 2 percent glutaraldehyde, stained with 1 percent OsO4 and 0.5 percent uranyl acetate, pelleted in 2 percent agarose (Millipore-Sigma, Type IX ultra-low gelling temperature), dehydrated in an acetone/water dilution series, and embedded in Araldite resin (EMbed 812, Electron Microscopy Sciences). Sections of 60 nm cut by ultramicrotome (Leica UCT) were mounted on square-mesh or oval-hole grids and imaged by TEM (FEI Tecnai G12) at 80 keV.

Mass spectrometry and measurement of CSF protein concentrations
Mice aged 18-21 days were anesthetized as described, their heads shaved, and immobilized stereotactically (Robot Stereotaxic, Neurostar) while maintaining body temperature at 37°C (TCAT-2LV controller, Physitemp). Anesthetic depth was verified by assessing reflexes before and during surgery. After peeling off the skin to expose the cranium, a 1.0 mm-diameter hole was drilled above the left lateral ventricle at -0.1 mm relative to the bregma, -0.80 mm from the midline, and 2.5 mm beneath the dura. A volume of 3-5 µL CSF was drawn from normal mice by a 10 µL syringe (Hamilton) at a rate of 1 µl/min and stored immediately on dry ice. CSF was drawn identically from hydrocephalic mice aside from the location of the hole that was drilled 2 mm left of the peak of the skull. A volume of 2 µL of each sample was reconstituted in 50 μL 6 M urea/100 mM Tris, pH 8.0 and digested by 3 μg trypsin overnight at room temperature. Samples were desalted by spin columns (Pierce™ C18, TFS) and reconstituted in 30 μL 1 percent acetic acid. Volumes of 50 μL of each sample were injected into a reversed-phase capillary chromatography column (Acclaim™ PepMap™ C18, Dionex), eluted by acetonitrile/0.1 percent formic acid gradient at a flow rate of 0.3 μL/min into the intake of a linear trap quadrupole (Orbitrap Elite™, TFS) hybrid mass-spectrometer and electro-sprayed at 1.9 kV. Amino acid sequences were determined from peptide molecular weights and ion collision-induced dissociation spectra by searching the Uniprot mouse protein database with MaxQuant (V1.5.2.8) application. CSF protein concentrations were measured by spectrometry of a 1:10 diluted CSF sample (NanoDrop, TFS).

Statistics
The significance of the difference between means was determined by two-tailed Student's t-test. The null hypothesis was considered untrue if the probability satisfied the condition P≤0.05. Based on initial sample variance, we increased sample size to test the statistical significance of inter-group difference, to fulfill the above criterion. The variances were similar between the tested groups, except for the control groups in densitometry measurements, which were set to 1. We excluded outliers if they were 1.5X the interquartile difference (between the medians of the upper and lower halves of the data set) above the top median or below to bottom median (Tukey's fence). This applied only to Fig. 7D. The normality of all data sets was confirmed by the Shapiro-Wilk test (Shapiro and Wilk, 1965), implemented by the Origin application (OriginLab). Sample identity was masked during data quantification of immunofluorescence images or of immunoblot bands. Blinding of investigators in regard to animals was not possible because the hydrocephalic mice were conspicuous.

DATA AVAILABILITY
The dataset of proteins identified in the CSF of Mpdz +/+ and Mpdz -/mice and their ratios is available in the PRIDE archive as PXD011535 (http://www.ebi.ac.uk/pride/archive/projects/PXD011535).

ACKNOWLEDGMENTS
We thank Dr. Kari Buck, Oregon Health & Science University) for sharing C57BL/6J Mpdz +/mice, Dr. Didier Trono, Lausanne Federal Polytechnic School, Switzerland, for sharing the pCMVR8.74 and pMD2.G plasmids, Dr. Gyorgy Csordas and Ms. Bodil Toma, Thomas Jefferson University, for assistance with TEM imaging; Prof. Elior Peles, Weizmann Institute of Science, Israel, for sharing a polyclonal antibody to MPDZ; Mr. Biao Zou and the University of Pennsylvania Electron Microscopy Resource Laboratory for preparing and mounting tissue thin sections for TEM; Drs. Yi Fan and Menggui Huang, University of Pennsylvania Department of Radiation Oncology, for help with MRI protocols; and Prof. Annie Laquerriere, University of Rouen, France, for sharing tissue samples.
This study was supported by the Cardeza Foundation for Hematologic Research (JY, CS, RK, IS, and AH); by the National Heart Lung and Blood Institute (R01 grant HL119984 to AH), by the National Cancer Institute (S10 grants OD012406 and RR23709 to MT; P30 CA56036 to the Bioimaging Shared Resource of the Sidney Kimmel Cancer Center, Thomas Jefferson University), and by the National Institutes of Health Office of the Director (S10 grant OD023436 to BW).

AUTHOR CONTRIBUTIONS
JY and CS designed and performed in vitro experiments; RK prepared mice for MRI and functional in vivo experiments, and isolated CPs; LO performed immunofluorescence experiments; MB harvested CSF and performed functional in vivo experiments; ZN performed TEM imaging; IS identified the first hydrocephalusharboring mouse, and managed the mouse cohorts; ST and FPK performed PET imaging, and MT supervised them and analyzed data; EL and PSV analyzed data and wrote the manuscript; BW supervised massspectrometry experiments and analyzed data; SP performed MRI and analyzed data; RS supervised the experiments performed by LO and MB, analyzed data, and wrote the manuscript; AH initiated the study, organized its performance, designed experiments, prepared mice for MRI, isolated CPs, analyzed data, and wrote the manuscript.

CONFLICT OF INTEREST
The authors have no competing financial interests.

THE PAPER EXPLAINED
Problem: Congenital hydrocephalus is a potentially life-threatening condition that occurs at a frequency of one case per 1000 births. Its main characteristic is accumulation of cerebrospinal fluid in the chambers of the brain, leading to swelling of the brain that is confined by the rigid skull. The only current treatment is shunting, an invasive procedure that fails within two years in 50 percent of the cases. Though this condition is hereditary, the manner by which a mutation results in hydrocephalus is not understood.
Results: Magnetic resonance was used to image contrast medium in 3-week old mice deficient of Mpdz, a large scaffold protein shared by humans. The medium leaked from the choroid plexus into the ventricles of the brain, showing that Mpdz is required for the integrity and function of the choroid plexus. The leakiness was caused by defects in the epithelial cell layer that surrounds the blood vessels of the choroid plexus. Another consequence of the loss of Mpdz was a more than twice higher protein concentration in the cerebrospinal fluid of the mutant mice.
Impact: Our study provides a straightforward connection between the mutation and the appearance of congenital hydrocephalus. The new insights may facilitate the development of non-invasive approaches for the treatment of the condition. Fig. 1: Hydrocephalus was detected by PET in Mpdz -/mice. A. Images of P4 (top) and P21 (bottom) Mpdz +/+ and Mpdz -/mice. Arrowheads point to the domed foreheads of the latter. B. Mean weights of P18-P21 Mpdz +/+ and Mpdz -/mice (n=8÷12, mean ± SD). C. Coronal, axial, and sagittal (top to bottom) PET images of P18-P21 Mpdz +/+ and Mpdz -/mice. The emission intensity is shown as a 7-point temperature scale from black (0) to white (7). D. Mean PET emission intensities in the indicated brain sections (n=4, mean ± SD).   The areas in the square frames are magnified in the bottom panels. B, C. CP sections from lateral ventricle villi immunolabeled for ZO1 (B) or E-cadherin (E-cad; C). The images are representative of 2 Mpdz +/+ mice and 3 Mpdz -/mice. Scale bars, top panels, 100 µm; bottom panels, 10 µm. D. Immunoblots with the indicated antibodies (top) and their quantifications. The densitometry measurements were normalized relative to the signal of the samples transduced by non-targeting shRNA (Ctrl). Note that the same b-actin immunoblot was used twice because the Jam-C and E-cadherin samples were immunoblotted on the same membrane (mean ± SD, n=3).  Fig. 6: Fluid-phase uptake by CPECs is higher in Mpdz -/relative to Mpdz +/+ mice. A. TEM images of CPEC sections from lateral ventricle villi of two Mpdz +/+ and two Mpdz -/-P14-P16 mice injected with HRP. The CPs were reacted with hydrogen peroxide and DAB ex-vivo. The dark particles are DAB deposits internalized by micropinocytosis. The magnified fields to the right show individual particles engulfed in macropinosomes. Note the layered structure of the particles, and the macropinosome that is open to the ventricular space in the Mpdz -/section. Scale bars, 1 µm; insets, 100 nm. B. Mean number of engulfed DAB particles per cell in Mpdz +/+ and Mpdz -/mice (mean ± SD, n=22). C. A CPEC section showing the engulfment of a DAB particle by its basal ruffles in the magnified field. Scale bars in panels C and D, 1 µm; insets, 200 nm. D. A CPEC section showing a preponderance of macropinosomes close to the apical face of the cell, and a magnified field that contains several macropinosomes. E. Mean numbers of DAB-containing macropinosomes close to the apical or basal sides of CPECs from Mpdz +/+ or Mpdz -/mice (mean ± SD, n=20).

C.
A volcano plot showing that 23 proteins are at least two-fold significantly more abundant in the CSF of P15-21 Mpdz -/mice (red circles, protein names are indicated) whereas only two proteins were more abundant in the CSF of Mpdz +/+ mice (green circles). The CSF composition was analyzed in 3 Mpdz -/and 3 Mpdz +/+ mice. D. Images of 10 µm-thick sections from lateral ventricle CP villi of Mpdz +/+ and Mpdz -/mice, immunolabeled for Nkcc1. The areas in the marked squares are magnified below. Scale bars, top, 50 µm, bottom 20 µm. The histograms show the mean fluorescence intensities per hCPEC in each cell group, normalized relative to the highest recorded intensity (mean ± SD, n=103). E. Immunoblots of LDLR and Mpdz, with b-actin as loading control. The MPDZ immunoblot shows the efficiency of the shRNA-mediated knockdown. Bands from three immunoblots were quantified by normalization to b-actin bands (mean ± SD). Table 1: Over-abundant proteins in the CSF of Mpdz -/mice, and their parameters in the LC-MS/MS analysis, grouped functionally into 7 sets (from top to bottom): extracellular matrix, blood coagulation cascade, lipoproteins, immune response, cytokines and cytokine-binding proteins, enzymes and enzyme-binding proteins, and protease inhibitors. LFQ, label-free quantification.

EXPANDED VIEW FIGURE LEGENDS
The Expanded View material consists of 3 figures and one data set.

Fig. EV1: No contrast medium is detected in the brain ventricles of Mpdz +/+ mice. A.
Coronal T2weighted MR image and an anatomically-corresponding HE-stained section. B. Triplicate rows of T2weighted and T1-weighted coronal images 1 min and 10 min post contrast-medium injection. The squares surround the locations of the lateral ventricles. Each row corresponds to one P18-P21 mouse. Scale bars, 1 mm. C. Time courses of the normalized T1-weighted image intensities corresponding to the locations of the areas surrounded by numbered squares in the T2-weighted images.