Rethinking the cilia hypothesis of hydrocephalus

Dysfunction of motile cilia in ependymal cells has been proposed to be a pathogenic cause of cerebrospinal fluid (CSF) overaccumulation leading to ventricular expansion in hydrocephalus, primarily based on observations of enlarged ventricles in mouse models of primary ciliary dyskinesia. Here, we review human and animal evidence that warrants a rethinking of the cilia hypothesis in hydrocephalus. First, we discuss neuroembryology and physiology data that do not support a role for ependymal cilia as the primary propeller of CSF movement across the ventricles in the human brain, particularly during in utero development prior to the functional maturation of ependymal cilia. Second, we highlight that in contrast to mouse models, motile ciliopathies infrequently cause hydrocephalus in humans. Instead, gene mutations affecting motile cilia function impact not only ependymal cilia but also motile cilia found in other organ systems outside of the brain, causing a clinical syndrome of recurrent respiratory infections and situs inversus, symptoms that do not typically accompany most cases of human hydrocephalus. Finally, we postulate that certain cases of hydrocephalus associated with ciliary gene mutations may arise not necessarily just from loss of cilia-generated CSF flow but also from altered neurodevelopment, given the potential functions of ciliary genes in signaling and neural stem cell fate beyond generating fluid flow. Further investigations are needed to clarify the link between motile cilia, CSF physiology, and brain development, the understanding of which has implications for the care of patients with hydrocephalus and other related neurodevelopmental disorders.


The standard model of hydrocephalus and the cilia hypothesis
Characterized by enlargement of the cerebrospinal fluid (CSF)-filled ventricles in the brain, hydrocephalus is a common neurological condition that can present at any age. In the newborn, hydrocephalus can arise as a secondary consequence from infections, intraventricular hemorrhage, or tumors. Infantile hydrocephalus that occurs in the absence of any known antecedent is classified as primary or congenital hydrocephalus. Affecting 1/1000 live births (Munch et al., 2012;Tully and Dobyns, 2014), hydrocephalus is the most common reason for brain surgery in children. On the opposite end of the lifespan, hydrocephalus often presents idiopathically in the older population as a condition termed normal pressure hydrocephalus with a prevalence of ~6% in individuals who are 80 years and older (Jaraj et al., 2014). Ventricular expansion in hydrocephalus is routinely attributed to pathologic overaccumulation of CSF in the ventricles due to increased fluid production, anatomical obstruction to CSF flow passageways (e.g. aqueductal stenosis), or decreased fluid reabsorption (Kahle et al., 2016;McAllister 2nd, 2012). Continued ventricular distention can compress the surrounding brain tissue and raise the intracranial pressure, leading to neurologic decline and even death if left untreated. The view of hydrocephalus as a fluid plumbing disorder has engendered neurosurgical shunting as the primary treatment strategy, which aims to reduce intraventricular CSF volume by insertion of a shunt that diverts CSF from the ventricles into the abdomen. Although neurosurgical CSF diversion can be life-saving, neurocognitive impairments still persist in some patients despite technically successful surgeries (Schiff et al., 2021;Riva-Cambrin et al., 2021;Vinchon et al., 2012), suggesting incomplete understanding of the pathogenic mechanisms leading to hydrocephalus.
Ciliary dysfunction has recently been proposed as an important pathophysiological mechanism of intraventricular CSF overaccumulation in multiple forms of hydrocephalus (Kumar et al., 2021;Ibañez-Tallon et al., 2004;Ji et al., 2022;Lee, 2013;Rodríguez and Guerra, 2017;Wallmeier et al., 2022;Wallmeier et al., 2020). Cilia are divided into two broad categories: primary cilia and motile cilia (Lovera and Lüders, 2021;Horani and Ferkol, 2021). Primary cilia are immotile and help to sense extracellular cues whereas motile cilia can rhythmically beat to help move fluid, mucus, or trapped bacteria along an epithelial surface (Horani and Ferkol, 2021;Fliegauf et al., 2007;Wallmeier et al., 2020). Motile cilia have two central microtubules and nine pairs of peripheral microtubules arranged in a "9 + 2" structure as well as two dynein arms that permit ciliary beating (Sakamoto et al., 2021). Primary cilia lack the central microtubules and thus have a "9 + 0" structure, and the absence of dynein motor arms render them immotile. However, there are some exceptions as not all motile cilia have the "9 + 2" and not all primary cilia have the "9 + 0" ultrastructure, but all motile cilia have dynein arms (Horani and Ferkol, 2021;Wallmeier et al., 2020). Some motile cilia also express signaling components that permit them to have sensory functions similar to primary cilia (Shah et al., 2009;Bylander et al., 2013). Prior to birth, primary cilia on ventricular neural stem cells detect extracellular signals to influence a variety of neurodevelopmental processes such as neuronal cell fate and migration (Guemez-Gamboa et al., 2014). After birth, the walls of the ventricles are lined by motile cilia emanating from ependymal cells (Coletti et al., 2018), and these ependymal cilia beat in an oar-like manner to generate a near-wall "ependymal flow" that has been proposed to help propel CSF throughout the ventricular system (Worthington Jr and Cathcart 3rd., 1963;Ibañez-Tallon et al., 2004;Ji et al., 2022;Delgehyr et al., 2015;Wallmeier et al., 2022;Olstad et al., 2019;D'Gama et al., 2021;Faubel et al., 2016).
The belief that motile cilia are important for CSF flow is based on observations of hydrocephalus and ventricular dilation in genetic mouse models of primary ciliary dyskinesia (PCD) (Lechtreck et al., 2008;Ibañez-Tallon et al., 2004;Abdelhamed et al., 2018;Chiani et al., 2019;McKenzie et al., 2015), a disorder of motile cilia dysfunction. These mouse models harbor loss-of-function mutations in genes that are essential for motile cilia function (notably Hydin and Mdnah5), and the observation of ventricular dilation in these mice is correlated with functional or structural perturbations in ependymal cilia that render them less able to generate fluid flow. Some human patients with mutations in motile cilia-related genes also exhibit hydrocephalus or ventricular dilation, such as patients with mutations in FOXJ1 (Jin et al., 2020;Wallmeier et al., 2019), CCNO Amirav et al., 2016), and MCIDAS (Boon et al., 2014;Robson et al., 2020). FOXJ1, CCNO, and MCIDAS are involved in ependymal cell differentiation and multiciliogenesis (Boon et al., 2014;Wallmeier et al., 2014;Jacquet et al., 2009;Núnez-Ollé et al., 2017;Lu et al., 2019). Other ciliarelated gene mutations have also been identified in normal pressure hydrocephalus patients , including CFAP43 (Morimoto et al., 2019). Complementing the genetic evidence, neuropathological examinations of post-mortem human brain tissues have revealed loss of multiciliated ependyma in primary and post-hemorrhagic hydrocephalus (McAllister et al., 2017;Guerra et al., 2015), suggesting impaired cilia-driven CSF flow in these hydrocephalic patients.
Here, we review and discuss the potential links between ependymal motile cilia, CSF movement, neural development, and the pathophysiology of hydrocephalus. Although it is clear that ciliary gene mutations are associated with hydrocephalus in the mouse, in humans it is rare for hydrocephalus to develop solely from motile cilia abnormalities, suggesting that other factors beyond cilia-mediated fluid dynamics are involved in disease pathogenesis (Sakamoto et al., 2021). We postulate that ciliary genes may have additional functions in neural signaling and neurodevelopment, and perturbations of these functions independently of cilia-mediated fluid movement may contribute to an increasingly recognized neurodevelopmental pathology in hydrocephalus. This may warrant a rethinking of the cilia hypothesis in hydrocephalus and a revisit of mouse mutant phenotypes previously attributed to loss of ciliagenerated fluid flow. A better understanding of cilia biology and pathophysiologic mechanisms in hydrocephalus will inform development of new treatment strategies for patients beyond the current standard practice of neurosurgical CSF diversion.

Cilia-generated flow is not a major mechanism for fluid dynamics before birth
Prior to birth, the walls of the ventricles are lined by neural stem cells that generate all neurons and macroglia of the cerebral cortex (Silbereis et al., 2016;Rakic, 1988;Bystron et al., 2008;Duy et al., 2022a). Ventricular neural stem cells abutting CSF harbor primary cilia that are important for detection of signaling cues involved in cerebrocortical development (Guemez-Gamboa et al., 2014). Unlike motile cilia emanating from ependymal cells, primary cilia on ventricular neural stem cells are immotile and incapable of generating fluid flow by themselves. At approximately midgestation in humans, some neural stem cells begin differentiating into ependymal cells, however ependymal cell maturation and the development of motile cilia along the ventricle wall are not completed until after birth (Coletti et al., 2018). Similarly, in the mouse, ependymal cilia are not functionally mature until approximately postnatal day seven (Duy et al., 2022b;Coletti et al., 2018;Spassky et al., 2005). Motile cilia are therefore unlikely to be a major contributor to CSF movement in the brain for much of gestation and perhaps even the early neonatal period.
Although hydrocephalus can present at any age, there are multiple forms of congenital hydrocephalus that present during fetal development and can be detected as early as gestation week 17, months before the birth of patients (Duy et al., 2022c;Patel et al., 2020;Pretorius et al., 1985). In fact, in a European cohort of congenital hydrocephalus patients, the majority (61%) was diagnosed prenatally at a median gestational age of 31 weeks (range 17-40 weeks) (Garne et al., 2010). Notable examples of genetic alterations leading to fetal hydrocephalus in human patients include L1CAM (Sullivan et al., 2021;D. Guo et al., 2020;Rosenthal et al., 1992), MPDZ (Al-Dosari et al., 2013;Shaheen et al., 2017), CCDC88C (Ekici et al., 2010), TRIM71 (Duy et al., 2022c), and SMARCC1 (Jin et al., 2020). The detection of fetal hydrocephalus in some patients thus correlates with developmental time points marked by neural stem cell proliferation and brain growth prior to the functional emergence of cilia-driven CSF flow at the ventricle wall. Multiple mouse mutant models, including those harboring mutations that disrupt motile cilia function (Tg737 and Ulk4 mutants), also exhibit hydrocephalus at birth before the maturation of ependymal cilia (Duy et al., 2022c;Banizs et al., 2005;Liu et al., 2016). Although loss of cilia-generated flow currents may theoretically exacerbate ventricular dilation in these mouse models, the early detection of hydrocephalus prior to cilia maturation suggests that ciliary dysfunction cannot be the initiating causal factor. Thus, developmental factors beyond cilia-related CSF movement should be considered in cases of fetal hydrocephalus that arise before the emergence of functional ependymal cilia.

Inconsistent link between loss of cilia-generated CSF flow and hydrocephalus in animal models
Despite the frequently reported association between ventricular dilation and ciliary dysfunction in the mouse literature, there are also reports suggesting that impaired cilia-driven CSF follow is neither required nor sufficient for the development of hydrocephalus. Numerous genetic mouse models of non-obstructive/communicating hydrocephalus do not exhibit ependymal cilia defects (Ito et al., 2021;Shimada et al., 2019;Duy et al., 2022a). There are also mouse and zebrafish models that surprisingly do not exhibit ventricular dilation and hydrocephalus despite altered ependymal ciliary function (Finn et al., 2014;Seo et al., 2021;Olstad et al., 2019), such as the mouse models nm1054 and bgh harboring mutations in the ciliary genes Pcdp1/Cfap221 and Spef2, respectively. The occurrence of ventricular dilation independently of motile cilia function suggests that perturbed ciliagenerated CSF flow is not the sole determinant of the hydrocephalus phenotype.
4. How much does cilia-generated CSF flow contribute to overall CSF dynamics in the human brain?
The standard model posits that CSF flows in one direction from its site of production to its site of reabsorption. Briefly, CSF flow starts from the lateral ventricle and moves into the third ventricle through the foramen of Monro, then crossing the cerebral aqueduct into the fourth ventricle, and finally entering the subarachnoid space where CSF is drained into the systemic circulation by arachnoid granulations. It is believed that this net unidirectional flow of CSF is assisted in part by flow currents generated by ependymal cilia (Khasawneh et al., 2018;Ji et al., 2022). Robust cilia-generated flow currents have been demonstrated by ex vivo imaging of ventricular explants from humans (Worthington Jr and Cathcart 3rd., 1963), rodents (Faubel et al., 2016;Ibañez-Tallon et al., 2004;Guirao et al., 2010;Abdelhamed et al., 2018), and pigs (Faubel et al., 2016). Similarly, in vivo imaging of zebrafish and frog models has also demonstrated the existence of CSF flow generated by beating of ependymal cilia (Olstad et al., 2019;D'Gama et al., 2021;Date et al., 2019).
Multiple lines of in vivo evidence from humans do not support a major role for cilia-mediated flow currents in driving CSF movement across different ventricular compartments. Noninvasive neuroimaging has shown that CSF movement in the human brain is not unidirectional as predicted by the standard model. Instead, intraventricular CSF flow is bidirectional and pulsatile in accordance with the heart beat (Shinya Yamada, 2014;Greitz, 1993;Greitz et al., 1992). In the human ventricle through the endoscope, a video motion analysis shows that the brain's ventricles have distributed motion at cardiac frequency, and that CSF pressure distributions also vary at cardiac frequency . CSF pulsations do not vary at known ependymal ciliary frequency (O'Callaghan et al., 2012), and the human ependymal cilia have no SA node innervated by a vagus nerve to signal them the cardiac frequency. These observations suggest that CSF dynamics in the human brain may reflect intracranial arterial and venous pulse wave coupling (Butler, 2017;Koch et al., 2022), pointing to the heart rather than cilia as the mechanical driver of CSF movement in the human brain. The movement of blood in and out of the cerebrovascular tree with every heartbeat results in brain motion that are in turn responsible for compression of the ventricular system and hence for the intraventricular flow of CSF (Greitz et al., 1992). Consistent with the human data, a larval zebrafish study found that flow currents generated by ependymal cilia actually serve to confine CSF within individual ventricular cavities, whereas movement of CSF across ventricular compartments is driven by the heartbeat or bodily movement (Olstad et al., 2019). These data suggest that the causal links between abnormal ependymal ciliary function, presumably altered CSF transport, and ventricular dilation remain to be established (Duy et al., 2022a;Sakamoto et al., 2021).
Rather than generating net CSF movement across ventricular compartments, other functions for cilia-generated flow currents in brain homeostasis have been proposed. One hypothesis is that cilia-generated flow is important for maintaining anatomical patency between narrow sites of CSF passage, such as the cerebral aqueduct that connects the third to the fourth ventricle (Ibañez-Tallon et al., 2004;Olstad et al., 2019). Thus, defective ciliary beating may lead to stenosis of the aqueduct and thereby obstruct the flow of CSF to cause fluid overaccumulation and hydrocephalus (Ibañez-Tallon et al., 2004). Ependymal cilia beating may also help to protect the ventricle wall against shear stress induced by intraventricular CSF pulsations (Shigeki Yamada et al., 2021). From the perspective of neuronal signaling, ciliagenerated flow currents may generate intraventricular boundaries of CSF and thereby modulate the spread of locally released chemical compounds in order to influence signaling at the brain-CSF interface (Faubel et al., 2016;Olstad et al., 2019).

Human genetics suggests altered neurodevelopment, rather than abnormal fluid dynamics, in hydrocephalus
Genetics is perhaps the most clinically relevant and unbiased approach to understanding the pathogenesis of complex diseases (Chong et al., 2015). The primary evidence underlying the cilia hypothesis of hydrocephalus has come from mouse models of motile ciliopathies (namely PCD) wherein genetic perturbations affecting motile cilia function frequently lead to hydrocephalus (Ibañez-Tallon et al., 2004;Lechtreck et al., 2008). Despite the association between hydrocephalus and loss of cilia-generated flow demonstrated by mouse genetics, motile ciliopathies rarely cause hydrocephalus in humans (Lee, 2013;Sakamoto et al., 2021;Ringers et al., 2020). For instance, while Hydin and Mdnah5 mouse mutant models are two of the most well-known PCD mouse models with hydrocephalus (Lechtreck et al., 2008;Ibañez-Tallon et al., 2004), human patients with mutations in HYDIN or DNAH5 have not been reported to exhibit hydrocephalus (Olbrich et al., 2012;Olbrich et al., 2002). In fact, the prevalence of hydrocephalus in human PCD has been reported to be as low as 1.3% in some clinical series (Behan et al., 2016). These observations suggest that a bona fide genetic defect affecting motile cilia structure and function may perhaps increase the risk of hydrocephalus in humans but may not be sufficient to cause hydrocephalus in all cases and that additional factors beyond fluid dynamics may be involved. The reduced frequency of hydrocephalus in human PCD compared to that in mouse models may reflect speciesspecific differences in the threshold to develop hydrocephalus when cilia-generated CSF flow currents are lost (Olbrich et al., 2012;Ibañez-Tallon et al., 2004;Ringers et al., 2020). Humans have much larger brain ventricles compared to the mouse, and given the hypothesis that ciliagenerated flow is particularly important at narrow sites of CSF passages (Ibañez-Tallon et al., 2004), the smaller and narrower mouse ventricles may be more susceptible to developing ventricular expansion when cilia-generated flow is lost due to genetic perturbations.
There are certain cilia gene mutations, however, that cause highly penetrant hydrocephalus in humans. The best known example is FOXJ1, de novo mutations in which are always associated with hydrocephalus (Wallmeier et al., 2019;Jin et al., 2020). However, FOXJ1 mutant patients also exhibit systemic symptoms such as chronic destructive airway disease and randomization of left/right body asymmetry (Wallmeier et al., 2019). The primary clinical features of PCD and other motile ciliopathies are related to dysfunction of motile cilia found in other organ systems outside of the brain, namely recurrent pulmonary infections, situs inversus (anatomical inversion of chest and abdominal organs), and fertility issues (Wallmeier et al., 2020). Thus, while mutations of some cilia genes can lead to highly penetrant hydrocephalus, these forms of hydrocephalus are usually always associated with systemic symptoms referable to dysfunction of extracerebral motile cilia (Wallmeier et al., 2019;Morimoto et al., 2019), unlike the majority of non-syndromic human hydrocephalus that does not exhibit the clinical concomitants of motile ciliopathies. Furthermore, the diagnosis of "hydrocephalus" may be debated in some of these cases that arise from cilia mutations. For instance, MCIDAS mutations are associated with enlarged ventricles on neuroimaging, however these patients do not exhibit adverse neurological sequelae or signs of elevated intracranial pressure despite ventriculomegaly (Robson et al., 2020;Boon et al., 2014). A diagnosis of hydrocephalus is based not only on radiographic findings of ventricular dilation, but also neurological symptoms suggestive of elevated intracranial pressure. Thus, some of these cases of asymptomatic "hydrocephalus" should actually be considered as radiographic ventriculomegaly (sometimes also termed arrested or compensated hydrocephalus) rather than clinical "hypertensive" hydrocephalus that often requires neurosurgical CSF diversion as a temporizing measure. Consistent with the finding of asymptomatic ventriculomegaly in some patients with cilia gene mutations, the intracranial pressures in two mouse models of motile ciliopathies are in fact normal despite ventricular enlargement (Xue et al., 2022), suggesting that impaired fluid dynamics stemming from ciliary dysfunction alone is not always sufficient to account for the neurological symptoms of clinical hydrocephalus.
Recent hypothesis-generating genomic studies in humans have suggested a neurodevelopmental component of hydrocephalus that perturbs brain structure and integrity leading to secondarily enlarged ventricles, rather than a primary disturbance of fluid dynamics (Jin et al., 2020;Duy et al., 2022a;Furey et al., 2018;Hale et al., 2021). Whole-exome sequencing studies of patients with surgically treated congenital hydrocephalus have discovered mutations in genes crucial for embryonic neural stem cell fate (Furey et al., 2018;Jin et al., 2020), including TRIM71 (Duy et al., 2022a;Chen et al., 2012;Duy et al., 2019;Ecsedi and Grosshans, 2013;Mitschka et al., 2015), SMARCC1 (Harmacek et al., 2014;Narayanan et al., 2018), PIK3CA , PTEN (Groszer et al., 2001;DeSpenza et al., 2021), and PTCH1 (Cohen et al., 2015). In this cohort of congenital hydrocephalus patients, mutations in genes related to cilia structure and function are not enriched (Jin et al., 2020). Genetic investigations of familial congenital hydrocephalus have also identified mutations in genes linked to prenatal neural stem cell progression and cortical formation (Shaheen et al., 2017), including EML1 (Jabali et al., 2022) and WDR81 (Carpentieri et al., 2022). Integrating genetic data with large-scale transcriptomic atlases of human and mouse brain development revealed the convergence of congenital hydrocephalus genetic risk in neuroepithelial cells (Duy et al., 2022a), the earliest neural stem cells of the human brain, but not in ependymal cells and choroid plexus cells. Similarly, a transcriptome-wide association study suggested that genetic risks for hydrocephalus are linked to brain structure and integrity rather than fluid dynamics (Hale et al., 2021).
Physiologically, a genetic defect in brain development may lead to deficient neural stem cell proliferation and cortical neurogenesis, resulting in a hypoplastic cortex that is mechanically unstable ("floppy") and unable to hold the pressure exerted by CSF due to loss of tissue elasticity and stiffness (Duy et al., 2022a;Peña et al., 2002;Fransen et al., 1998;Pang and Altschuler, 1994;Zee and Shapiro, 1989). Indeed, studies of brain biomechanics have demonstrated abnormal viscoelasticity and reduced stiffness in the brain tissue of hydrocephalic human patients and animal models (Duy et al., 2022a;Zee and Shapiro, 1989;Olivero et al., 2016;Wagshul et al., 2021). Altered brain structure and integrity may therefore disrupt brain-CSF biomechanical interactions, facilitating passive intraventricular CSF pooling and secondary enlargement of the ventricles even in the absence of a primary disturbance in CSF flow such as loss of cilia-generated currents or anatomical obstruction. In other words, the pathology is arising not arising from the fluid compartment but rather from the "vessel" (brain parenchyma) that is unable to contain the fluid (Duy et al., 2022a;Peña et al., 2002;Fransen et al., 1998;Pang and Altschuler, 1994;Zee and Shapiro, 1989), shifting the perspective of hydrocephalus from that of a fluid "plumbing" condition to that of a neural parenchyma disease with secondary enlargement of ventricles.

Roles of ciliary genes in signaling and development beyond fluid dynamics
A hydrodynamics-focused view of cilia biology in the context of hydrocephalus may have the unintended consequence of overlooking the developmental functions of nonmotile primary cilia in regulating neural stem cell fate that may also contribute to brain parenchymal pathology in hydrocephalus (Guemez-Gamboa et al., 2014;Duy et al., 2022a). Indeed, embryonic neural stem cells lining the in utero ventricles harbor nonmotile primary cilia that project into the CSF, and these cilia are crucial for the detection of chemical signals such as Shh that influence a variety of neurodevelopmental processes, including neural tube patterning, neuronal cell fate, neuronal migration, differentiation, and white matter development (Guemez-Gamboa et al., 2014;J. Guo et al., 2019;Hasenpusch-Theil and Theil, 2021). Cilia are also found in the choroid plexus, where they are thought to be involved in choroid plexus signaling and regulation of fluid transport (Narita and Takeda, 2015). Impaired primary cilia function could therefore lead to dysregulated neural signaling that contributes to a neurodevelopmental pathology in hydrocephalus. Indeed, mouse models of defective primary cilia function exhibit hydrocephalus, the mechanism of which is attributed to impaired neuroprogenitor proliferation and neurogenesis rather than a disturbance in CSF flow (Carter et al., 2012;Foerster et al., 2017). The classic division between one type of cilium as a sensory organelle and another as motile is blurring as sensory functions have been identified in motile cilia (Jain et al., 2012;Shah et al., 2009). This suggests the possibility that gene mutations previously thought to affect exclusively affect ciliary beating and fluid dynamics may also impact signaling functions that may affect brain parenchymal development. For instance, the progressive hydrocephalus (prh) mutant mouse that harbors a point mutation in the Ccdc39 gene exhibits not only loss of ciliamediated CSF flow, but also brain parenchymal phenotypes such as decreased numbers of neuroprogenitors in the ventricular/subventricular zones, reduced myelination, impaired maturation of excitatory synapses (Iwasawa et al., 2022;Abdelhamed et al., 2018). FOXJ1, which is a transcription factor classically associated with motile cilia, has also been found to influence the sensitivity of neuroprogenitor cells to Shh signaling by modulating the length of primary cilia (Cruz et al., 2010). Thus, ciliary gene mutations may contribute to hydrocephalus not only via problems from improper ciliary beating and generation of CSF flow currents, but rather via the role of cilia in signaling from the CSF to the brain compartment.
Beyond directly influencing cilia function, such as motility or ciliamediated signaling, ciliary gene mutations may also impact other cellular processes that contribute to development of hydrocephalus. For instance, there are classes of cytoplasmic dynein motor proteins that are involved in interkinetic nuclear migration of ventricular neural stem cells during neurogenesis (Hu et al., 2013;Tsai et al., 2010) and retrograde transport in axons (Twelvetrees et al., 2016;Schnapp and Reese, 1989;Wang et al., 1995). Although only the dyneins that are assembled into the dynein arms are involved in cilia motility, it is possible that these cilia dynein arm proteins may have nonciliary functions. Indeed, we examined a published single-cell RNA sequencing dataset of the developing human brain (Li et al., 2018) and found robust expression of DNAH14 in neuroepithelial and radial glia neuroprogenitors (see Fig. 1), suggesting a neurodevelopmental role of DNAH14 beyond its known function in motile cilia. Thus, while ventricular dilation observed in dynein mutant mouse models and human patients (such as the Mdnah5 mutant mouse and DNAH14 patients) has been attributed to loss of ciliagenerated CSF flow (Ibañez-Tallon et al., 2004;Kageyama et al., 2016), there remains the possibility that the pathology may have actually originated earlier from impaired dynein function during brain parenchymal development. Impaired fluid dynamics stemming from motile cilia dysfunction may thus be a secondary consequence that could exacerbate the hydrocephalus phenotype but may not be the initiating causal factor for disease pathogenesis. Consistent with the hypothesis that dynein mutations may impact neurodevelopment in addition to fluid dynamics, mutations in the dynein genes DNAH2 and DNAH14 have been reported in patients with microcephaly (J. Li et al., 2022;Duerinckx et al., 2021), a disorder classically associated with deficient neuroprogenitor proliferation and cortical hypoplasia (Lancaster et al., 2013). As discussed above, altered neurodevelopment may lead to hydrocephalus by altering the biomechanical stability of the brain parenchyma and thus facilitate secondary enlargement of ventricles from a "floppy" cortical mantle that is unable to resist the pressure exerted by CSF. This warrants future investigations to examine the nonciliary impact of ciliary gene mutations and clarify the neurodevelopmental mechanisms underlying hydrocephalus previously attributed to loss of cilia-generated CSF flow. Of particular importance will be the use of cellspecific knockout models to more thoroughly examine the spatiotemporal requirement of cilia genes in brain development and hydrocephalus, since most of the existing cilia mutant mouse models are full body knockouts that do not permit the assessment of which cell types and timepoints in development are involved in disease pathogenesis.

Conclusion
The conceptual framework of hydrocephalus as a fluid "plumbing" problem has not changed over the past century since Walter Dandy introduced the bulk flow model of CSF circulation (Dandy, 1919). Treatment strategies for hydrocephalus over the past decades are still directed almost exclusively towards neurosurgical CSF diversion and reduction of ventricle size albeit with incremental improvements in surgical technique and hardware. Paradoxically, even with technically successful reduction in CSF volume, neurodevelopmental outcomes may still fail to improve and the cortical mantle may not re-expand after surgery in children with hydrocephalus Schiff et al., 2021;Preuss et al., 2015;Sobana et al., 2021). The persistence of abnormal neurodevelopmental outcomes suggest that irreversible damage from neural tissue compression has already occurred by the time of shunting, or alternatively, that CSF overaccumulation is not the primary disease mechanism but rather a secondary consequence of another process that is not adequately addressed by neurosurgical CSF diversion alone.
Despite the clinical recognition that neurosurgical CSF diversion does not satisfactorily address the mechanisms of disease, the hydrocephalus literature has only further reinforced the role of impaired CSF circulation in disease pathogenesis based on the more recent hypothesis that dysfunction of motile cilia begets CSF overaccumulation in hydrocephalus (Ibañez-Tallon et al., 2004;Lechtreck et al., 2008). The cilia hypothesis of hydrocephalus, in particular the fluid dynamics-focused view of ciliary genes, has therefore only further reinforced CSF diversion as the primary treatment strategy. However, as we have reviewed here and also discussed by others (Sakamoto et al., 2021), loss of ciliagenerated CSF flow is unlikely to be the sole cause of hydrocephalus, especially in human patients. A rethinking of the cilia hypothesis of hydrocephalus is therefore warranted in order to consider the signaling and developmental functions of ciliary genes beyond fluid dynamics, which may also spur a revisit of many mouse mutant phenotypes previously attributed to loss of cilia-mediated flow currents. Clinically speaking, a rethinking of the cilia hypothesis also suggests the need to develop alternative treatment strategies in addition to CSF diversion in hydrocephalus, such as gene therapy or pharmacological approaches directed at optimizing brain development (Duy et al., 2022b;Duy et al., 2022c;Rodríguez and Guerra, 2017;Guerra et al., 2015). Given that ependymal cilia have also been hypothesized to be involved in the pathogenesis of neuropsychiatric conditions such as depression (Seo et al., 2021) and schizophrenia (Eom et al., 2020), which also exhibits ventricular dilation as a neuroradiographic feature (Styner et al., 2005;Bethlehem et al., 2022), a better understanding of cilia biology in the Fig. 1. Expression of DNAH14 in different cell types of the prenatal human brain. Gene expression was analyzed from the Psy-chENCODE single-cell RNA sequencing of the prenatal human brain dataset (Li et al., 2018). The violin plot was downloaded from http://development.psychencode.org. Astro: astrocytes, Endo: endothelial cells, ExN: excitatory neurons, InN: inhibitory neurons, IPC: intermediate progenitor cells, NasN: nascent neurons, NEPRGC: neuroepithelial/ radial glia cells, Oligo: oligodendrocytes, OPC: oligodendrocyte precursor cells.
context of brain function and homeostasis will lead to improved care of not only patients with hydrocephalus but also other related neurodevelopmental disorders (Duy et al., 2022b).

Data availability
Human brain gene expression data were downloaded from the publicly available PsychENCODE database at http://development.psychencode.org/.