The morphological and functional diversity of apical microvilli

2.1. Photoreceptor cells

Photoreceptor cells are retinal cells tailored to detect photons of light and create a neuronal response. To manage this challenging task, vertebrate photoreceptors develop an enlarged and modified cilium, known as the outer segment (OS), which accommodates large amounts of photopigment in numerous stacked membranous disks or lamellae (Goldberg et al., 2016). With the OS devoted to phototransduction, the metabolic demand of the cell is fulfilled by the adjacent inner segment, and the two regions are connected by a narrow connecting cilium. A microtubule‐based axoneme originates from a basal body in the apical inner segment and passes through the connecting cilium to form the backbone of the OS. Also emerging from the apical inner segment are the actin‐based CPs, which encircle the base of the OS. Although CPs were identified in early photoreceptor studies (Brown et al., 1963; Schultze, 1872), how they function to support vision remains to be elucidated.

The sensory endings of photoreceptors are constantly being renewed, with new disks or lamellae evaginating at the base of the OS, then translocating upwards before being shed off at the distal end (Goldberg et al., 2016). The shed fragments are phagocytosed by the adjacent retinal pigment epithelium (RPE). Depending on the OS shape and light sensitivity, photoreceptors can be classified as rods or cones. In the cylindrically‐shaped OS of rods, which mediate monochromatic vision in low light conditions, newly generated disks become enclosed by the plasma membrane of the OS. Conversely, in the conical‐shaped cone OS, the membranous folds remain largely continuous with the plasma membrane as lamellae. Cones are active in well‐lit environments to mediate color and high‐acuity vision.

In rods and cones, light triggers the activation of an opsin molecule, leading through the phototransduction cascade to the degradation of cGMP. As cGMP maintains cyclic nucleotide‐gated (CNG) channels in the plasma membrane in the open state, light triggers closure of the CNG channels, reduced influx of cations (mainly Na⁺ and Ca²⁺), and hyperpolarization of the cell. The resulting reduction in synaptic glutamate release has the surprising function of activating the downstream ON‐bipolar cells. As the OS is the site of phototransduction, the need for the surrounding CPs is unclear.

CP presence and structural features vary according to the species. Photoreceptor microvilli have been observed in, among others, humans, monkeys, pigs, chickens, and various amphibian and fish species (Brown et al., 1963; Olson, 1979; Sahly et al., 2012). Photoreceptors of rhesus macaque monkeys exhibit ~10–16 CPs evenly spaced around cone and rod OS (Steinberg et al., 1980). Analysis of the related cynomolgus macaque photoreceptors revealed similar morphology: cones have 14–16 processes with an average length of 3 μm or one‐fifth of the length of the OS; rods show more irregularity, with ~10 processes per cell and an average length of 1.8 μm, but the occasional presence of long processes up to 15 μm (Sahly et al., 2012). Cones of green sunfish were found to have 23–26 CPs of approximately 5 μm in length (Nagle et al., 1986). A recent report used expansion microscopy to image the photoreceptors of cynomolgus macaques, revealing that CPs form a C‐shaped cage around the base of the cone OS, with a gap at the connecting cilium and axoneme. Rod CPs were not detected, which could be an artifact of the technique (Verschueren et al., 2022).

Notably, the actin cores of CPs extend several microns into the inner segment as rootlets, often merging into fewer bundles before reaching the level of the outer limiting membrane (OLM), which is the location of adherens junctions between photoreceptors and Müller glia (Höfer & Drenckhahn, 1993; Steinberg et al., 1980).

Although present in many species ranging from fish to humans (Figure 2), CPs are largely absent in the photoreceptors of nocturnal rodents (Sahly et al., 2012; Zou et al., 2019). One study identified only an occasional vestigial CP on mouse photoreceptors and an absence of actin filament bundles in the inner segment (Sahly et al., 2012). Volland et al. (2015) reported one large tongue‐like process emerging from mouse rods, although it did not resemble CPs found in other species. Since CP function is not yet fully understood, it is unclear whether rodents utilize compensatory mechanisms or simply do not require CPs for their photoreceptor homeostasis.

2.2. Sensory hair cells

Hair cells are the mechanosensory receptors in the auditory, vestibular, and lateral line (acoustico‐lateralis) organs of vertebrates. By a process called mechano‐electrical transduction (MET), all sensory hair cells transform mechanical stimuli into electrochemical signals (Hudspeth, 1985). In the vertebrate inner ear, the mechanosensory organs are subfunctionalized to detect either vestibular or auditory information. Vestibular hair cells are organized in sensory patches located in either the semicircular canal cristae or the utricle and saccule (vestibular maculae); the hair cells in the latter are associated with acellular carbonate stones known as otoliths or otoconia. The vestibular system detects angular and linear acceleration of the head, providing a sense of spatial orientation, balance, and gaze stability (Burns & Stone, 2017). Morphologically, the vestibular organs are similar in all vertebrates. Conversely, vertebrate auditory organs are much more diverse. For example, the primary auditory organ in fish (the saccule) is a field of hair cells associated with a single, large otolith (Popper & Fay, 1993). By comparison, the coiled structure of the mammalian cochlea and its precisely patterned rows of inner and outer hair cells (OHCs) is highly specialized for sound detection in a terrestrial environment (Robles & Ruggero, 2001). A single row of inner hair cells (IHCs) acts as the true acoustic detector, while three rows of OHCs provide signal amplification. Despite the anatomical diversity in vertebrate hearing organs, all auditory hair cells detect sound pressure waves, whether they originate in the air or in the water. Lastly, the mechanosensory lateral line is a hair cell‐based sensory system distributed over the surface of aquatic vertebrates (fish and amphibians) (Bleckmann, 2006). The hair cells of the lateral line are clustered in sensory organs known as neuromasts. By virtue of their direct contact with the surrounding aqueous environment, the lateral line provides information about near‐field water particle motion, such as might be produced from predators, prey, and conspecifics, as well as other hydrodynamic features such as water flow (Mogdans, 2019). The physical and molecular mechanisms of MET are fundamentally similar in all vertebrate hair cells (Holt et al., 2021). The fact that different hair cells detect different types of mechanosensory stimuli is more a reflection of anatomical and organizational differences between the auditory, vestibular, and lateral line organs and less about fundamental differences in hair cell function.

The defining feature of hair cells is the “hair bundle”—a collection of projections (“hairs”) on the apical surface that function as the mechanosensory organelle of the cell. Hair bundles consist of at least three rows of actin‐filled microvilli called stereocilia that are ordered in a unique staircase arrangement (Tilney, Tilney, & DeRosier, 1992b). Stereocilia can vary in length dramatically between species and between types of hair cells within species, with documented lengths of around 1 μm up to 120 μm (Manor & Kachar, 2008). There is also diversity in the number of stereocilia per hair cell, with ranges of sometimes fewer than 20 up to 300 (Roberts et al., 1988). Despite the diversity in vertebrate hair cell morphology, each type of hair cell follows a stereotyped course of development with high fidelity regarding the timing of bundle development, the number of individual stereocilia and their number of rows, stereocilia height, width, and shape. This impressive, multiparameter precision in bundle morphology suggests intricate regulation of the actin core as the stereocilia develop from an initial “lawn” of short, undifferentiated microvilli (Barr‐Gillespie, 2015).

In addition to their distinctive microvilli, hair bundles also feature a non‐motile, microtubule‐based primary cilium (kinocilium) that abuts the middle of the tallest row of stereocilia (Elliott et al., 2018; Wang & Zhou, 2021). Kinocilia vary in length depending on the species and type of hair cell. Although the kinocilium plays an essential role in organizing the bundle during the development of all hair cells, its role in mature hair cells differs depending on the type of hair cell and the vertebrate species. In mammalian cochlear hair cells, the kinocilium degenerates as the bundle matures and therefore is not directly involved in mammalian auditory perception. However, degeneration of the kinocilium is the exception rather than the rule among vertebrate hair cells. Mammalian vestibular hair cells retain their kinocilia, and likewise, the auditory, vestibular, and lateral line hair cells in fish and frogs all feature prominent kinocilia that contribute to the mechanosensory function of the hair bundle.

Together, the hair bundle is a morphologically polarized collection of terraced stereocilia with a kinocilium neighboring the tallest row. This structural polarization plays a central role in hair cell function. The stiff actin core of each stereocilium is inserted into the apical cell surface by a more compliant actin‐based structure called the rootlet (Pacentine et al., 2020). In response to mechanical stimuli, the hair bundle pivots as a cohesive unit at the tapered, rootlet insertion‐point of each stereocilium. The tip of each stereocilium is connected to the shaft of the adjacent taller stereocilia by protein bridges known as tip links (see Section 4.1). When the hair bundle pivots toward the kinocilium, tension on the tip links transfer force to the MET channel, thereby allowing the influx of cations into the hair cell. Conversely, deflection of the hair bundle away from the kinocilium will decrease tension on the tip links and close the MET channels. The MET channel is a multi‐subunit complex composed of the pore‐forming transmembrane channel‐like proteins TMC1 and TMC2 (Jia et al., 2020; Pan et al., 2018) together with integral transmembrane subunits TMIE (Cunningham et al., 2020; Zhao et al., 2014) and LHFPL5 (Beurg et al., 2014; Xiong et al., 2012), and auxiliary cytoplasmic subunits CIB2 and CIB3 (Liang et al., 2021). The channel is nonselectively permeable to monovalent cations but selective for Ca²⁺ (Ohmori, 1985). Like photoreceptors, mature hair cells do not fire action potentials. Instead, stimulus characteristics are encoded by graded depolarizations that modulate the rate of glutamate release at the afferent synapse (Matthews & Fuchs, 2010). Interestingly, activity‐dependent Ca²⁺ influx through the MET channel also regulates the morphology of the stereocilia themselves. Genetic or pharmacological inactivation of the MET channel causes dynamic instability at the tips and changes in stereocilial length (Krey et al., 2020; Vélez‐Ortega et al., 2017).

2.3. Cerebrospinal fluid‐contacting neurons

Cerebrospinal fluid‐contacting neurons (CSF‐cNs) line the central canal of the spinal cord in vertebrate species where they monitor CSF composition and flow (Sternberg et al., 2018; Vigh‐Teichmann & Vigh, 1983; Wu et al., 2021). CSF‐cNs are divided into two populations based on location: ventral and dorsolateral, originating from different progenitor domains of neuroepithelial cells (Djenoune et al., 2017). Differences in structure and physiology between populations suggest they may have unique functions. The ventral population is also termed the Kolmer–Agduhr (KA) neurons in Xenopus and zebrafish, after the researchers who first hypothesized the sensory role of CSF‐cNs (Agduhr, 1922; Kolmer, 1921). CSF‐cNs are GABAergic and project ascending ipsilateral axons (Djenoune et al., 2014).

In contrast to photoreceptors with an overgrown static cilium encircled by CPs and hair cells with a much smaller non‐motile kinocilium connected to stereocilia, CSF‐cNs have a single motile cilium surrounded by microvilli. Further, the neurons have a brush or tuft of microvilli rather than any regular arrangement. The CSF‐cN apical extension, which may also be referred to as the “dendritic terminal,” varies by species. Amphibians and teleosts typically have a greater number of thin microvilli than other species studied—up to 20 or more in Xenopus (Alibardi, 1990; Djenoune & Wyart, 2017). Additionally, several mammalian species lack a motile cilium in the apical extension. In Xenopus laevis, KA neurons have a cell body 16 μm in diameter topped by a brush of microvilli that are each 3–4 μm long with an average diameter of 114 nm (Dale et al., 1987). Notably, CSF‐cNs resemble the morphology of chemosensory tuft cells present in various mucosa, including the lining of gut and respiratory tract, but are distinct based on their identity as neurons and the lack of microtubules in their microvillar tuft (Djenoune et al., 2017; Höfer & Drenckhahn, 1996; Nevo et al., 2019).

CSF‐cNs have both chemosensory and mechanosensory roles in the central canal. Changes in pH and osmolarity are likely detected via the ASIC3 (acid‐sensing ion channel) and PKD2L1 (polycystic kidney disease‐2‐like 1) channels (Jalalvand et al., 2016; Orts‐Del'Immagine et al., 2012). PKD2L1, a non‐selective cation channel covering the apical extension, also functions as a mechanoreceptor to detect changes in CSF flow and spinal bending (Sternberg et al., 2018). The zebrafish pkd2l1 mutants have inactive CSF‐cNs and mature fish develop an unnatural spine curvature. In addition, recent work showed that the rostral‐most CSF‐cNs project to motoneurons in the hindbrain, and ablation impairs the escape response in fish (Wu et al., 2021).

The length of the microvilli has been shown to modulate the degree of CSF‐cN activation (Desban et al., 2019). In zebrafish mutants with shortened microvilli, Ca²⁺ flux is significantly reduced, possibly due to less surface area and fewer Pkd2l1 channels.

2.4. Enterocytes

The most abundant cells in the intestinal epithelium are enterocytes, which have distinct apicobasal polarity, simple columnar morphology, and a microvilli‐covered apical surface known as the brush border (Crawley, Mooseker, et al., 2014). Microvilli contribute to the enormous surface area of the digestive tract, housing enzymes that digest macromolecules, transport substrates, and provide immunological defense. The intestinal brush border also plays a role in monitoring the luminal contents. In contrast to long‐lived stereocilia, the brush border is designed to remodel and regenerate. Further, microvilli have been shown to adapt to feeding and fasting conditions, as well as microbial infection (Misch et al., 1980).

Enterocytes have 2000 or more densely packed microvilli covering the entire apical surface of each cell (Pinette et al., 2019). The separation distance between microvilli is only 93.5 nm in mice, a packing efficiency made possible by arrangement in a hexagonal lattice pattern. Intestinal microvilli are connected by extracellular protein links (discussed in Section 4.2), which contribute to their regular positioning and highly uniform morphology characterized by diameters of ~100 nm and lengths of 1–3 μm (all projections have an equal length within a particular region; Crawley, Mooseker et al., 2014; Pinette et al., 2019; Weck et al., 2016).

Brush border epithelial cells are extremely short‐lived. Pluripotent stem cells present in the intestinal crypts continually proliferate, giving rise to new enterocytes that translocate apically along the villi (Flier & Clevers, 2009). Within 3 days, the cells apoptose and are sloughed off. Even within the short lifetime of the cell, brush border microvilli can undergo dynamic disassembly and repair after injury or infection (Delacour et al., 2016). In rats, specific dietary lectins cause acute damage to intestinal microvilli, leading to their vesiculation and length reduction (Weinman et al., 1989). After 20 h, normal brush border morphology is restored by repair processes and not by expedited cell renewal. Weinmen et al. propose that microvilli naturally cycle through damage and repair with the consumption of dietary lectins. Following epithelial damage, enterocytes can rapidly disassemble the brush border to migrate, which likely involves the reallocation of microvillar actin filaments (Ubelmann et al., 2013).

2.5. Epithelial cells of the renal proximal tubule

Similar to enterocytes, epithelial cells of the renal proximal tubule form a dense microvillar brush border with hexagonal packing (Coudrier et al., 1988; Li et al., 2016). Renal microvilli contact filtrate from the blood as it flows through the nephron in the process of forming urine. The brush border provides an extensive surface area for the exchange of solutes and localization of surface enzymes, while also sensing fluid flow (Du et al., 2004; Guo et al., 2000).

Renal microvilli are typically taller than enterocyte microvilli, but have a similar diameter (Coudrier et al., 1988). For example, rabbit renal microvilli are 3–5 μm long and 100 nm in diameter. Renal epithelial cells also have a relatively slow turnover rate and are considered stable cells (Fujigaki, 2012; McCreight & Sulkin, 1959). Nevertheless, regeneration is possible after epithelial damage via proliferation of existing epithelial cells and stem cell populations. Similar to the enterocytes, renal epithelium can rapidly eliminate and rebuild microvilli. For example, microvilli of renal cells exposed to anoxic stress show shortening, reduced density, and evidence of blebbing but can recover within 30 min after the restoration of normal conditions (Chen et al., 1994). Short periods of renal ischemia cause the total disassembly of the brush border, which can be rebuilt within hours (Paddock et al., 1981).

Glomerular filtration rate (GFR) and exchange of materials by the tubular epithelium are regulated by complex neuronal and hormonal inputs (DiBona & Kopp, 1997). However, there is strong evidence for autoregulation of glomerotubular balance by epithelial cells. The reabsorption of water and ions by individual cells can be precisely matched to natural variations in GFR (Guo et al., 2000). The flow of filtrate through the proximal tubule exerts a torque on microvilli and the torque is proportional to the rate of Na⁺ reabsorption into epithelial cells (Du et al., 2004). Based on the current model, the deflection of microvilli leads to increased transport of the Na⁺/H⁺ ion exchanger 3 to the luminal membrane (Wang et al., 2017).

Additionally, Duan et al. (2008) provide evidence for a “junctional buttressing” response to fluid shear stress detected by microvilli. Increased force on microvilli is associated with dramatic redistribution of actin and associated proteins, leading to reinforced cell–cell junctions and improved epithelial resilience. Thus, brush border microvilli may serve as receptors for multiple signaling cascades that alter physiological activity.

2.6. Functional variety of apical microvilli

The basic functions of apical microvilli encompass surface area enlargement, cell compartmentalization, and regulation of ion currents (Lange, 2010). We have so far introduced three types of sensory neurons, which will be discussed in detail below and highlight how sensory cells utilize microvillar protrusions either directly for sensation or in supportive roles. Taste and olfactory receptor neurons are less studied and therefore are not discussed here but are further examples of diversity in apical projections of sensory cells. Taste buds in the tongue contain 50–100 clustered cells, most of which protrude apically into a taste pore (Yang et al., 2019). The sensory taste cells extend a single microvillus into the pore, while supportive cells form multiple microvilli. Mammalian olfactory receptor neurons sense odors through cilia‐localized receptors and lack microvilli, but nonetheless depend on adjacent microvillated support cells; further, other vertebrate species use both ciliated and microvillated olfactory sensory cells (Hansen & Zielinski, 2005; Morrison & Costanzo, 1990, 1992).

While we focus on sensory neurons, microvilli are present throughout the body on a wide variety of cell types, including epithelial and immune cells (Cebecauer, 2021). For each cell type, the microvilli can be adapted to the particular niche through modifications to the width and length of the actin core, and to the complement of actin‐binding proteins, membrane proteins, and surrounding extracellular matrix components. Below, we discuss the current state of knowledge about stereocilia, CPs, CSF‐cN microvilli, and the brush borders, varying the order to begin with the best characterized for each topic. The next section addresses the actin core and actin‐associated proteins found in microvilli.

3.1. Actin filament organization

Actin is the most abundant eukaryotic intracellular protein, found either as globular G‐actin or in polymers of filamentous F‐actin (Dominguez & Holmes, 2011). The three major actin isoforms share 93% or greater sequence identity—the α‐actin isoform is found predominantly in muscle, while β‐ and γ‐isoforms are universally expressed and both contribute to the microvillar cytoskeleton (Bretscher & Weber, 1978; Kashina, 2020; Perrin & Ervasti, 2010). There is evidence for distinct functions, localization, and affinity for actin‐binding proteins between β‐ and γ‐isoforms. As evidence, γ‐actin knockout mice are viable while β‐actin knockouts are embryonic lethal (Belyantseva et al., 2009; Shawlot et al., 1998). Hair cell‐specific conditional mouse knockouts (cko) of β‐ or γ‐actin each form normal hair cell stereocilia that subsequently exhibit distinctive degeneration patterns (Perrin et al., 2010). The double cko mice do not form stereocilia. Analysis of actin isoforms in other microvillar types remains to be done.

The growth of membrane protrusions like microvilli can be driven by the force of actin polymerization (Svitkina, 2018). The length and stability of the protrusions depend on the balance of elongation, retrograde flow of actin through the filament, and disassembly. The barbed ends (plus ends) that favor actin polymerization is always oriented toward the apical tip of the microvillus, while the pointed ends (minus ends) are directed toward the cell body (Mooseker & Tilney, 1975). Interestingly, a recent study suggested that initiation of a membrane protrusion may require the temporary severing of attachment between cortical actin and the plasma membrane (Welf et al., 2020).

Although a core of actin filaments underlies all microvilli, the number of filaments, dynamics of filament turnover, and complement of actin‐associated proteins are variable and adapted to the environment and physiological role of the particular microvillus. In the ear, the stereocilia of the hair cells contain large bundles of actin filaments; for example, over 3000 single filaments were counted in an alligator lizard stereocilium cross section (Tilney et al., 1980). However, the number of filaments is reduced at the base, with only 18–29 filaments connecting the stereocilium to the cell body. The unique taper creates flexibility and a fulcrum for stereocilia to pivot. In contrast, CSF‐cNs have ~20–25 bundled filaments, renal, and intestinal microvilli have 20–30 bundled filaments, and the number for CPs is unknown, but likely similar.

Parallel actin filaments in stereocilia and intestinal microvilli are arranged with semi‐regular spacing, an order described as “paracrystalline” based on the cross‐sectional appearance (Bartles, 2000; Brown & McKnight, 2010; Metlagel et al., 2019; Tilney et al., 1980). Further, electron tomography‐based reconstruction of rat intestinal microvilli revealed the bundled actin filaments to be twisted 120° in a clockwise manner, a feature dependent on tension generated by interaction with the membrane (Ohta et al., 2012).

Microvillar actin cores extend rootlets into the cell, anchoring the processes, and connecting to underlying networks. Hair cells have a cuticular plate, and enterocytes have a terminal web, each a meshwork of actin, intermediate filaments, and associated proteins. The networks play integral functions in the growth, turnover, and mechanics of the microvilli and we direct the reader to reviews of the structure and components of each (Delacour et al., 2016; Pacentine et al., 2020; Pollock & McDermott, 2015). Notably, the terminal web is linked to the adherens and tight junctions of the enterocyte, whereas the cuticular plate appears to be unconnected to the circumferential belt of cytoskeleton associated with hair cell junctions. In photoreceptors, no meshwork of fibers analogous to the terminal web or cuticular plate has been identified, which is unsurprising given that the processes cover a small portion of the apical cell surface. However, CPs do extend rootlets into the inner segment to the level of the cell–cell junctions at the OLM (Höfer & Drenckhahn, 1993). Further, the recent expansion microscopy work showed a vertical array of microtubules lining the cone inner segment membrane and closely associated with the actin filaments, suggesting that the CPs are well anchored (Verschueren et al., 2022).

Of note, myosins contribute to most aspects of microvilli biogenesis, including transport, anchoring, length regulation, and membrane tension, and their functions are reviewed elsewhere in detail (Houdusse & Titus, 2021). Specific myosins will be mentioned in the relevant sections below.

3.2. Actin dynamics

Two distinct patterns of microvillar actin dynamics have been uncovered through detailed analysis of stereocilia and brush border microvilli.

The actin of intestinal and renal brush border microvilli exhibits so‐called “treadmill” dynamics, similar to filopodia; microvillar cores are maintained in a steady state by actin addition at the barbed end and disassembly at the pointed end in the rootlet, resulting in a continuous turnover of the actin filaments without disruption to microvillar morphology. In a kidney epithelial cell line (CL4), live imaging demonstrated that photobleached GFP‐actin translocates down to the rootlet as new GFP‐actin is added at the tip (Loomis et al., 2003). A rapid rate of 1.5 monomers replaced per second (0.24 μm of actin filament replaced per minute) was measured, although this is still over two times slower than in filopodia. Other photobleaching experiments confirmed treadmill dynamics but observed a rate of only ~0.3 monomers per second in CL4 cells (Tyska & Mooseker, 2002).

In hair cells, early models based on static images of cultured tissue concluded that stereocilia have complete actin turnover within 1–2 days, mediated through similar treadmill dynamics (Rzadzinska et al., 2004). However, live cell imaging of EGFP‐actin in utricle hair cells revealed dynamic elongation and disassembly occurring only at the tip (Drummond et al., 2015). In support, newly synthesized 15 N‐labeled actin was found to be added only to the most apical 0.3–0.5 μm of stereocilia, as assessed by multi‐isotope imaging mass spectrometry (Zhang et al., 2012). Further, photobleaching of fluorescent‐labeled actin uncovered long‐lasting actin below the tip region: even after 40 weeks, very few newly synthesized actin monomers appear in the shaft (Narayanan et al., 2015). Notably, tip turnover is still significantly slower than polymerization in solution and is stringently regulated by capping proteins such as Twinfilin and CapZ (see Section 3.5; Shin et al., 2013). Also, the stereocilia tip turnover model may not be conserved across all vertebrates. In zebrafish stereocilia, rapid recovery of β‐actin‐mCherry could be seen in the shaft following photobleaching, although replacement was incomplete before stalling (Hwang et al., 2015).

The dynamics of actin turnover within CSF‐cNs and CPs remain unknown. However, one report suggested that CP length adjusts according to light conditions, at least in certain species (Pagh‐Roehl et al., 1992). Photoreceptors in fish, amphibians, and birds undergo retinomotor movements, a retinal adaptation to a change in light intensity. During light exposure, rod inner segments lengthen, cone inner segments shorten, and pigment granules translocate into villous processes of the RPE cells, thus protecting the rod outer segments while making the cone outer segments available for vision (Burnside, 2001). As part of the retinomotor movements, CPs of isolated rods from green sunfish were found to shorten upon light exposure (Pagh‐Roehl et al., 1992). Although such movements may suggest a form of actin dynamics in CPs or their rootlets, the data is lacking. Also, evidence from OHCs suggests microvillar length can be modified without dynamic actin: OHC stereocilia undergo sound‐induced length changes despite having a largely stable actin core (Hakizimana et al., 2012). In summary, CP retinomotor movements have yet to be confirmed in vivo or for cones, and the underlying actin dynamics and adaptive purpose remain unclear.

3.3. Actin bundling proteins

Actin bundling proteins (ABPs) connect adjacent actin fibers, creating crosslinks that stabilize actin bundles and establish the packing arrangement, density, stiffness, and unipolar organization of the actin core in microvilli. Common ABPs include plastins (also known as fimbrins), fascins, espin, and villin (Figure 3).

Individual plastin (PLS) proteins contain two actin‐binding domains, each formed by two calponin homology domains, and crosslink actin fibers with a gap of approximately 12 nm (Volkmann et al., 2001). Mammals have three plastin genes, which encode 68 kDa proteins originally referred to as I‐plastin (plastin 1), L‐plastin (plastin 2), and T‐plastin (plastin 3) based on their distinct expression patterns (Shinomiya, 2012).

Fascin (FSCN) is a 55 kDa protein that binds actin filaments via two ß‐trefoil domains to create a tight interfilament spacing of ~8 nm (Jansen et al., 2011). In vitro experiments suggest actin‐binding by fascin is inhibited by serine phosphorylation (Lin‐Jones & Burnside, 2007).

Espin (ESPN) is a single gene that generates four classes of isoforms ranging from ~25 to ~110 kDa, all capable of bundling actin but with variable domains for additional interactions (Bartles, 2000; Sekerkova, 2004). For example, longer isoforms may include binding sites for phosphatidylinositol 4,5‐bisphosphate (PIP₂) and profilin, both of which function in regulation of the actin cytoskeleton (Sekerková et al., 2006).

Villin (VIL) is a member of the gelsolin superfamily of Ca²⁺‐dependent actin interactors, which also includes CapG (Khurana & George, 2008; Nag et al., 2013). Villin binds actin via a headpiece and through a domain that includes six 15 kDa gelsolin‐like repeats, creating an interfilament distance of ~12.8 nm (Bartles, 2000; Brown & McKnight, 2010). A versatile protein, villin has the ability to bundle, nucleate, cap, and sever actin filaments. The regulation of its various functions is not fully understood but tyrosine phosphorylation and high Ca²⁺ levels have each been suggested to convert villin activity from actin bundling to actin severing (Khurana & George, 2008).

On hair cells, each stereocilium is reinforced by tens of thousands of ABP molecules, particularly plastin, fascin, and espin, together generating an interfilament distance of 10 nm (Krey et al., 2016; McGrath et al., 2017; Shin et al., 2013). Of note, the spacing measurements may depend on the preservation method and an electron cryo‐tomography study revealed wider filament spacing (12.5 ± 2.1 nm) than typically observed (Metlagel et al., 2019). The particular complement of ABPs can determine the packing arrangement of actin bundles, as is suggested by differences between mouse and chick stereocilia: in chick, the filaments are arranged in regular lattice array (Tilney et al., 1983), while in mice the filaments exhibit so‐called liquid packing (DeRosier et al., 1980). The difference may arise from the predominance of fascin 2 as the major actin crosslinker in chick stereocilia, and plastin 1 as the most abundant crosslinker in mouse stereocilia (Krey et al., 2016). Indeed, mouse stereocilia lacking plastin 1 adopt hexagonal packing, and the animals display moderate and progressive hearing loss (Krey et al., 2016; Metlagel et al., 2019; Taylor et al., 2014). Fascin 2 is nevertheless required in mouse stereocilia, as Fscn2 mutants form shorter stereocilia that degenerate over time (Liu et al., 2018). In zebrafish hair cells, fascin 2b was shown to be highly dynamic and to cause stereocilia lengthening when overexpressed (Chou et al., 2011; Hwang et al., 2015).

Espin is the least abundant ABP in stereocilia, and yet the most essential; the deaf jerker mice carry homozygous Espn mutations and develop thin, short stereocilia that quickly degenerate (Sekerková et al., 2011; Zheng et al., 2000). Espin drives stereocilia growth and is also an important stabilizer of sensory microvilli because it is insensitive to Ca²⁺, an important component of the sensory response (Sekerkova, 2004). Notably, the concentration of espin is proportional to the length of the stereocilia across different regions of the cochlea, and espin overexpression leads to stereocilia lengthening (Loomis et al., 2003; Rzadzinska et al., 2005). Although classically considered an ABP, espin was recently shown by stimulated emission depletion (STED) microscopy imaging to primarily localize at the actin‐membrane interface, bringing into question its specific role (Qi et al., 2020; Sekerková et al., 2006). The paralogue espin‐like (ESPNL) is also present in stereocilia, but lacks a second actin‐binding domain and therefore can bind but not bundle actin. ESPNL localizes to tips of shorter stereocilia during development and Espnl mouse mutants show high‐frequency hearing loss and the absence of the shortest row stereocilia in the basal cochlea (Ebrahim et al., 2016).

Class III myosins are implicated in the regulation of sensory microvilli length, at least partly through association with espin (Desban et al., 2019; Lelli et al., 2016; Lin‐Jones et al., 2004; Liu et al., 2016). MYO3A and MYO3B contribute to the maturation of hair bundles and the establishment of graded lengths of stereocilia rows, with their functions partially overlapping (Lelli et al., 2016). However, despite a proposed essential role of MYO3 proteins in transporting cargo to the tips of stereocilia, espin and other interacting partners still localize properly in the absence of both MYO3A and MYO3B, suggesting a more nuanced mechanism of length regulation (Ebrahim et al., 2016; Lelli et al., 2016). We direct the reader to a recent review by Cirilo et al. (2021) for a model of Class III myosin function in stereocilia.

Xin‐repeat domain‐containing protein 2 (XIRP2) is the most recent addition to the family of ABPs in stereocilia (Francis et al., 2015; Scheffer et al., 2015). Localized to the stereocilia of both IHCs and OHCs, although with variation in isoforms, XIRP2 is necessary for the paracrystalline arrangement of the actin core. Mouse knockouts develop normal stereocilia, which subsequently degenerate. While the long isoforms may cross‐link actin filaments through their multiple Xin domains, a short isoform of XIRP2 lacking Xin domains was localized to the actin‐membrane interface.

Actin filaments in brush border microvilli form a dense paracrystalline arrangement with the cross‐linkers plastin, villin and espin (Brown & McKnight, 2010; Crawley, Mooseker, et al., 2014; Delacour et al., 2016). Approximately 19 parallel filaments at the center of the bundle are organized in a hexagonal arrangement, separated by an inter‐filament distance of 12 nm and studded by ABPss located at regular intervals.

Plastin 1 is the most abundant cross‐linking protein of the intestinal epithelium and forms a fundamental component of the paracrystalline matrix (Brown & McKnight, 2010; Volkmann et al., 2001). Surprisingly, mouse Pls1 mutants exhibit normal microvillar structure, suggesting redundancy with villin and espin, or compensation by plastin 3, which is expressed developmentally (Chafel et al., 1995; Grimm‐Günter et al., 2009). The most notable phenotype of the mutants is disruption of the microvillar rootlets, where plastin 1 likely mediates interactions between actin and intermediate filaments. As a result, the brush border in the mutants is fragile and susceptible to insult.

Cross‐linking by villin is orderly but appears less crucial to structural integrity, as microvilli are retained in Vil knockout mice (Matsudaira et al., 1983; Pinson et al., 1998). However, villin plays an essential regulatory role because of its ability to switch to a severing activity, which may account for the high plasticity of the brush border and explain the absence of villin in the longer‐lived sensory microvilli (Ferrary et al., 1999; Flock et al., 1982; Höfer & Drenckhahn, 1993).

Only a short splice variant of espin is expressed in mature brush borders and at much lower abundance in comparison to espin in sensory microvilli or villin and plastin 1 in the brush border (Bartles et al., 1998). Consistent with observations in other systems, espin overexpression in renal proximal tubule CL4 cells reduces the actin depolymerization rate and lengthens microvilli in a concentration‐dependent manner (Loomis et al., 2003; Zheng et al., 2010).

Mouse triple knockouts of espin/villin/plastin‐1 are viable and form normal‐appearing intestinal epithelia (Revenu et al., 2012). Even more surprising, the enterocytes can still produce microvilli, although ones that are substantially shorter, lack rootlets, and contain disorganized actin filaments. The transport of digestive enzymes to the apical tips of microvilli is severely impaired in the triple knockouts, likely accounting for the delayed growth of the young mice. Mass spectrometry analysis failed to detect upregulation of any other ABPs that could be compensating for the lack of espin, villin, and plastin‐1, arguing that ABPs are not essential for brush border formation. Nevertheless, actin bundling is still critical to proper organization and functioning of the intestinal microvilli.

Espin and Myo3b are components of the CSF‐cN microvilli in zebrafish and cells in espn mutants or expressing a dominant negative Myo3b protein display reduced apical extensions (Desban et al., 2019). Unlike in hair cells, Espn localization in CSF‐cN microvilli is dependent on Myo3b. It is unclear whether other ABPs are present in CSF‐cNs, but it is possible that fewer is required given that the microvilli are less orderly than those found in other systems.

Analysis of ABPs in CPs is limited, but so far has identified the presence of plastin 1, fascin 2, and espin, suggesting a rigid actin core similar to stereocilia (Höfer & Drenckhahn, 1993; Lin‐Jones & Burnside, 2007; Sahly et al., 2012; Schietroma et al., 2017; Verschueren et al., 2022). Notably, all three proteins are also associated with the actin bundles that extend through the inner segment before forming the cores of the CPs. Further, myosin 3a is associated with inner segment and CP actin, as shown in striped bass and macaque photoreceptors, and transgenic overexpression of Myo3a in Xenopus rods produces abnormally long and broad CPs (Dosé et al., 2003; Lin‐Jones et al., 2004; Sahly et al., 2012).

3.4. Actin‐membrane connections

Tight alignment between the overlying plasma membrane and the internal cytoskeletal core is essential for microvilli to maintain a defined morphology and exist as individual structures. Electron microscopy studies in the 1970's first identified lateral links connecting actin filaments to the membrane in intestinal microvilli (Matsudaira & Burgess, 1979; Mooseker & Tilney, 1975; Mukherjee & Staehelin, 1971; Sauvanet et al., 2015). Subsequent work identified the primary protein to be MYO1A, a plus‐end directed motor that is arranged in a spiral pattern around the actin bundle (Mooseker & Coleman, 1989).

Most microvillar MYO1A turns over at a rapid pace, much faster than the treadmilling actin core, and yet it mediates slow transport of membrane needed for brush border vesicle release (McConnell & Tyska, 2007; Tyska & Mooseker, 2002). Myo1a null mice form intestinal microvilli that show a clear loss of lateral links between the actin core and the plasma membrane, are generally more disorganized, and display vesiculations (Tyska et al., 2005). Further, calmodulin, the myosin 1 light chain, is severely depleted in the microvilli of Myo1a mutants, thereby reducing the Ca²⁺ buffering in the microvilli and potentially activating the actin‐severing function of villin.

MYO6 is a minus end‐directed motor that functions in protein transport, endocytosis, and tethering of the plasma membrane to the terminal web of enterocytes and the cuticular plate of hair cells. The absence of MYO6 results in irregular brush border microvilli and stereocilia, each marked by elevation of the apical plasma membrane and fusion of adjacent processes (Crawley, Mooseker, et al., 2014; Gotoh et al., 2010; Hegan et al., 2012; Hertzano et al., 2008; Seiler et al., 2004; Self et al., 1999). In the intestine, mice lacking both MYO1A and MYO6 exhibit defects in microvillar protein transport but surprisingly normal brush border morphology, suggesting the proteins execute opposite functions when regulating membrane forces (Hegan et al., 2015).

Ezrin is a member of the ERM (Ezrin, Radixin, Moesin) family of proteins, which are characterized by an N‐terminal FERM domain connected via an alpha‐helical domain to a C‐terminal ERM association domain (C‐ERMAD; Pelaseyed & Bretscher, 2018). The FERM domain binds to the plasma membrane either directly or via transmembrane or membrane‐associated proteins, while the C‐ERMAD domain binds to actin filaments. Regulated by phosphorylation and phosphatidylinositol 4,5‐bisphosphate (PIP₂) binding, ezrin cycles between a folded inactive state and a phosphorylated and open active state. Ezrin was first identified as a component of intestinal microvilli, where it acts as another link between the plasma membrane and the actin cytoskeleton (Berryman et al., 1993; Bretscher, 1983). When mice have ezrin fully knocked out or selectively removed from adult intestines, microvilli are short, thick, and irregular, and sometimes replaced with large membrane blebs (Casaletto et al., 2011; Saotome et al., 2004). However, loss of ezrin also disrupts the terminal web and maturation of the enterocyte apical domain, leaving open the question of ezrin's specific role within the microvilli.

Characterization of membrane‐actin connections in stereocilia has mainly focused on the components of the intermicrovillar adhesion complex, to be discussed in Section 4.1. Ezrin may support the initial formation of the stereocilia, but its expression is not maintained in adulthood (Kitajiri et al., 2004). Another ERM protein, radixin (RDX), is localized around the base of hair cell stereocilia and blocking RDX function results in increased deflection but reduced sensitivity in response to sound (Pataky et al., 2004; Prasad et al., 2020).

BAR domain proteins are a large group that share little sequence similarity, but all contain a BAR (Bin1, Amphiphysin, and Rvs) domain consisting of coiled alpha helices with variable curvature and the ability to bind the plasma membrane through electrostatic interactions, thereby inducing membrane curvature (Carman & Dominguez, 2018; Zhao et al., 2011). I‐BAR proteins have an inverse curvature well adapted for creating membrane protrusions such as microvilli, as well as a WH2 domain that can interact with actin regulatory proteins or bind actin directly, thus linking cytoskeletal assembly to membrane outgrowth.

In mouse intestinal organoids, the I‐BAR protein IRTKS (insulin receptor tyrosine kinase substrate, also known as brain‐specific angiogenesis inhibitor 2‐associated protein‐like 1, BAIAP2L1) localizes to the tips of the brush border microvilli, particularly on the crypt enterocytes (Postema et al., 2018). Live imaging revealed IRTKS tracking along the growing tips, and knockdown impaired brush border formation. IRTKS promotes microvillar elongation through recruiting EPS8 (see Section 3.5) and possibly through a direct interaction with actin.

F‐BAR proteins have an inverse curvature, thereby pulling the membrane inward. PASCIN2 (protein kinase C and casein kinase substrate 2, also known as Syndapin 2) is an F‐BAR protein present in the intermicrovillar region of the enterocyte apical membrane, and knockout mice form shorter microvilli with long rootlets (Grega‐Larson et al., 2015; Postema et al., 2019). The reduced membrane coverage of the microvilli core is likely caused by impaired endocytosis in the PASCIN2 mutants, suggesting an important mechanism for regulating microvillar length.

Stereocilia also use an I‐BAR protein, but in a distinct manner (Carlton et al., 2021). BAIAP2L2 (also known as pinkbar) localizes to the tips of short stereocilia in cochlear hair cells, and both conditional and full knockout mouse mutants show progressive stereocilia degeneration and hearing loss (Carlton et al., 2021; Yan et al., 2021). BAIAP2L2 is transported to the tips in a complex with EPS8 and MYO15A (long isoform) and appears after the stereocilia have already formed (Carlton et al., 2021; Halford et al., 2022). Subsequently, MET channel activity is required to maintain BAIAP2L2 at the tips, possibly through Ca²⁺ influx and/or binding to the channel subunit CIB2 (Halford et al., 2022; Yan et al., 2021). Notably, BAIAP2L2 has a unique almost‐flat membrane‐binding surface, which may better suit it for the wide morphology of stereocilia. Further, baiap2l2b is expressed in the zebrafish ear and lateral line hair cells, suggesting a widely conserved function in stereocilia (Erickson & Nicolson, 2015). On the other hand, vestibular hair cell stereocilia are functionally and morphologically unperturbed in the mouse knockouts despite similar BAIAP2L2 localization (Yan et al., 2022).

Zebrafish CSF‐cNs transiently express baiap genes during early development, consistent with a possible function of generating membrane protrusions for apical microvilli formation (Desban et al., 2019). However, functional studies and analysis of other actin‐membrane interactors remain to be done.

To date, none of the actin‐membrane interactors mentioned above have been identified in vertebrate photoreceptors, but some do function in neighboring cells. Ezrin localizes to the apical extension of the RPE and is notably absent from photoreceptors (Bonilha et al., 1999, 2006; Höfer & Drenckhahn, 1993). Similarly, myosin 6 functions in phagosome transport in the RPE, and myosin 1 is restricted to the microvilli of Müller glial cells (Hewage & Altman, 2018; Höfer & Drenckhahn, 1993).

3.5. Actin capping and severing proteins

Localized to the tips of microvilli are dense protein complexes, components of which regulate microvillar actin dynamics and length through either capping or severing of F‐actin barbed ends. Stereocilia, with their staircase‐like arrangement, are a particularly interesting study, with strict distinction existing between proteins localized to the tips of tall row 1 stereocilia versus those localized to the shorter, transducing stereocilia in rows 2 and 3. Notably, the tip turnover rate of actin is greater in the shorter microvilli, a difference that is dependent on intact mechanotransduction (McGrath et al., 2021).

Proteins of the epidermal growth factor substrate 8 (EPS8) family can regulate actin dynamics directly through capping or bundling, or indirectly through enhancement of Rac GTPase function (Hertzog et al., 2010; Luo et al., 2021). EPS8 is primarily localized to the tips of tall stereocilia, where it is thought to act mainly as an actin bundler. Loss of EPS8 leads to a maintained embryonic state, which includes shortening of stereocilia, near absence of the staircase arrangement, continued presence of additional rows of stereocilia, disrupted functional maturation, and profound congenital hearing loss (Manor et al., 2011; Zampini et al., 2011). Notably, EPS8 acts as part of a complex with MYO15A (short isoform), whirlin (short isoform), GPSM2 (G protein signaling modulator 2), and GNAI3 (G protein subunit alpha I3) in row 1 stereocilia, and loss of any one of these factors produces a similar phenotype (Anderson, 2000; Hadi et al., 2020; Holme et al., 2002; Mauriac et al., 2017; Tadenev et al., 2019; Tarchini et al., 2016). Of interest is how the entire hair bundle is affected in all of the mutants despite the localization of the complex to the tips of row 1, suggesting widespread early developmental functions of the proteins and/or indirect effects on shorter rows.

So far, EPS8 is the only protein with potential actin capping or severing function that has been identified in the tips of row 1 stereocilia. In contrast, the shorter stereocilia show tip enrichment of multiple actin capping/severing proteins. For example, EPS8 is also present at the tips of transducing stereocilia, although at lower levels than in row 1 (Zampini et al., 2011), as is the related protein EPS8L2 (Furness et al., 2013). Loss of EPS8L2 leads to a progressive loss of hearing associated with slow degeneration of stereocilia, a distinct phenotype from the congenital deafness caused by EPS8 mutations (Furness et al., 2013). Combined Eps8/Eps8L2 mouse mutants have highly disorganized stereocilia. Unlike in row 1, EPS8/EPS8L2 can traffic to the tips of the shorter stereocilia without the assistance of MYO15A. Nevertheless, the long isoform of MYO15A is part of the tip complex in rows 2 and 3 and is required for the maintenance of the short stereocilia (Fang et al., 2015; Hadi et al., 2020).

Twinfilin 2a (TWF2A) is an actin capping protein that appears at the tips of smaller stereocilia, especially in row 2, just as their growth is finishing (Peng et al., 2009; Rzadzinska et al., 2009). When TWF2A is ectopically expressed in IHCs of mice, the row 1 stereocilia are shortened. In cells lacking MYO7A, TWF2A localization is lost and row 2/3 stereocilia grow abnormally long.

Also present in transducing stereocilia are alpha and beta proteins of the capping actin Protein of Muscle Z‐line (CAPZ) family (Avenarius et al., 2017). CAPZB2 in mouse is initially present in all stereocilia, both at the tips and in punctae along the shaft, where it may be regulating the addition of growing filaments needed to thicken the bundle. Subsequently, CAPZB2 resolves to the tips of the mature row 2 stereocilia, and tip localization depends on BAIAP2L2 (Avenarius et al., 2017; Yan et al., 2021). Interestingly, much of the CAPZB2 in hair cells is bound to TWF2 instead of forming a typical CAPZA/B heterodimer (Avenarius et al., 2017). When CABZB2 was removed from hair cells in a conditional null mouse, auditory function was severely diminished and stereocilia thinned, shortened, and were eventually lost.

The faster actin turnover and higher number of free barbed F‐actin endings in shorter stereocilia, particularly row 2, may be a result of enrichment of the actin severing protein actin depolymerizing factor (ADF) and its paralogue cofilin (CFL) (McGrath et al., 2021). Notably, ADF/CFL1 localizes only once mechanotransduction begins, and impaired mechanotransduction will disrupt the concentration of both ADF/CFL1 and free actin barbed ends in row 2 tips. The combined loss of ADF and CFL1 in mice leads to severe disruption of stereocilia, whereas single mutant, conditional, and mixed alleles result in a variety of milder phenotypes that variably affect stereocilia in all three rows, perhaps because of both direct effects in shorter stereocilia and indirect effects on row 1 through changes to the pool of available actin monomers (McGrath et al., 2021; Narayanan et al., 2015). In some cases, reduced ADF/CFL1 also results in the persistence of medial microvillar processes that typically undergo pruning, suggesting that ADF/CFL1 aid in their breakdown (McGrath et al., 2021).

Actin inhibitory protein (AIP1), also known as WD‐repeat protein 1 (WDR1), is another actin‐severing protein, but one that preferentially interacts with ADF/CFL‐bound actin filaments and is therefore often considered a cofactor (Ono, 2018). Loss of AIP1 results in some missing and shortened row 2/3 stereocilia, a phenotype very similar to the ADF single mutant (Narayanan et al., 2015).

Finally, gelsolin (GSN) is one of eight members of the gelsolin superfamily of proteins (which also includes villin) with actin capping, severing, and nucleating properties (Feldt et al., 2018). Transient expression of gelsolin was identified in the shafts of short stereocilia of murine OHCs from P0 to P15 (Mburu et al., 2010). Stereocilia initially form normally in Gsn mutant mice, but the OHC stereocilia in the apical region of the cochlea subsequently grow long and straggly.

Less is known about actin capping/severing proteins in other types of microvilli. Eps8 is localized to tips of intestinal microvilli in Caenorhabditis elegans and knockdown leads to irregular microvilli, but selective mutations suggest a predominant role in actin bundling rather than capping (Croce et al., 2004; Hertzog et al., 2010). Eps8 knockout mice have shortened intestinal microvilli, resulting in reduced body weight and altered metabolism (Tocchetti et al., 2010). EPS8 and EPS8L1A are found at tips of microvilli on CL4 cells, derived from porcine renal proximal tubule epithelial cells (Zwaenepoel et al., 2012). Similar to their disparate roles in stereocilia, EPS8 facilitates actin bundling in the renal brush border, whereas EPS8L1A promotes actin capping and limits microvillar elongation (Zwaenepoel et al., 2012). In this case, ezrin was shown to be a necessary partner for EPS8/EPS8L1A functions in microvilli.

Villin was discussed previously as a prominent actin cross‐linking protein in the brush border, but it is also a member of the gelsolin superfamily and can switch to actin capping and severing functions depending on phosphorylation and the concentration of Ca²⁺ (Khurana & George, 2008). The presence of villin in brush border microvilli may therefore preclude the need for the many actin regulatory factors found in stereocilia. Indeed, ADF will only translocate into brush border microvilli to promote breakdown in conditions of stress, such as following ischemic injury (Ashworth et al., 2001; Schwartz et al., 1999).

No localization or functional data is available for actin capping/severing proteins in the microvilli of CSF‐cNs or photoreceptors.

4.1. Stereocilia protein links

The structural and functional integrity of the hair bundle is maintained by an heterogenous array of protein–protein linkages between adjacent stereocilia and between stereocilia and the kinocilium (Richardson & Petit, 2019). The intermicrovillar linkages are categorized according to the proteins involved and by their location within the hair bundle. Cadherin 23 (CDH23) and protocadherin 15 (PCDH15) participate in various types of bundle linkages: developmentally transient lateral links (Ahmed et al., 2006; Michel et al., 2005), kinocilial links between stereocilia and the kinocilium (Goodyear et al., 2010; Lagziel et al., 2005; Michel et al., 2005), and tip links (see below). As bundles mature, the linkages between stereocilia become more refined. Ankle links are formed at the base of the stereocilia by Adhesion G‐protein coupled receptor V1 (ADGRV1/GPR98/VLGR1/USH2C) and Usherin (USH2A) (Adato et al., 2005; McGee et al., 2006). In mature bundles, shaft and top connectors (PTPRQ, STRC) adjoin neighboring stereocilia along their lengths (Goodyear et al., 2003; Verpy et al., 2010). While these connections all maintain hair bundle integrity and function, the tip links have received the most attention because of their direct role in MET.

Tip links connect the tip of a transducing stereocilium to the shaft of a neighboring stereocilium in the adjacent taller row. As measured in guinea pig, links are 5 nm thick and 150–300 nm in length (Furness & Hackney, 1985). CDH23 forms the upper portion of the tip link, located on the shaft of the taller partner and connected to the actin core by the scaffolding protein harmonin and the motor protein MYO7A (Grati & Kachar, 2011; Grillet et al., 2009; Schietroma et al., 2017). PCDH15 forms the lower end of the tip link at the distal tips of the shorter adjacent stereocilia where it directly binds to the TMC1 and TMC2 pore‐forming subunits of the MET channel (Kazmierczak et al., 2007; Maeda et al., 2014). Sans is another scaffolding protein important for MET, but of unclear localization, having been identified at the location of lower tip links in one study (Caberlotto et al., 2011) and upper tip links in another study (Grati & Kachar, 2011). Sans interacts with MYO7A and harmonin in vitro and localization of both proteins is disrupted in sans mouse mutants (Ush1g−/−), arguing that sans is at least developmentally necessary in the upper tip link. However, sans localization is disrupted in both Cdh23 and Pcdh15 mutants, and each adhesion protein is capable of localizing sans to the plasma membrane in COS‐7 cells, keeping open the possibility of a lower tip link role.

The N‐termini of CDH23 and PCDH15 interact on the external face of the stereocilia in a Ca²⁺‐dependent manner (Sotomayor et al., 2012). When hair cells are deflected toward the tallest stereocilia, force is conveyed through the tip links and opens the MET channels (Pepermans & Petit, 2015). The tip link densities anchoring the atypical cadherins to the cytoskeleton are therefore essential for hair cell signaling.

One of the best‐studied hereditary stereocilia‐related diseases is Usher syndrome, a congenital condition leading to a combined loss of hearing and vision (El‐Amraoui & Petit, 2005). Usher syndrome is clinically and genetically heterogeneous and subdivided into types 1, 2, and 3, based on the phenotype of the patients and the course of the disease (Jouret et al., 2019). Many of the Usher genes code for components of the interstereocilia links. Usher type 1 (USH1) proteins (CDH23, PCDH15, MYO7A, harmonin/USH1C, sans/USH1G) interact to form tip link complexes and are therefore required for mechanotransduction. USH2 proteins (ADGRV1, USH2A, WHRLN), together with USH1 proteins, participate in the extensive network of lateral and ankle links between adjacent stereocilia.

Mice mutants for each of the five USH1 proteins have disorganized stereocilia at misaligned angles and often a mispositioned kinocilium (Lefevre et al., 2008). In Myo7a mouse mutants, stereocilia are abnormally tall. In mutants for other USH1 genes, the shorter rows are reduced in length while the tallest row remains intact. The early morphological phenotypes suggest that the links are required during development for hair bundle cohesion and regulation of stereocilia length.

4.2. Brush border intermicrovillar links

Microvilli in the intestinal brush border are joined by tip‐links that structurally resemble those seen in stereocilia, but differ in protein composition. The one common component is harmonin (USH1C). Indeed, in a case study of individuals with a severe deletion in the USH1C gene, patients suffered from renal and intestinal dysfunction, as well as expected sensorineural deafness (Bitner‐Glindzicz et al., 2000).

Brush border extracellular links are 50 nm long, as measured in cultured CACO‐2 (colorectal adenocarcinoma‐2) cells, and are restricted to the apical 20% of the microvilli (Crawley, Shifrin, et al., 2014). The links are composed of CDHR5 (cadherin‐related family member 5, also referred to as mucin‐like protocadherin) and CDHR2 (cadherin‐related family member 2, also referred to as PCDH24), bound in a Ca²⁺‐dependent manner (Crawley et al., 2016; Li et al., 2016). Intracellularly, CDHR2 interacts with harmonin, which complexes with ANKS4B (ankyrin repeat and sterile α motif domain‐containing 4B) and MYO7B. ANKS4B shares 40% sequence similarity with sans (USH1G), but the two proteins are not interchangeable in binding affinity experiments. ANKS4B is essential for the formation of the tip link complex and for the clustering of brush border microvilli. MYO7B, related to the MYO7A protein in stereocilia, anchors the complex to the actin core and likely transports the components to the microvillar tips (Weck et al., 2016, 2020). Calmodulin‐like protein 4 (CALML4) acts as the myosin light chain for MYO7B in the brush border (Choi et al., 2020).

A hybrid protein consisting of the MYO10 motor domain fused to the MYO7B cargo‐binding tail was designed to test MYO7B function using HeLa cell filopodia as a model (Choi et al., 2020). As expected, the MYO10‐MYO7B chimera could transport USH1C, ANKS4B, and CDHR2 to the tips of filopodia. Surprisingly, the MYO10‐MYO7B chimera could also transport CDHR2 in the absence of USH1C and ANKS4B, suggesting potential for direct interaction between MYO7B and CDHR2. Notably, the CDHR5 end of the protein link has not yet been characterized, but the MYO10‐MYO7B chimera could also transport CDHR5 to filopodia tips in an USH1C‐independent manner.

Knockdown of any component of the intermicrovillar adhesion complex results in a significant loss of organized clustering in the brush border (Crawley et al., 2016; Crawley, Mooseker, et al., 2014; Weck et al., 2016). Microvilli become irregularly spaced compared to the hexagonal packing in wild‐type cells. Furthermore, microvilli lacking CDHR2 are variable in length and shorter on average. When the MYO10‐MYO7B chimera was expressed in HeLa cells, filopodia not only developed tip links but also exhibited height matching (Weck et al., 2020). Intermicrovillar links are therefore required during brush border formation to create correct hexagonal packing and regulate the length of the microvilli.

In the mature intestinal brush border, intermicrovillar links may be required to stabilize microvilli against mechanical damage caused by peristaltic contraction and abrasion. Additionally, the links support dense packing and uniform length that may help impede microorganisms from colonizing the space near the apical cell body (Crawley, Mooseker, et al., 2014).

A recent study using cryo‐electron tomography identified “nanobristles” extending laterally from the shafts of intestinal microvilli in C. elegans (Zhu et al., 2022). Although nanobristles from adjacent cells are not in contact, they still function in intermicrovillar spacing. In the absence of the protocadherin CDH8, the nanobristles fail to form and the microvilli exhibit denser packing.

Intermicrovillar links in the renal brush border are poorly characterized, but likely similar to those of the intestine. Supporting evidence thus far is the expression of CDHR5 and USH1C in the brush border of the human proximal convoluted tubule, and renal dysfunction in three patients with large mutations overlapping the USH1C locus (Bitner‐Glindzicz et al., 2000; Bläsius et al., 2017; Scanlan et al., 1999). Presumably, the links would be required to establish consistent height and orderly packing of the renal microvilli, as in the intestine, but could also support the proposed mechanosensory function of the renal brush border (Wang et al., 2017).

4.3. Extracellular links of CPs

Extracellular links are proposed to connect CPs to the outer segment and to one another (Miles et al., 2021; Sahly et al., 2012; Schietroma et al., 2017). Spacing between CPs is similar to stereocilia, suggesting‐related link components. In support, PCDH15 and CDH23 have been localized to the CPs and OSs of macaque, human, and Xenopus rods and cones, and the links are sensitive to chelation of Ca²⁺, a necessary component of PCDH15‐CDH23 binding (Sahly et al., 2012; Schietroma et al., 2017; Sotomayor et al., 2012). Knockdown of Pcdh15 in Xenopus laevis results in overgrown rod disks, abnormally tapered cones, reduced actin staining, and fewer and fragmented CPs (Schietroma et al., 2017). Electroretinogram amplitudes, a measure of the neural response to retinal light detection, are reduced in knockdown Xenopus tadpoles.

Zebrafish have two PCDH15 orthologues, pcdh15a expressed in the ear and pcdh15b expressed in the eye (Seiler et al., 2005). Fish carrying homozygous pcdh15b mutations display shortened, disorganized, and missing OSs with detached CPs (Miles et al., 2021). The binding partner for Pcdh15b remains uncertain though as cdh23 expression in the zebrafish retina is limited to a subset of amacrine cells (Glover et al., 2012).

CPs are proposed to provide structural support for the OS and restrain the growth of new disks, a view supported by the Pcdh15 knockdown and mutant phenotypes; however, Pcdh15 is not localized exclusively to the CPs but is also present at the inner segment‐outer segment interface, and zebrafish photoreceptors lacking Pcdh15b show detachment of the two cellular regions (Miles et al., 2021; Sahly et al., 2012; Schietroma et al., 2017). Further, mice, which lack CPs, also display visual deficits when homozygous for Pcdh15 mutations (Sethna et al., 2021).

Functions of the remaining USH1 proteins, including harmonin, sans, and MYO7A, are even less clear in photoreceptors. While evidence exists for an association with CPs, expression and localization of the three proteins are inconsistent across species, and the formation of a retinal USH1 complex has not yet been demonstrated (Hasson et al., 1995; Hodel et al., 2014; Sahly et al., 2012; Schietroma et al., 2017; Sorusch et al., 2017, 2019). At the least, it remains possible that these USH1 proteins have additional functions in other retinal cell types or other regions of the photoreceptors, any of which could contribute to patient visual deficits.

Evidence discussed so far suggests that CPs are a stable form of microvilli. However, CPs are linked to an OS undergoing constant renewal, which would suggest that either the CPs need to grow and shed fragments at the same rate as the OS, or the protein links to the OS undergo regular remodeling.