The architecture of the cell surface is crucial for metazoan development and physiological homeostasis. Microvilli are membrane-bound cell surface protrusions that contain a core bundle of actin filaments enveloped in the plasma membrane 1-3. Many epithelial cells develop this organelle above their apical surface to enhance functional capacity for a range of physiological tasks, including nutrient absorption in the intestine 4, solute uptake in the renal tubules 5, mechanosensation in sensory stereocilia of the inner ear 6, and chemosensation in the gut, lung, and urogenital tracts 7-9. Abnormal microvillar structure and function lead to human disorders such as life-threatening nutrient malabsorption and osmatic imbalances and inherited deafness in Usher syndrome 1,3,4,10.
An intestinal enterocyte develops up to one thousand densely packed microvilli in an array known as the "brush border" (BB). These fingerlike outward projections enhance the functional surface area for nutrient absorption and provide the barrier for host defense against pathogens and toxins 1,3,4. Despite their importance, we know little about how microvilli are formed with striking uniformity in sizes and how these protrusions are maximally packed in a hexagonal pattern. Because the gut epithelium undergoes the constant regenerative renewal, microvillus assembly is a process that continues throughout our lifetime 1,3,4,11. However, an individual microvillus is only 0.1 μm in diameter and ~2 μm in height (Fig. 1a), the tiny size of which, along with their high density and lumen localization, becomes a technical hurdle for in situ structural investigations at high resolution 1-3.
Cryo-ET of C. elegans larvae
Recent methodology advance of cryo-electron tomography (cryo-ET) makes the platform well suited to address the challenges of studying microvillus structure in animals 12,13. As a proof-of-concept, we initiated the cryo-ET study of intestinal microvilli in the nematode Caenorhabditis elegans. Classical transmission electron microscopy (TEM) documented the anatomy and molecular organization of the C. elegans intestine and microvilli, which highly resemble those in mammals, supporting the notion that microvilli are evolutionarily conserved organelles (Fig. 1a, Extended Data Fig. 1a) 14,15. Studies of microvilli in other species require tissue dissection or in vitro cell culture systems. In contrast, the thickness of the C. elegans L1 larvae, which is approximately 12.5 μm in diameter (Extended Data Fig. 1a) 14,15, allows high-pressure freezing (HPF) of an entire animal, thereby enabling structural analysis of microvilli in their close-to-native conditions.
We employed cryo-focused ion beam milling (cryo-FIB) to produce thin lamellas from the frozen-hydrated vitrified worms 12. By taking advantage of the invariance of the C. elegans gut development and anatomy 14,15, we focused on the intestine from the newly hatched first-stage juvenile (L1 larva, 6 hours after hatching). The larval intestine runs approximately 70% of body length between the anterior pharynx and the posterior hindgut 14,15. We thinned the intestine's central region, approximately 150 μm posterior from the animal's nose, corresponding to the intestinal cells from Int4 to Int7 (Fig. 1b, Extended Data Fig. 1a and Supplementary Video 1). Cryo-FIB milling removed the biological material above and below the area, generating a 150- to 200-nm-thick cellular lamella traversing the whole animal with high consistency and reproductivity (Fig. 1b). Next, we applied cryo-ET to image the cryo-lamella, which facilitates the three-dimensional (3D) visualization of microvilli at unprecedented molecular resolution. Each lamella provided a suitable imaging area of ~10×10 µm2, from which we obtained 2~3 cryo-ET tilt series. In agreement with the early TEM observations 14,15, our tomograms of the C. elegans microvilli revealed that the plasma membrane surrounds the actin bundles in the core (Fig. 1c, Supplementary Video 1). The plasma membrane appears as two smooth and parallel black lines, and actin bundles from the cross-section were resolved as individual black dots (Fig. 1d, Supplementary Video 1). The excellent contrast and rich structural details from the reconstructed tomogram demonstrate the high quality of prepared samples (Fig. 1d, Supplementary Video 1).
Discovery of nanovilli from C. elegans intestine
Strikingly, we found hundreds of stick-like structures projecting out from each microvillus's lateral surface (Fig. 1c,f and Supplementary Video 2). These structures are highly uniform, with a constant diameter of 4.5 ± 0.5 nm (n = 4, 983 mean ± SD). They appear to be the transmembrane structure whose extracellular length is 37.5 ± 0.2 nm (mean ± SD). Each tomographic slice contains 50 ± 5 (mean ± SD, from 12 focal planes) of the stick-like structures (Fig. 1d, e), highlighting their abundance. These structures are distinct from the glycoprotein-rich glycocalyx, a region between the apical tip of microvilli and the luminal space (Fig. 1a), which provides a barrier for pathogens and serves as the interface for nutrient digestion 3,4. These sticks are also different from protocadherin-based adhesion tip links between adjacent microvilli, which only localize at microvillar tips (Fig. 1a) 16. By contrast, the stick-like structures distribute along the lateral surface of microvilli (Fig. 1c, f). Microvilli are considered tiny micrometer appendages that decorate the epithelium surface, and our results reveal that these previously unrecognized yet highly abundant nanometer attachments further decorate microvilli's lateral surface. Therefore, we have named a single stick-like structure nanovillus because of its size and location relative to microvillus.
We wondered why the previous TEM studies did not describe nanovillus. The TEM sample preparation involves chemical fixation, dehydration, and heavy metal staining, which are not compatible with high-resolution structural studies. Although TEM cannot resolve an individual nanovillus, this technique may detect an electron-dense layer that contains numerous nanovilli at the lateral surface of microvilli. To test this idea, we performed TEM of HPF, freeze-substituted C. elegans. We determined an excellent TEM quality microvillus if the plasma membrane’s lipid bilayer is resolved as two parallel black lines. From such microvilli, we indeed found an electron-dense area along microvilli’s lateral surface from either longitudinal or cross-sections (Fig. 2a). The thickness of these layers is 37.3 ± 0.5 nm (51 nanovilli layers from 3 animals, mean ± SD) (Fig. 2a), comparable to that of cryo-ET tomograms. We noticed a thin band that does not have electron density localized between two adjacent nanovilli layers (Fig. 2a). Such electron transparent bands indicate the separation of nanovilli layers and suggest that nanovilli might be involved in intermicrovillar spacing.
Nanovilli in mouse intestine
We asked whether mammals develop a similar nanovilli structure on microvilli. First, we screened the published TEM images of microvilli. Most of the sample preparation relied upon chemical fixation, which did not yield high-quality pictures on the microvillar membrane. Considering that membrane preservation is probably a prerequisite to visualize the fine structure on its surface, we performed HPF and freeze-substituted mouse intestinal tissues, allowing the visualization of a lipid bilayer in the microvillar membrane as two black lines (Fig. 2b). We detected a similar electron-dense layer on the lateral surface of mouse microvilli in the longitudinal or cross-sections (Fig. 2b). Next, we employed the same cryo-FIB milling and cryo-ET platform that detects C. elegans nanovilli to examine mouse intestinal tissues. In agreement with our TEM results, 13.8 ± 0.8 nm long stick-like structures decorate the lateral surface of mouse microvilli (64 mouse nanovilli from 4 tomograms, mean ± SD) (Fig. 2c). Despite the difference in length, these stick-like structures' morphology and location suggest that nanovilli, like microvilli, might be widespread from nematodes to mammals.
Nanovillus formation requires CDH-8
We sought to identify the molecular composition of nanovilli. We performed single-cell mRNA sequencing of C. elegans larvae to gain insights into the intestine transcriptome (Extended Data Fig. 1b, c, Table s1, details in Materials and Methods). A previous study documented the proteome of the mouse brush border 17. Among the overlapped hits between the C. elegans intestinal transcriptome and mouse brush border proteome, the intriguing candidates are the protocadherin family proteins (Extended Data Fig. 1d, and 2a). The transmembrane protocadherin family proteins mediate cell-cell adhesion, such as PCDH24 and PCDH15 in the intermicrovillar link at microvillar tips 3,16,18,19. however, an electron transparent band between two adjacent microvilli indicates that nanovilli layers are not linked but well separated (Fig. 2a). Considering that the mouse protocadherin 8 protein plays a repulsive role in mediating the dendritic self-avoidance during neural development 20,21, we asked whether protocadherin might be a component of nanovilli.
We obtained the protocadherin cdh-3, cdh-8, and cdh-10 mutant alleles defective in C. elegans protocadherin genes to probe this possibility. Because TEM offers the throughput to screen for the nanovilli layer and electron-transparent intermicrovillar band, we used this method to assess nanovilli in multiple mutants quickly. The nanovilli and transparent bands in cdh-3 and cdh-10 deletion mutants are indistinguishable from those in wild-type (WT) animals (Fig. 2d, > 200 examined microvilli from three animals of each genotype). By contrast, we did not detect the nanovilli layer or the electron transparent band in cdh-8(ok628) mutant animals (530 examined microvilli from five animals), despite excellent preservation of the microvillar membrane (Fig. 2d). Our cryo-ET tomograms further confirmed the loss of nanovilli in cdh-8(ok628) animals. We occasionally found stick-like structures between microvilli in cdh-8(ok628) mutants (Fig. 3c, yellow arrows), but their more prolonged length (56.8 ± 5.4 nm, n =50, mean ± SD, versus 37.5-nm-long nanovilli) indicates that they are distinct structures from nanovilli.
Protocadherin genes are well-known for their alternative splicing 21, and the cdh-8(ok628) allele has an in-frame deletion that removes 119 residues in its extracellular domain (Fig. 2e). To rule out the effects of a large variety of protocadherin isoforms, we implemented the CRISPR-Cas9-based homologous recombination strategy to precisely remove the transmembrane domain along with the entire intracellular domain of CDH-8, which are shared by all the CDH-8 isoforms, generating a cdh-8(cas1109) mutant allele (Fig. 2e). Our TEM datasets uncovered that the nanovilli layer and the electron transparent band disappeared in cdh-8(cas1109) mutants (Fig. 2d, 365 examined microvilli from 4 animals). The cryo-ET pipeline also revealed the loss of nanovilli in cdh-8(cas1109) animals (Fig. 3a-f and Supplementary Video 3-5). Our results from two mutant strains show that CDH-8 is essential for nanovilli formation.
To define the molecular architecture of nanovilli, we performed unbiased subtomogram averaging of nanovilli from cryo-ET datasets. A total of 4983 nanovilli were hand-picked from 11 tomograms, aligned, and averaged to produce an average structure (Fig. 3g, h). While the resolution did not allow us to speculate the molecular composition of nanovilli, the density map shows that a single nanovillus compromises nine repeats in the extracellular domain, the transmembrane region, and the intracellular domain (Fig. 3g, h). Consistently, CDH-8 is predicted to consist of nine extracellular cadherin (EC) domains, a single transmembrane (TM) helix, and an intracellular domain (Fig. 3g, h). X-ray crystallography characterized the atomic resolution structures of EC domains 22. Using its characteristic curvature, we manually fitted nine copies of the human protocadherin-15's EC8 domain (PDB code 4XHZ) into the nanovillus extracellular region of our sub-tomogram averaging structure (Fig. 3h and Supplementary Video 6), further supporting the molecular connection between CDH-8 and nanovilli.
Consistent with the structural results, our GFP-based transcriptional reporter indicated that the cdh-8 gene was expressed in the C. elegans intestinal cells (Extended Data Fig. 2b), validating our single-cell sequencing data. We expressed the GFP-tagged CDH-8 full-length protein in the C. elegans intestinal cells. By constructing a red fluorescence protein mScarlet knock-in animal to label an established microvillus marker ERM-1 (the Ezrin/Radixin/Moesin family protein) 2,3, we showed the colocalization of the green CDH-8::GFP and the red ERM-1::mScarlet fluorescence in the intestinal lumen (Extended Data Fig. 2c), providing evidence that CDH-8 is the component of nanovilli.
Loss of Nanovilli slows down animal growth
Next, we used the cdh-8 mutant animals to dissect the function of nanovilli. We did not notice any apparent defects in the overall animal morphology or the number of progenies; however, our quantifications of the body length at each developmental stage revealed a reduction of animal growth in cdh-8 mutant animals (Extended Data Fig. 3a). Because the eggshell prevents embryos from accessing the bacterial food, our measurements started from the newly hatched larvae, when the worm body lengths are indistinguishable between cdh-8 mutants and WT. With the progression of animal growth in the presence of the same food, the body lengths of cdh-8 mutant alleles were shorter than that of WT animals at the various larval stages (Extended Data Fig. 3a). At 122 hours after hatching, when C. elegans already developed into the day two adults, cdh-8 deletion animals reached the WT body length (Extended Data Fig. 3a). Notably, the GFP-tagged cdh-8 gene expression using an intestinal-specific promoter Pges-1 fully rescued the slow growth phenotype in both cdh-8 mutant alleles (Extended Data Fig. 3b), demonstrating a cell-autonomous function of cdh-8 in promoting animal growth. These data indicate that nanovilli function from the C. elegans intestine to regulate animal growth. We suggest that the growth delay might result from the less efficient food update and that the slow but continuous growth eventually restored the cdh-8 mutant body length to the WT level.
Y-shaped microvilli and the role of nanovilli in spacing microvilli
To address if nanovilli contribute to microvilli assembly and organization, we compared microvilli structure in cdh-8 mutant and WT animals. Our TEM analyses detected the forked or branched (collectively called Y-shaped) microvilli in cdh-8 mutant animals with 4% (ok628, 530 microvilli from 5 animals) or 16% (cas1109, 351 microvilli from 4 animals) penetrance (Fig. 4a). The Y-shaped microvilli were rarely detectable in WT animals (Fig. 4a, 468 microvilli from 5 animals); neither did our literature search find Y-shaped microvilli under the standard conditions. Curiously, half a century ago, the Y-shaped microvilli were described during the salamander microvilli regeneration (Extended Data Fig. 2d) 23. Using hydrostatic pressure to eliminate microvilli on the small intestine, Tilney and Cardell observed the Y-shaped regenerating microvilli four minutes after pressure release but not later on, suggesting that Y-shaped microvilli might represent an intermediate structure during microvilli regeneration.
In a developing WT C. elegans larva, one of the cryo-ET tomograms showed a Y-shaped microvillus with two separated tips but a single base (Fig. 4b, c, and Supplementary Video 7). Consistent with our TEM results, we only detected one Y-shaped microvillus from more than 100 examined ones, which indicates the occurrence of such a structure during animal development and suggests that a Y-shaped microvillus might be transient. In cdh-8 mutant animals, the Y-shaped microvilli may be an arrested status of a splitting microvillus; and the lack of nanovilli failed to space two daughter microvilli efficiently, thereby enriching Y-shaped microvilli (Fig. 4a). Unlike the straight nanovilli on the well-spaced microvilli, nanovilli on the newly split microvilli were bent (red sticks in Fig. 4c lower), implying that nanovilli straightening might push two nascent microvilli apart. In support of this idea, loss of nanovilli markedly reduced the distance between two adjacent microvilli from 88.1 ± 15.9 nm in WT (Mean ± SD, n = 44) to 61.3 ± 7.8 nm in cdh-8(ok628, n = 32) or 51.0 ± 9.1 nm in cdh-8 (cas1109, n = 49) mutant animals (Fig. 3c, e, and Extended Data Fig. 2e), further demonstrating a role of nanovilli in spacing microvilli. Thus, we propose a microvilli division model through which nanovilli regulate microvilli formation: an existing microvillus divides from the tip to the base, generating two nascent microvilli, and nanovilli insert into the newly split microvilli to push them apart (Fig. 4d).
Live imaging of microvilli division
To test the microvilli division model directly, we developed a high-resolution live-cell imaging method to follow the dynamic microvilli assembly process in C. elegans. Fluorescence live-cell imaging recorded microvilli dynamics in cell cultures 24,25; however, time-lapse visualization of microvilli in a living organism is challenging. The 88 nm intermicrovillar distance does not allow conventional confocal microscopy to resolve adjacent microvilli. Neither could the super-resolution imaging platform be applied because their limited working space does not determine the intestinal lumen structures, which is at least five μm inside the C. elegans cuticle. Using the Airyscan technology in a GFP::ERM-1 KI animal, we successfully resolved individual microvilli along the intestinal lumen in developing C. elegans larvae (Fig. 4e). Line scan from the microvillar top or bottom or the whole microvilli revealed separate fluorescence peaks, and the average distance between the peaks of adjacent microvilli is 201.0 ± 9.1 nm (Mean ± SD, n = 44, Fig. 4e), which is comparable to the measurements from the cryo-ET tomogram slice (205.8 ± 28.0 nm, Mean ± SD, n = 49). Time-lapse recording started from a Y-shaped microvillus (Fig. 4f, yellow dotted box), and its two tips gradually moved apart, forming a V-shaped microvillus (0.8 and 1.6 min in Fig. 4f). Eventually, two nascent microvilli were separated (Fig. 4f). A stick-like microvillus underwent a similar process to assemble two new microvilli, and the Y- or V-shaped microvilli were intermediate structures when a mother microvillus split into two daughter microvilli (Fig. 4g and more examples in Extended Data Fig. 4). Our live imaging data support the microvilli division model in which two nascent microvilli are generated by dividing an old microvillus. In this model, the daughter microvilli formation uses the mother microvilli as their templates, forming two daughter microvilli of identical sizes, reminiscent of chromosome or centriole duplication in a dividing cell. In support of microvilli duplication, the Y-shaped microvilli base contains more actin bundles and is larger than the regular stick-like microvilli (Fig. 4b, c).