Brush cells in the human duodenojejunal junction: an ultrastructural study


  • Manrico Morroni,

    1. Institute of Normal Human Morphology, School of Medicine, Polytechnic University of Marche, Ancona, Italy, and Electron Microscopy Unit, University Hospital, Ancona, Italy
    Search for more papers by this author
  • Angela Maria Cangiotti,

    1. Institute of Normal Human Morphology, School of Medicine, Polytechnic University of Marche, Ancona, Italy, and Electron Microscopy Unit, University Hospital, Ancona, Italy
    Search for more papers by this author
  • Saverio Cinti

    1. Institute of Normal Human Morphology, School of Medicine, Polytechnic University of Marche, Ancona, Italy, and Electron Microscopy Unit, University Hospital, Ancona, Italy
    Search for more papers by this author

Manrico Morroni MD, Istituto di Morfologia Umana Normale, Facoltà di Medicina e Chirurgia, Università Politecnica delle Marche, Via Tronto 10/A, 60020 Ancona, Italy. T: +39 071 2206085; F: +39 071 2206087; E:


Brush cells have been identified in the respiratory and gastrointestinal tract mucosa of many mammalian species. In humans they are found in the respiratory tract and the gastrointestinal apparatus, in both the stomach and the gallbladder. The function of brush cells is unknown, and most morphological data have been obtained in rodents. To extend our knowledge of human brush cells, we performed an ultrastructural investigation of human small intestine brush cells. Six brush cells identified in five out of more than 300 small intestine biopsies performed for gastrointestinal tract disorders were examined by transmission electron microscopy. Five brush cells were located on the surface epithelium and one in a crypt. The five surface brush cells were characterized by a narrow apical pole from which emerged microvilli that were longer and thicker than those of enterocytes. The filamentous core extended far into the cell body without forming the terminal web. Caveolae were abundant. Filaments were in the form of microfilaments and intermediate filaments. Cytoplasmic projections containing filaments were found on the basolateral surface of brush cells. In a single cell, axons containing vesicles and dense core granules were in close contact both with the basal and the lateral surface of the cell. The crypt brush cell appeared less mature. We concluded that human small intestine brush cells share a similar ultrastructural biology with those of other mammals. They are polarized and well-differentiated cells endowed with a distinctive cytoskeleton. The observation of nerve fibres closely associated with brush cells, never previously described in humans, lends support to the hypothesis of a receptor role for these cells.


Brush cells (BCs), also designated tuft cells (the second most common term), caveolated, multivesicular or fibrillovesicular cells, were described over 50 years ago in the mouse stomach and duodenum (Järvi & Keyriläinen, 1955) and in rat trachea (Rhodin & Dalhamn, 1956). Since then, they have been detected in a variety of mammals in lung (Meyrick & Reid, 1968), gastrointestinal (GI) tract (Luciano et al. 1968a,b), including gallbladder (where they are particularly abundant) (Luciano & Reale, 1969), pancreas (Weyrauch & Schnorr, 1976), parotid and submandibular gland (Sato & Miyoshi, 1988).

Cells with similar morphological features have also been described in amphibians and fishes (Sugimoto et al. 1985; Whitear, 1992; Podkowa & Goniakowska-Witalinska, 2002), suggesting that they may be evolutionarily well conserved.

In humans, BCs have been identified in the adult (Rhodin, 1959, 1966; Watson & Brinkman, 1964; Basset et al. 1971) and fetal (DiMaio et al. 1990) respiratory tract; in conditions such as desquamative interstitial pneumonitis (DiMaio et al. 1988) and immotile cilia syndrome (Gordon & Kattan, 1984; Cerezo & Price, 1985), but not in healthy alveolar lining, and in gastric mucosa (Johnson & Young, 1968) and gallbladder (Gilloteaux et al. 1989). Carstens et al. (1976) described tumour cells with the ultrastructural characteristics of BCs in a human ileal carcinoma, and suggested that they could be the malignant counterparts of intestinal BCs. Brush cells have also been identified in two human colon carcinoma cell lines (Barkla et al. 1988).

The majority of papers examining BC morphology refer to rodents. Studies of the ultrastructure of human small intestine BCs are not available. We thus performed an ultrastructural investigation of BCs identified in duodenal biopsies from five patients with GI tract disorders with a view to exploring their morphology and gaining insights into their elusive functional role.

Materials and methods


Six BCs were identified in five out of 300 small intestinal (duodenojejunal junction) peroral mucosal biopsies from patients with GI disorders examined at our institution in the last 30 years. The five patients, three males and two females, aged 9 months to 4 years (mean 24.2 months), suffered from chronic diarrhoea (n = 2), coeliac disease in dietary treatment, intractable diarrhoea requiring parenteral nutrition, and cystic fibrosis with intractable diarrhoea requiring total parenteral nutrition.

All biopsies were performed with fully informed parent consent as part of the routine investigation for GI tract disorders.

Electron microscopy

Small intestinal biopsy specimens were processed for transmission electron microscopy (TEM). Specimens were fixed in 2% glutaraldeyde/2% paraformaldehyde in 0.1 m phosphate buffer for 3 h at 4 °C, postfixed in 1% osmium tetroxide in the same buffer solution, dehydrated in graded alcohols, and embedded in an Epon–Araldite mixture. For each biopsy, 5–10 random semithin sections (2 µm) were obtained with a MICROM HM 355 microtome (Zeiss, Oberkochen, Germany) and stained with toluidine blue. Thin sections were obtained with an MTX ultramicrotome (RMC, Tucson, AZ, USA), stained with lead citrate and examined with a CM10 transmission electron microscope (Philips, Eindhoven, the Netherlands).

The length and thickness of BC microvilli were measured with a micrometer; enterocytes were also measured for comparison. Despite being from patients with GI tract diseases, the latter cells had a normal ultrastructural appearance and microvilli were of normal length and thickness (Fawcett, 1994). Only perfectly longitudinally sectioned microvilli were included in the morphometric analysis. Length was calculated from the base of the finger-like projections of the luminal plasma membrane to the apex, and thickness in the intermediate and basal areas. The morphometric data of the two cell types were analysed by non-parametric Wilcoxon's test for independent samples. Values were expressed as median and interquartile range (25th–75th percentiles). A level of 0.05 was established for statistical significance.


Six BCs were detected in five out of more than 300 biopsies examined by TEM at our institution, two cells being identified in one sample. Five BCs were located in the epithelium of intestinal villi (intermediate and basal zones) and one in a crypt.

Although easily identified by TEM, BCs were not readily recognized under the light microscope on toluidine blue-stained semithin sections. In fact, they were identified only after they had been evidenced by electron microscopy (Fig. 1).

Figure 1.

Toluidine blue-stained semithin section of a human intestinal villous portion. A cell among the enterocytes (arrow) has been identified as a brush cell by electron microscopy. Some lymphocytes (L) and goblet cells (*) among the enterocytes are also evidenced. E, epithelium; LP, lamina propria. Scale bar = 13.5 µm.

The five BCs in the surface epithelium were scattered among enterocytes. Four had a cylindrical form with a usually rounded or oval basal nucleus containing both marginated (n = 2) and finely dispersed (n = 2) chromatin. One had a cuboid shape and contained a horseshoe nucleus with marginated chromatin. Sometimes the nucleus contained a small nucleolus. All cells had a narrow apex projecting distinctive microvilli into the luminal surface (Fig. 2). These were longer (P < 0.001, mean length 1.7 µm, 25th–75th percentiles: 1.4–2.0 µm) and larger (P = 0.007, mean thickness 0.12 µm, 25th–75th percentiles: 0.11–0.17 µm) than those of enterocytes (mean length and thickness 1.0 µm, 25th–75th percentiles: 0.8–1.1 µm and 0.09 µm, 25th–75th percentiles: 0.09–0.10 µm, respectively) (Fig. 2 and inset).

Figure 2.

General appearance of a human small intestinal brush cell: note the narrow apex and the tuft of microvilli (Mi). The cellular surface forms finger-like projections (arrowheads). Inset: enlargement of the apex showing the long filamentous core (FC) and caveolae (Ca). The cytoplasmic projections also contain filaments. Gl, glycogen; RER, rough endoplasmic reticulum. Scale bar = 0.7 µm; inset scale bar = 0.4 µm.

In the apical part of the cytoplasm, a well-developed filamentous core extended from the microvilli into the cell body to form unusually long rootlets projecting downwards to the supranuclear region (Figs 2 and 3). Unlike absorptive cells, BCs had no terminal web in the apical cytoplasm. Core rootlet filaments had a mean diameter of about 6 nm (microfilaments). In addition, BCs had a thinner glycocalyx than enterocytes.

Figure 3.

High magnification of the apical part of the brush cell shown in Fig. 1. Note both the long bundles of straight filaments (FC) extending from the core of microvilli (Mi) deep into the cytoplasm and a cytoplasmic projection (*) on the lateral border penetrating into the enterocyte (E). F, filaments; G, Golgi apparatus; Ly, lysosomes; M, mitochondria; N, nucleus. Scale bar = 0.25 µm.

Between and below the basal parts of microvilli the plasma membrane formed electronically empty apical caveolae (vesicles) (Fig. 2 and inset). Caveolae were also detected between the filament bundles of the core, always in supranuclear position.

The rough endoplasmic reticulum (RER) was in perinuclear position; it was relatively scanty and organized into short cisternae in three BCs, whereas it was abundant and exhibited long, parallel cisternae in the other two (Fig. 2). Rare cisternae of smooth endoplasmic reticulum were also detected.

BCs consistently showed a well-developed Golgi apparatus and relatively abundant mitochondria with variable morphology, both located in supranuclear position (Fig. 3).

The filaments, besides forming long rootlets, were also scattered in the cytoplasm. Some were also slightly wavy and had a diameter of about 8–9 nm (intermediate filaments). Microtubules were rare.

Lysosomes, multivesicular bodies and small lipid droplets were observed, as were numerous free and aggregate ribosomes (Fig. 3).

Glycogen was always present (abundant in one cell) in the form of alpha and beta particles (Fig. 2).

The basal and lateral cell borders were irregular for the presence of finger-like projections that penetrated into corresponding invaginations of surrounding enterocytes. These cytoplasmic projections also contained filaments (Fig. 2).

Junctional complexes were observed at the apical perimeter and numerous small desmosomes were scattered over the lateral surfaces of BCs.

Finally, in one cell two unmyelinated nervous fibres were in close contact with the BC: one was on the basal surface (Fig. 4 and inset) and the other on the lateral surface of the apical part (Fig. 5) of the cell. These axons contained clear, sparse synaptic vesicles approximately 40–50 nm in diameter, while numerous dense core granules (80–90 nm in diameter) were intermingled with the vesicles (inset Fig. 4). A membrane density on both nerve ending and the BC, suggestive of a synapse, was not observed.

Figure 4.

Two nerve fibres (arrows) in contact both with the basal and with the lateral portion of a human small intestine brush cell (BC). Inset: high magnification showing vesicles and dense core granules in the basal axon. Scale bar = 1.3 µm; inset scale bar = 0.2 µm.

Figure 5.

High magnification of the apical part of the brush cell (BC) shown in Fig. 4. Note the close contact with the lateral nerve fibre (arrow). Scale bar = 0.45 µm.

The BC identified in the crypt was found between two goblet cells (Fig. 6). It was cylindrical and, like all BCs, it was narrower at the apex, which contacted the crypt lumen, than at the base. The nucleus was round and contained finely dispersed chromatin. The cell was characterized by prominent microvilli and a long filamentous core, but the content in other organelles (RER, Golgi apparatus, mitochondria, etc.) was scarce compared with the BCs located on villi.

Figure 6.

Human small intestinal crypt. A brush cell (BC) between two goblet cells (GC). Note the immature appearance of the cell. L, lumen. Scale bar = 1.3 µm.


BCs are quite uncommon in human small intestine, as in rodents (Isomäki, 1973). Indeed, we found only six BCs in five out of 300 biopsies examined at our institution, even though their density may have been underestimated given that neither serial sectioning nor immunohistochemical study were performed. Nevertheless, we believe the ultrastructural characteristics of their cytoplasm make them easily identifiable independently of the presence of a prominent tuft of microvilli. There was no apparent correlation between their presence and the condition that had led to the ultrastructural study.

Human small intestine BCs are also particularly difficult to distinguish from other cell types, such as enterocytes or neuroendocrine cells, by light microscopy on semithin sections. This difficulty can be ascribed to the fact that the tuft of the microvilli is masked by the different thickness of the glycocalyx in the various cell types. Although BCs have more developed microvilli than enterocytes, as demonstrated herein and elsewhere (Isomäki, 1973; Nabeyama & Leblond, 1974; Trier et al. 1987), they have a thinner glycocalyx (Luciano et al. 1968a; Isomäki, 1973; Luciano & Reale, 1990); the enterocyte glycocalyx is thicker and may extend up to 0.5 µm beyond the tips of the microvilli (Fawcett, 1981). This is in line with evidence that the glycocalyx of BCs differs considerably from that of enterocytes and goblet cells (Clark et al. 1995; Gebhard & Gebert, 1999).

Our ultrastructural study of human small intestinal epithelial BCs confirms previous reports (Isomäki, 1973; Nabeyama & Leblond, 1974; Trier et al. 1987) by describing a narrow apical pole from which emerges a well-developed tuft of microvilli, a filamentous core extending into the cytoplasm without forming the terminal web, numerous apical vesicles and abundant filaments both in the cell body and in the basolateral cytoplasmic projections. Filaments are of two types: micro- and intermediate filaments. The latter have already been described by TEM (Wattel & Geuze, 1978; Luciano et al. 1981; Gomi et al. 1991). Interestingly, in the BCs of the present specimens microtubules were rare, at variance with the data reported by Isomäki (1973) who considered their abundance as a characteristic element. By contrast, two BCs exhibited abundant RER, which was scarce in the cells described previously (Isomäki, 1973; Trier et al. 1987). The consistent finding of glycogen in the cytoplasm of BCs does not necessarily point to increased metabolic activity (Ghadially, 1997). Lysosomes, lipid droplets and multivesicular bodies also lack specificity.

The only BC localized in a crypt had a less mature morphological aspect than those on villi, in line with the stem function of the intestinal crypt (Tsubouchi & Leblond, 1979; Karam, 1999).

BCs resemble GI tract neuroendocrine cells, both being polarized cells with a narrow apical pole and a tuft of microvilli extending into the lumen. However, neuroendocrine cells lack both the developed filamentous core and the caveolae, while they contain numerous small neuroendocrine granules aggregated in the basal cytoplasm (Langley, 1994; Rindi et al. 2004). BCs do not contain secretory granules (Isomäki, 1973; Nabeyama & Leblond, 1974; Trier et al. 1987; and present study) and should thus be considered as a distinct cell type unrelated to classic enteroendocrine cells of the gut. This view is further supported by the lack of expression in BCs of two markers typical of neuroendocrine cells, chromogranin A and serotonin (the latter specific for enterochromaffin cells) (Höfer & Drenckhahn, 1996a, 1998). However, Sato & Miyoshi (1997) observed electron-dense granules in main excretory duct epithelium BCs of rat submandibular gland. As also discussed by the authors, these granules do not correspond to neuroendocrine granules, but are glycocalyceal bodies found on the apical surface of normal and neoplastic intestinal-type epithelial cells, respiratory tract and oesophagus. The function and nature of glycocalyceal bodies are still unknown.

The high concentration of cytoskeletal proteins, such as cytokeratin 18, tubulin and ankyrin, described in rat GI tract BCs (Höfer & Drenckhahn, 1996a), suggests that they might confer mechanical stability and polarization. Interestingly, BCs maintain structural polarity also after isolation (Luciano et al. 1993).

Some cytoskeletal proteins are found not only in the apical microvilli but also in the basolateral projections of BCs (Höfer & Drenckhahn, 1992, 1996a). These immunohistochemical data agree with the observation of filaments in the cytoplasmic blebs evidenced in the present study of human BCs and by other authors in mice (Luciano & Reale, 1990).

The polarization and high level of cytoplasmic organization of BCs also allow us to exclude their nature as undifferentiated cells, as suggested by Johnson & Young (1968).

The significance of the caveolae is elusive. Using cationized ferritin and horseradish peroxidase it has been demonstrated that these vesicles are not resorptive, as they do not show tracer uptake (Luciano & Reale, 1979; Trier et al. 1987; Sato & Miyoshi, 1997). Gebert et al. (2000) reported that some BCs examined by electron microscopy showed an inhomogeneous distribution of lectin-bound glycoconjugates. The authors speculate that this arrangement may reflect a relatively high membrane turnover due to continuous loss of membrane fragments to the lumen and replenishing by cytoplasmic vesicles that fuse with the apical membrane at the base of the microvilli.

There are no doubts regarding the morphology (Isomäki, 1973; Nabeyama & Leblond, 1974; Trier et al. 1987; present study) and the turnover (Tsubouchi & Leblond, 1979; Karam & Leblond, 1993; Karam, 1999) of BCs, even though their function is as yet unknown. Most authors believe that they have a sensory function and probably serve as chemoreceptors with a role in certain functional aspects of the GI tract. Indeed, BCs share several morphological characteristics with taste bud sensory cells, for example the shape of the microvilli and a specialized glycocalyx (Luciano et al. 1968a; Witt & Miller, 1992; Gebhard & Gebert, 1999); a similar composition of the cytoskeleton (cytokeratin 18, villin, fimbrin, ankyrin, neurofilaments) (Höfer & Drenckhahn, 1992, 1996a, 1999; Kasper et al. 1994; Zeng et al. 1995; Gebert et al. 2000; Luciano et al. 2003); and the presence of the G-protein α-gustducin. The last of these is specific for taste receptor cells of the tongue (Höfer et al. 1996b; Höfer & Drenckhahn, 1998; Sbarbati et al. 2004) and is involved in sweet and bitter gustatory function (Wong et al. 1996). A further element in favour of a possible receptor role of BCs is the observation that rat stomach and pancreatic duct BCs are particularly rich in nitric oxide synthase I (NOS-I) and might thus use nitric oxide (NO) as a paracrine gaseous messenger molecule (Kugler & Drenckhahan, 1994; Kugler et al. 1994). Finally, unmyelinated nervous fibres in close contact with the lateral surface of the basal portion of BCs have been demonstrated in rat trachea (Luciano et al. 1968a; Chang et al. 1986) and in main excretory duct epithelia of the three major rat salivary glands (Sato & Miyoshi, 1996, 1997; Sato et al. 2002).

Our ultrastructural study of human small intestine BCs confirmed the presence of nervous fibres in contact not only with the basal portion, but also with the lateral surface of BCs. Identification of nerve fibres allows to hypothesize that the nerves could influence the functions of BCs. To our knowledge, this is the first ultrastructural report of axonal structures associated with human BCs.

In conclusion, our data indicate that: (1) BCs in human small intestine are uncommon and difficult to identify by light microscopy; (2) they share similar ultrastructural features with the BCs of other mammals; and (3) the contacts detected between axons and BCs confirm their sensory function, suggesting a diffuse chemosensory system in the GI and respiratory tracts having analogies with the chemosensory systems described in aquatic vertebrates (Sbarbati & Osculati, 2005a,b; Tizzano et al. 2006).


We are grateful to Drs Mariella Marelli and Maria Cristina Zingaretti for their excellent technical assistance. We are indebited to Professor Flavia Carle for the statistical analysis. This study was supported by grant from the Polytechnic University of Marche (2006 FAR, formerly 60%) to M.M.