The Transcription Factor c-Maf Controls Touch Receptor Development and Function

See allHide authors and affiliations

Science  16 Mar 2012:
Vol. 335, Issue 6074, pp. 1373-1376
DOI: 10.1126/science.1214314


The sense of touch relies on detection of mechanical stimuli by specialized mechanosensory neurons. The scarcity of molecular data has made it difficult to analyze development of mechanoreceptors and to define the basis of their diversity and function. We show that the transcription factor c-Maf/c-MAF is crucial for mechanosensory function in mice and humans. The development and function of several rapidly adapting mechanoreceptor types are disrupted in c-Maf mutant mice. In particular, Pacinian corpuscles, a type of mechanoreceptor specialized to detect high-frequency vibrations, are severely atrophied. In line with this, sensitivity to high-frequency vibration is reduced in humans carrying a dominant mutation in the c-MAF gene. Thus, our work identifies a key transcription factor specifying development and function of mechanoreceptors and their end organs.

Low-threshold mechanoreceptors respond to innocuous mechanical stimulation and are crucial for touch sensation (1, 2). Mechanoreceptors are physiologically and morphologically diverse and are classified as rapidly adapting, slowly adapting, and D-hair mechanoreceptors (RAMs, SAMs, and D-hairs) on the basis of responses to sustained stimuli and conduction velocities. RAMs/SAMs and D-hairs have large- and medium-diameter myelinated axons, respectively, and are further distinguished by their end-organ morphology. Pacinian corpuscles, Meissner corpuscles, and lanceolate endings are RAMs, whereas Merkel cell-neurite complexes and Ruffini corpuscles are SAMs (3, 4). Distinct mechanoreceptors detect different stimuli, for example, high- or low-frequency vibration, movement, or static skin indentation. Mechanosensory neurons in dorsal root ganglia (DRGs) derive from neural crest cells, but little is known about their molecular characteristics and development (5, 6). Low-threshold mechanoreceptors and proprioceptors are born during an early neurogenic wave and depend on the basic helix-loop-helix transcription factor Ngn2 for their generation (7). The tyrosine kinase receptor Ret is expressed in early- and late-born neurons (810). Recent work showed that “early Ret” neurons are RAMs and depend on Ret for their development (1012). RAMs also express the transcription factor MafA; however, mutation of MafA has little impact on these receptors (11, 13). The Maf family encodes basic leucine-zipper factors, whose founding member, v-Maf, was identified as a retroviral oncogene (14). We show that c-Maf/c-MAF, a gene known to control eye and lens development (1518), is crucial for mechanosensory function in mice and humans.

In a screen for genes expressed in the developing murine nervous system, we found c-Maf and MafA coexpressed in mechanosensory DRG neurons (Fig. 1) and lamina III of the dorsal spinal cord [compare (19) with fig. S1A]. c-Maf expression in lumbar DRGs starts around embryonic day 11 (E11), and these “early” c-Maf+ neurons coexpress Ret and MafA (Fig. 1A), identifying them as RAMs (10, 11). c-Maf+/Ret+/MafA+ neurons are maintained during development and postnatal life (Fig. 1B and fig. S1, B to D). Around E13, additional sensory neuronal classes begin to express c-Maf (Fig. 1C). Early and late c-Maf+ neurons in postnatal DRGs represent 7 and 15% of all neurons, respectively (fig. S1E). All c-Maf+ neurons coexpress NF200, indicating that they have myelinated axons (Fig. 1, D). Furthermore, c-Maf expression in DRG neurons is evolutionarily conserved and detectable in chicks and humans (fig. S1, F and G).

Fig. 1

Sensory neuron classes expressing c-Maf. (A and B) c-Maf, MafA, and Ret expression in lumbar DRG neurons at E11 and postnatal day 15 (P15). (C) Quantification of c-Maf and Ret coexpression at different stages; the vast majority of c-Maf+ neurons are Ret+ at E11, but subsequently c-Maf+/Ret neuronal classes appear. Error bars indicate SEM. (D to I) At P15, all c-Maf+/Ret+ neurons coexpress NF200 (D). Further, c-Maf+ subpopulations coexpress TrkB and Ret (E), CB and Ret (F), CB and TrkB (G), TrkC and Ret (H), and CR and Ret (I). Arrows and arrowheads point toward c-Maf+/CB+ and c-Maf+/CB neurons, respectively, in (G) and toward c-Maf+/Ret+ and c-Maf+/Ret neurons, respectively, in other panels. (J) Quantification of distinct c-Maf+ neuronal populations at P15. Scale bars indicate 30 μm.

To characterize early c-Maf+ neurons, we used a marker panel that defined three c-Maf+/Ret+/NF200+ subtypes on a molecular basis: TrkB+/calbindin+ (CB), TrkC+, and calretinin+ (CR), which represent 3, 3, and 1% of all DRG neurons, respectively (arrows in Fig. 1, D to I, and fig. S2). Late c-Maf+/NF200+ neurons are also heterogeneous, comprising Ret/TrkB+ mechanoreceptors, Ret/TrkC+/parvalbumin+ mechanosensory muscle afferents, and a few Ret/TrkA+/CGRP+/IB4 nociceptors (arrowheads in Fig. 1, D to I; figs. S2 and S3; summary in Fig. 1J). Furthermore, FABP7+ glia also express c-Maf (fig. S3F).

We used mouse genetics to define the role of c-Maf in sensory neurons. Because the null mutation is lethal (15), we generated conditional mutant mice (Isl1cre c-Mafflox/– mice, hereafter called c-Maf mutants). With use of this strategy, recombination in DRG neurons occurred before c-Maf expression; hence, conditional mutant DRG neurons never expressed c-Maf. However, c-Maf was present in the spinal cord and glia (see fig. S4 for additional information). In c-Maf mutant DRGs, there was no significant change in NF200+ neuron quantity. We also assessed the integrity of peripheral nerves. No loss of myelinated axons from the saphenous nerve, a cutaneous nerve rich in mechanoreceptors, was observed (fig. S5, A and B). In c-Maf mutants, axonal size distribution in the saphenous nerve was altered such that large diameter axons became thinner. Myelinated fibers are classified as Aβ (large diameter, fast conducting, >10 m/s) and Aδ (medium diameter, medium conducting, <10 m/s). The shift of axon diameter in the saphenous nerve of c-Maf mutants correlated with a reduced conduction velocity of Aβ but not Aδ fibers (Fig. 2A). Most myelinated axons in the interosseous nerve terminate in Pacinian corpuscles, a scarce mechanoreceptive subpopulation. In contrast to the saphenous nerve, many myelinated axons >4 μm were lost in the interosseous nerve of c-Maf mutants (Fig. 2B).

Fig. 2

Disrupted RAM function and end organs in c-Maf mutant mice. (A and B) c-Maf loss affects myelinated axons of saphenous and interosseous nerves, resulting in a shift of axon diameter in saphenous nerves and a loss of large (>4 μm) axons in interosseous nerves. WT, wild type. (C to F) c-Maf loss results in altered RAM responses to mechanical stimuli in the saphenous nerve. (C) Spike trains of RAMs and SAMs in response to a 512-μm mechanical stimulus. (D) Distribution of RAM, SAM, and mechanically nonresponsive (n.r.) fibers. (E) RAM responses to stimuli of variable magnitudes (spike counts) and (F) variable velocities (interspike intervals). The inset in (F) shows representative RAM spike trains at the beginning of a 512-μm stimulus; top and bottom trains were recorded from a control and a c-Maf mutant mouse, respectively. (G to J) c-Maf loss affects morphologies of lanceolate and circumferential endings associated with tylotrich hair of back skin (G) and hair of dorsal hindpaw (H and I), and of Meissner corpuscles (J). Arrowheads and arrows in (G) and (H) point to type I and II lanceolate endings, respectively. In (I), the arrow and the arrowhead marked by an asterisk point toward type II and III lanceolate endings, respectively. Scale bars for (B) and (J inset), 20 μm; (G to I), 12.5 μm; and (J), 100 μm. Statistical significance is indicated in all figures (*P < 0.05; **P < 0.01; ***P < 0.001). Error bars, SEM.

We next investigated whether the c-Maf mutation alters mechanoreceptor function and used an in vitro skin-saphenous nerve preparation to record mechanically evoked responses from single nerve fibers (20). In controls, skin indentation evokes RAM firing solely during the onset of stimulus movement. In contrast, RAM firing in c-Maf mutants was prolonged and even continued into the beginning of the static phase. SAMs fired during the entire static phase in controls and c-Maf mutants, and therefore remained distinguishable from mutant RAMs (Fig. 2C and fig. S5, D and E). RAMs and SAMs were present in normal proportions in c-Maf mutants (Fig. 2D). Increased RAM spiking was observed over a wide range of displacement amplitudes (Fig. 2E), and spike intervals were shorter (Fig. 2F). In contrast, firing properties of SAMs were not significantly altered, but their von Frey thresholds were reduced (fig. S5, C to F). D-hair receptors and Aδ nociceptors were unaffected (Fig. 2A and fig. S5, G and H). Thus, the c-Maf mutation profoundly disrupts cutaneous RAM function.

Next, we studied mechanoreceptor end organs in c-Maf mutant skin. We observed that lanceolate endings were heterogeneous with respect to marker expression and c-Maf–dependence, indicating functional heterogeneity. We visualized lanceolate endings and associated terminal Schwann cells and defined three neurochemical types: type I (S100+/NF200+/CB+/TrkBlow) (arrowheads in Fig. 2, G and H), type II (S100+/NF200/CB/TrkBhigh) (arrows in Fig. 2, H and I), and type III (S100+/NF200/CB/TrkB) (asterisks in Fig. 2I) (21). Type I endings are associated with tylotrich and nontylotrich hair in back skin and with the morphologically distinct dorsal hindpaw hair. Type I lanceolate endings displayed lower levels of NF200 and CB, but not TrkB, in c-Maf mutants, but type II and III endings were unaffected (22); the morphology of circumferential endings was also disrupted (Fig. 2, G and H, and fig. S6, B to D). Meissner corpuscles represent an additional type of RAM ending in the glabrous skin and are, like type I lanceolate endings, innervated by axons coexpressing NF200/TrkBlow/CB (fig. S6E). In c-Maf mutants, the quantity of Meissner corpuscles was reduced by 75% ± 9% and remaining corpuscles appeared rudimentary (Fig. 2J), but the NF200+ axons innervating dermal papillae were not lost (8.0 ± 0.2 and 7.1 ± 0.5 innervating axons per section in control and c-Maf mutants, respectively). No changes were observed in Merkel cell-neurite complexes (SAMs), muscle spindles, and nociceptive skin endings (fig. S6, F to I).

We next analyzed Pacinian corpuscles, the primary detectors of high-frequency vibration. Pacinian corpuscles are abundant in human palms and fingers, but in rodents they cluster in periostea (23). In c-Maf mutants, the quantity of Pacinian corpuscles associated with fibulae was reduced by 58%, correlating with the axonal loss (61%) in interosseous nerves; remaining corpuscles were small and possessed irregularly shaped S100+ cores (Fig. 3, A and B). Thus, four mechanoreceptor classes depend on c-Maf: Three terminate in the murine skin (type I lanceolate endings, circumferential endings, and Meissner corpuscles), and the fourth, Pacinian corpuscles, is abundant in murine periostea. Axonal loss is, however, restricted to Pacinian corpuscle innervation.

Fig. 3

Pacinian corpuscles in c-Maf mutant mice and sensitivity to vibration in patients with dominant c-MAF mutation. (A and B) c-Maf mutation in mice causes a loss of Pacinian corpuscles and disrupts the morphologies of remaining corpuscles. (C) Pedigree of a family with the R288P c-MAF mutation; members carrying the mutant allele (black symbols) suffer from early-onset cataracts. (D and E) Detection threshold of vibrotactile stimuli. (D) Example of a vibration stimulus and response and (E) vibration detection threshold in control and four carriers of the R288P c-MAF mutation. In healthy participants and unaffected family members, the detection threshold decreases with increasing frequencies. The detection threshold of affected family members is aberrant at high but not low frequencies. Scale bars, 100 μm. Error bars, SEM.

In humans, dominant c-MAF mutations are associated with ocular developmental abnormalities and cataracts (1618), but their effects on mechanosensation have not been examined. Encouraged by our results in mice, we tested touch sensitivity in a family comprising four carriers of the dominant Arg288→Pro288 (R288P) c-MAF mutation (16) (Fig. 3C). This mutation in the auxiliary DNA binding domain interferes with c-MAF–dependent transcriptional activation but does not eliminate c-MAF function (24) and represents one of three known cataract-causing point mutations in c-MAF (1618). To assess the function of Meissner and Pacinian corpuscles, we tested vibrotactile acuity over a wide range of frequencies (5 to 240 Hz; Fig. 3, D and E). Pacinian corpuscles are essential for the detection of small-amplitude high-frequency vibrations (2527). We observed a large increase in the vibration amplitude required to elicit responses in c-MAF mutant carriers at high but not low frequencies (Fig. 3E). Tactile spatial acuity, that is, the ability to distinguish grids of different spacing, is thought to depend on hair follicle afferents/Meissner corpuscles/Merkel cell-neurite complexes (25, 28, 29) and was not significantly changed (fig. S6J). Thus, the R288P c-MAF mutation interferes with normal vibration detection in humans.

To understand the mechanism underlying mechanoreceptive deficits, we identified changes in gene expression in DRG neurons of c-Maf mutants. Ret and MafA expression was strongly reduced at birth as assessed by counting Ret+/NF200+ and MafA+ neurons (Fig. 4, A, B, and F) and modestly or not affected at E13.5 (fig. S7). Thus, c-Maf is required to maintain but not initiate Ret and MafA expression. We also analyzed DRGs of Ret mutant mice and found that c-Maf expression was unchanged (fig. S7). Thus, c-Maf acts upstream to maintain but not initiate Ret expression during RAM development.

Fig. 4

c-Maf and the molecular mechanisms of mechanosensory dysfunction. (A to F) The numbers of RAM neurons that express Ret and MafA and the numbers of neurons that express TrkB, Cav3.2, and Kcnq4 were quantified in control and c-Maf mutant DRGs. (G and H) Interspike interval of RAMs measured in the skin-nerve preparation of the saphenous nerve in control and c-Maf mutant mice in the presence and absence of linopirdine. Representative traces (G) and quantification of interspike intervals (H) are shown. Scale bars for (A) and (B), 30 μm; (C) to (E), 100 μm. Error bars, SEM; n.s., not significant.

We also examined the expression of receptors and ion channels characteristic of mechanoreceptors like TrkB, Cav3.2, and Piezo2, which were unchanged in c-Maf mutants (Fig. 4, C, D, and F, and fig. S7M). Further deregulated genes were identified by using microarray analysis and in situ hybridization. Among these were crystallins (Cryba2 and Crygs), known c-Maf targets encoding structural lens proteins (15). Crystallins also act as chaperones and can impinge on neural physiology. Further, genes encoding the potassium channels Kcng4, Kcnq4, Kcna1, and Kcnh5 were down-regulated in c-Maf mutants (Fig. 4, E and F, and table S1). Kcnq4 was recently found to fine-tune mechanoreceptor function (30). The Kcnq channel blocker linopirdine shortens interspike intervals of RAMs in control but not c-Maf mutants (Fig. 4, G and H), directly implicating Kcnq4 in one aspect of the functional changes that are observed in RAMs of c-Maf mutant mice.

Our analyses show that c-Maf directs mechanoreceptor development. c-Maf acts upstream of Ret, a receptor known to control RAM development, providing a mechanistic basis for this phenotype. c-Maf and MafA are similar in structure and might act redundantly. The fact that MafA expression is not maintained in c-Maf mutant mechanoreceptors can explain the salient function of c-Maf. c-Maf controls many parameters of RAM development, morphology, and function and modulates functional aspects of SAMs. Our analysis also reveals that the most strongly affected RAM subtype in the c-Maf mutant mice are Pacinian corpuscles that specialize in the detection of high-frequency vibration. Pacinian corpuscles and associated axons were largely absent, and residual corpuscles exhibited disrupted morphologies. Because the deformation of the lamellar end organ of Pacinian corpuscles initiates axonal firing (31), remaining corpuscles are expected to be functionally impaired. In line with this, humans carrying a dominant cataract-causing c-MAF mutation displayed reduced acuity to high-frequency vibration. Other mechanisms like aberrant spinal processing of vibrotactile information might also contribute to this deficit in humans. We conclude that the transcription factor c-Maf directs RAM development and formation of RAM mechanoreceptive end organs.

Supporting Online Material

Materials and Methods

Figs. S1 to S7

Table S1

References (3240)

References and Notes

  1. Type II lanceolate endings are TrkBhigh, and TrkBhigh neurons do not express c-Maf (<5%). TrkBhigh neurons coexpress Cav3.2 and correspond thus to D-hairs (fig. S6A). Type III lanceolate endings are NF200, and NF200 neurons do not express c-Maf.
  2. Heterogeneity of lanceolate endings was recently also noted by L. Li et al., Cell 147, 1615 (2011). They report that Aβ-, Aδ-, and C-low threshold mechanoreceptors terminate in lanceolate endings, and comparisons of marker expression indicate that these correspond to type I, II, and III endings, respectively.
  3. Acknowledgments: We are grateful to I. Schiffner, M. Terne, S. Buchert, C. Päseler, P. Stallerow, and the MDC transgenic and microarray units for technical support and animal husbandry and R. Schmidt-Ullrich, R. Hodge, A. Garratt, and W. Birchmeier (MDC, Berlin) for advice during analysis of hair types and critical reading of the manuscript. This work was funded by grants to C.B. and G.R.L. (SFB 665, NeuroCure). The authors declare no conflicts of interest.

Stay Connected to Science

Navigate This Article