Macrophage migration inhibitory factor acts as a neurotrophin in the developing inner ear

Over the last 20 years, a number of investigators have demonstrated interactions between the immune and neural systems during normal organismal development (Patterson, 1994; Armstrong et al., 2003; Siemion et al., 2005; Wrona, 2006). Our laboratories have been investigating proteins that regulate innervation during the earliest stages of inner ear development (reviewed by Barald and Kelley, 2004; Gerlach et al., 2000), and our recent work highlights the importance of chemokines and cytokines in the early stage peripheral auditory system (Bianchi et al., 2005; Holmes et al., 2011; Shen et al., 2012).

We began investigating the neurotrophic role of immune system chemokines and cytokines during the earliest stages of neuronal innervation, using the ear as a model system (Bianchi and Cohan, 1991; Barald et al., 1997; Germiller et al., 2004; Bianchi et al., 2005; Holmes et al., 2011; Shen et al., 2012). In earlier reports, we used gain- and loss-of-function approaches to demonstrate that the cytokine macrophage migration inhibitory factor (MIF) acts as a neurotrophic factor in the developing zebrafish inner ear (Holmes et al., 2011; Shen et al., 2012). In this report, we show that MIF plays a similar role in the chick and mouse inner ear; in culture, recombinant MIF promotes both directional neurite outgrowth and the survival of statoacoustic ganglion (SAG) neurons in both species.

Previous studies from our laboratories and others have indicated that cells in the otocyst secrete target-derived factors [otocyst-derived factor(s) (ODF)] to promote the initial outgrowth and survival of the SAG [at embryonic day (E) 4-6 in chick, E11-14 in the mouse and rat] (Bianchi and Cohan, 1991; Bianchi and Cohan, 1993; Bianchi et al., 1998; Bianchi et al., 2000). However, the identity of the ODF has remained elusive. Although growth factors such as the classical neutrophins have been shown to play a role at later developmental stages in the auditory system, the active components of ODF do not contain identified neurotrophins and have been difficult to ascertain. The small size of the otocyst makes generation of sufficient quantities of bioactive material difficult, further hampering biochemical analysis of active components.

Because so little ODF can be obtained from embryonic otocysts, we generated cell lines from the E9.5 Immortomouse otocyst (IMO) (Barald et al., 1997) that secrete a functionally bioactive ODF equivalent (Germiller et al., 2004; Bianchi et al., 2005). Using cytokine arrays on IMO-generated ODF from these cell lines we first identified the cytokine monocyte chemoattractant protein 1 (MCP1; also known as CCL2) as an active, but not the sole, component of ODF (Bianchi et al., 2005). Because MCP1 alone could not entirely replicate the activity of ODF, we investigated whether other cytokines or cytokine-interacting molecules are also present in ODF.

In this study, through 2D gel analysis and proteomic screens we identified MIF as a major component of ODF, both in abundance and activity. Subsequent in vitro experiments in the developing mouse and chick, as well as analysis of Mif knockout (KO) mice, demonstrated that MIF can account for the major functional activities of ODF, including promoting SAG neuron survival and directional neurite outgrowth, and is required for normal hearing and inner ear innervation patterning. We also demonstrate that residual bioactivity in chick or mouse otocyst-generated ODF that is not removed by function-blocking antibodies to MIF can be entirely eliminated by function-blocking antibodies to the cytokine MCP1, which we previously identified as a component of ODF (Bianchi et al., 2005). Together, these data and our studies of zebrafish otic development (Holmes et al., 2011; Shen et al., 2012) show that the cytokines MIF and MCP1 (Bianchi et al., 2005) play crucial roles in inner ear development, directly demonstrating another important connection between the immune and nervous systems.

Conditioned medium containing ODF (CMODF) was prepared from H2kbtsA58 transgenic IMO cell lines as previously described (Bianchi et al., 2005; Germiller et al., 2004), as well as from E4-6 chick and E9.5-10.5 mouse otocysts as previously described (Bianchi and Cohan, 1993; Barald et al., 1997; Germiller et al., 2004; Bianchi et al., 2005; Thompson et al., 2003).

Sandwich ELISA assays were performed to determine MIF and MCP1 cytokine levels in IMO 2B1DT cell line CMODF and in conditioned medium from E5 chick and E9.5-10.5 mouse otocysts, using an anti-human MIF IgG monoclonal antibody that cross-reacts with human, mouse and chick MIF (monoclonal mouse IgG1 clone #12302 MAB289, R&D Systems, Minneapolis, MN, USA). The assays were performed in triplicate at room temperature; absorbance was read at 450 nm in a Halogen BELLAPHOT plate reader (Fisher Scientific, Pittsburgh, PA, USA).

SAG neurite outgrowth and survival bioassays were performed as described (Bianchi et al., 2005). The effects of MIF and MIF function-blocking antibodies (and in some experiments anti-MCP1 antibodies) were also assessed in these two bioassays. Images were taken on a Leitz Diavert inverted microscope (Leica) with Nikon ACT-1 software.

Recombinant mouse MIF (rmMIF; 1978MF/CF, R&D Systems) or recombinant human MIF (rhMIF; R&D Systems) was added to basal medium or inactive IMO conditioned medium (e.g. from cell line 3C6) at 5 pg/ml to 500 ng/ml (mouse) or 2 ng/ml to 200 ng/ml (human) to evaluate the effects on SAG outgrowth and survival. Extensive dose-response evaluations were first performed for both neurite outgrowth and neuronal survival to determine the optimal concentrations for each bioactivity in each species (mouse, chick). Initial results demonstrated that the low and mid-range concentrations produced the most consistent outgrowth and survival, and 10 ng/ml was used in subsequent experiments. All trials were performed in triplicate, with wells assigned randomly by computer.

In the experiments shown in Fig. 1, the effects of rmMIF were compared with IMO CMODF, and in both cases a MIF function-blocking antibody was used. MIF function-blocking antibody (AB289 goat anti-mouse IgG, R&D Systems) was added to 10 ng/ml rmMIF (1978MF/CF, R&D Systems), a previously determined optimal concentration in basal medium without γ-interferon or FBS, or to 2B1DT IMO CM. As a control, similar samples were treated with an antibody to bovine serum albumen (BSA) (A4338, Sigma-Aldrich, St Louis, MO, USA).

Before use in the SAG neurite outgrowth bioassay or dissociated SAG survival assay, each sample of rmMIF or 2B1DT IMO CMODF was either (1) untreated, (2) treated with 300 ng/ml MIF function-blocking antibody (determined as the optimal concentration after testing 10-1000 ng/ml) or (3) 300 ng/ml anti-BSA antibody. Samples were incubated with protein-G beads, which were removed by spin column (IP50 Protein-G Immunoprecipitation Kit, Sigma-Aldrich). P-values were calculated using a one-tailed t-test with Bonferroni correction; 16 different comparisons were performed (statistical significance is reported in supplementary material Table S1). In experiments in which the anti-MIF-treated conditioned media were subsequently treated with MCP1 function-blocking antibodies (see Bianchi et al., 2005) (supplementary material Fig. S1), all residual bioactivity could be eliminated from either type of ODF.

The D3 mouse embryonic stem cell (mESC) line (S6) was cultured as described (Roth et al., 2007; Roth et al., 2008) to assess the effects of MIF bioactivity on differentiation. Control media included 1 μg/l ciliary neurotrophic factor (CNTF) (R&D Systems), 1 μg/l nerve growth factor (NGF) (Chemicon, Temecula, CA, USA), or various concentrations (1 pg/ml to 500 ng/ml) of rmMIF. mESCs were cultured for another 5 days at 37°C, fed with differentiation media (NGF, CNTF or MIF) every other day. Similar concentrations of rhMIF or rmMIF produced similar effects on neuronal development from the mESCs (data not shown). Extensive dose-response experiments were conducted to determine the optimum concentration of rmMIF or growth factors that resulted in maximum neuronal differentiation as defined by morphology and labeling with anti-neurofilament or anti-TUJ1 (also known as TUBB3) in 3-4 days.

RT-PCR was used to detect expression of Mif and of the MIF receptor Cd74 in embryonic and adult mouse and chick inner ear tissues. Primer sequences (5′-3′; F, forward; R, reverse) were: MIF-F, CCAGAACCGCAACTACAGTAAGC; MIF-R, TTGGCAGCGTTCATGTCGTAATAG; CD74-F, GCTTCCGAAATCTGCCAAACC; CD74-R, AAGAGCCACTGCTTCATCCAGC.

Inner ears were isolated from mice at postnatal day (P) 3 (Chen et al., 2009); organs of Corti (OC) and spiral ganglia (SG) were isolated from age-matched wild-type mice (Balb/c, WT) and Mif KO mice. The OC explants (WT or Mif KO) were then placed 1 cm from the SG explants (n=12 per condition). Tissues were observed daily by light microscopy for the ensuing 7 days. Phase-contrast images (40×) to produce composites of the entire field of view were taken daily beginning at 24 hours. To determine the density of contacts between SG and WT or Mif KO OC, composite images magnified in Adobe Photoshop to the limits of pixel resolution (n=11) were analyzed using a superimposed grid. Only clearly visible contacts (with no underlying neurite extensions) between the neurite and the surface of the OC were counted; these therefore represent a minimum number of contacts owing to the omission of overlapping contact points in many regions. No contacts were seen between the WT SG and the Mif KO OC at any stage or under any conditions.

Developmental expression of MIF protein was assessed in sectioned embryonic inner ears and adult inner ear tissues, as well as in whole-mounted tissues (Gerlach-Bank et al., 2002) using rabbit anti-MIF (1:500; ZMD.321, Invitrogen, Camarillo, CA, USA), goat anti-MCP1 (1:40; R&D Systems) and anti-myosin VIIa (1:200; Sigma), followed by donkey anti-rabbit Alexa Fluor 488 or donkey anti-goat Alexa Fluor 555 (both 1:200; Invitrogen) secondary antibodies and observation by confocal microscopy.

MIF expression was also examined in dissociated and reaggregated inner ear ‘domes’ prepared from E13/14 mouse embryonic otocysts (Bianchi et al., 2002; Bianchi et al., 2006). Dome cells were double labeled for myosin VI (a marker for sensory hair cells; 10 mg/ml; T. Hasson, UCLA) and MIF (1:3000; Santa Cruz Biotechnology, Santa Cruz, CA, USA), or for cytokeratin (a marker for supporting cells; 1:100; Sigma) and MIF. Alexa Fluor 488 goat anti-rabbit and Alexa Fluor 594 rabbit anti-goat secondary antibodies (both 1:2000; Molecular Probes, Eugene, OR, USA) were used to detect myosin VI and MIF, respectively; and Alexa Fluor 488 goat anti-rabbit and Alexa Fluor 555 goat anti-mouse secondary antibodies were used to detect MIF and cytokeratin, respectively.

Auditory brainstem response (ABR) was assessed for 4-week-old WT (Balb/c) (n=5) and Mif KO mice (n=4) (Saul et al., 2008). Responses were collected at 12, 24 and 48 kHz for a range of stimulus levels, separated by 10 dB steps at higher intensities and 5 dB steps near threshold. Thresholds were interpolated between the lowest stimulus intensity that produced a response and the highest stimulus at which no response was observed (5 dB increments).

siRNAs were used to knock down mouse MIF or avian CD74 proteins by introducing a homologous double-stranded (ds) RNA. The nucleotide sequences of the dsRNA and complementary dsRNA for mouse Mif mRNA and avian Cd74 mRNAs are (5′-3′): mouse Mif siRNA duplex, CCGCAACUACAGUAAGCUGdTdT and CAGCUUACUGUAGUUG - CGGdTdT; Mif missense control RNA duplex, GCGCGCUUUGUAGG - AUUCGdTdT and CGAAUCCUACAAAGCGCGCdTdT; avian Cd74 siRNA duplex, GCAACAAGACUGAGGAUCAAATdTd and UUUG - AUCCUCAGUCUUGUUGCTdTd; Cd74 missense control RNA duplex, GCAAGAAGACAGAGGUUCAAATdTd and UUUGAACCUCUG - UCUUGUUGCTdTd. All constructs were confirmed by enzymatic digestion (HindIII) and DNA sequence analysis (DNA Sequencing Core, University of Michigan).

IMO 2B1DT cells expressing MIF and chick neuronal cells expressing CD74 (Bryan et al., 2008) were used to titrate siRNA concentrations necessary for maximal inhibition of MIF or CD74 protein, as determined by western blotting (the maximal inhibition achieved was ∼70%). IMO 2B1DT cells were transfected with mouse Mif siRNA or its missense control and chick neuronal cells were transfected with avian Cd74 siRNA or its missense control (Mirus Reagent, Madison, WI, USA) according to the manufacturer’s protocol. Twenty-four hours post-transfection, western blotting assays (Bianchi et al., 2005) were performed to examine the effect of these siRNAs on MIF and CD74 levels. SAG bioassays (Bianchi et al., 2005) in the presence of siRNA were used to monitor any effect of Cd74 siRNA on ODF-induced SAG neurite outgrowth and survival (supplementary material Fig. S2).

For function-blocking antibody neurite outgrowth permutation test results, pairwise differences between experimental conditions were evaluated using approximate permutation tests. In all cases, the test statistic was the difference between the means of the two groups, and , being compared. In each test, the values and were pooled. We then drew m values from the pool without replacement and assigned them to the first group, assigning the remaining n values to the second group. The statistic was then recomputed and the procedure was repeated. P-values were based on 10⁶ independent draws and evaluations of the statistic. One-sided P-values were found by computing the fraction of replicates in which the value of the resampled statistic was at least as great as its empirically observed value. Two-sided P-values were found by computing the fraction of replicates in which the resampled statistic was at least as great in magnitude as the absolute value of its empirically observed value.

For general statistical analysis of data in other experiments, a one-tail t-test was used. P<0.05 was considered significant in all data.

IMO cell lines, chick (E4-5) and mouse (E11) otocysts were used as sources of ODF bioactivity. Initially, serum-free conditioned medium (CM) from IMO cell lines that promoted both robust outgrowth and significant survival of SAG neurons (bioactive CM) was compared with CM from IMO cell lines that did not promote outgrowth or survival and with serum-free control basal medium alone (inactive or basal medium). Two-dimensional gel analysis and a subsequent proteomic screen (MALDI/TOF/TOF spectroscopy; Proteomics Core, University of Michigan) were used to identify components of the biologically active samples of ODF.

MIF was found in all the bioactive cell lines (e.g. cell line 2B1DT), but not in inactive IMO cell line (3C6) CM or the basal cell-free and serum-free CM samples. Similar findings were obtained in analyses of CMODF from E5 chick and E11 mouse otocysts. Following the identification of MIF and verification that MCP1 was also present in ODF by proteomic analysis, sandwich ELISA assays were performed to quantify these proteins from IMO CMODF and chick and mouse otocyst ODF. MIF was found at 2.40±0.69 ng/ml (n=6) in IMO CMODF, 1.21±0.84 ng/ml (n=46) in chick ODF and 1.32±0.61 ng/ml (n=16) in mouse ODF.

Proteomic analysis of the ODF samples also confirmed the presence of MCP1, which we had previously detected in ODF using a cytokine protein array that did not include MIF (Bianchi et al., 2005). ELISA assays were used to document the presence of MCP1 at 1.22±0.06 ng/ml (n=6) in the IMO cell lines, 1.26±0.22 ng/ml (n=6) in chick ODF and 1.18±0.17 ng/ml (n=14) in mouse ODF.

IMO CMODF and chick and mouse otocyst ODF produce similar effects on neurite outgrowth from SAG explants in vitro. To determine whether MIF alone mediates either SAG neurite outgrowth or SAG neuronal survival, recombinant MIF (rMIF) was tested in mouse and chick in extensive dose-response SAG neurite outgrowth and survival assays. Concentrations of MIF tested varied in the pg/ml to μg/ml ranges, all of which elicited outgrowth and promoted survival. The optimal concentration was determined to be close to the physiological concentration of MIF found in embryos (see above) and in ODF; therefore, recombinant mouse MIF (rmMIF) was used at 10 ng/ml. To determine whether MIF could account for the bioactivity found in ODF, ODF pretreated either with MIF function-blocking antibody or anti-BSA antibody and then activity depleted with protein-G beads, or untreated ODF was added to cultures of explanted SAG. Neurite outgrowth was measured at 24 hours (Fig. 1). Untreated IMO-derived ODF (CM from IMO cell line 2B1DT; Fig. 1D) produced a neurite outgrowth score of 4.1 (Thompson et al., 2003; Germiller et al., 2004). The average neurite outgrowth score in 10 ng/ml rmMIF was 4.2 (Fig. 1D) (Bianchi et al., 2005).

SAG cultured in rMIF pretreated with anti-MIF antibody followed by protein-G beads to remove MIF activity displayed little or no neurite outgrowth (average neurite outgrowth score of 0.2; Fig. 1D), demonstrating a direct effect of MIF on SAG neurite outgrowth. Pretreatment of rMIF with anti-BSA did not diminish the neurite outgrowth produced by rMIF (average neurite outgrowth score of 4.0). However, although pretreatment of chick ODF with the same MIF function-blocking antibody also diminished the effect of chick ODF on neurite outgrowth, the average neurite outgrowth score was 1.1 (Fig. 1D), suggesting that additional outgrowth-promoting factors are present in ODF, such as MCP1, which was found in our earlier study (Bianchi et al., 2005). This indeed proved to be the case. When the anti-MIF-depleted CM from IMO-generated ODF and the CM from E5 chick-generated ODF were subsequently treated with the same MCP1 function-blocking antibody that we used in our earlier study (Bianchi et al., 2005), no significant neurite outgrowth was seen. However, subsequent treatment of the anti-MIF-treated CM with anti-BSA had no effect on the CM activities (supplementary material Fig. S1).

To assess the effects of rMIF on neuronal survival, dissociated SAG neurons were exposed to either IMO ODF or rmMIF, each with and without MIF function-blocking antibody pretreatment. Anti-BSA was again used as the control and did not affect neuronal survival promoted by ODF from any source. Under normal culture conditions, SAG treated with IMO CMODF or 10 ng/ml rMIF had 67% (n=31) or 77% (n=17) of neurons surviving at 24 hours, respectively (Fig. 1E). However, addition of anti-MIF antibody to either rmMIF or ODF significantly reduced SAG neuronal survival. Only 4% of neurons survived with anti-MIF-treated rmMIF (Fig. 1E), whereas 17% of SAG neurons survived when cultured in anti-MIF-treated ODF. Again, this residual survival-promoting activity in ODF is eliminated by subsequent treatment of anti-MIF-treated CM with the same MCP1 function-blocking antibody that we used in our earlier study (Bianchi et al., 2005) (supplementary material Fig. S1). No significant neuronal survival was noted in these ‘twice-treated’ CM; again, subsequent treatment with anti-BSA antibody had no effect on the anti-MIF-treated CM. Therefore, MIF and MCP1 can together account for essentially all the bioactive neuronal survival and outgrowth functions of IMO- and otocyst-generated ODF. There are trace amounts of other cytokines in ODF (Bianchi et al., 2005) but they do not appear to contribute significantly to its bioactivity, although synergies among these cytokines have not been investigated extensively.

Furthermore, treatment of chick SAG explants or dissociated SAG with siRNA to the MIF receptor Cd74 also completely inhibited SAG neurite outgrowth in response to either MIF or ODF and significantly reduced neuronal survival in the presence of MIF or ODF (see supplementary material Fig. S2).

To further examine the capacity of MIF to impact neuronal development, we tested the ability of rmMIF to induce a neuronal phenotype in D3 mouse embryonic stem cells (mESCs) (Doetschman et al., 1985). Concentrations of rmMIF ranging from 1 pg/ml to 500 ng/ml caused a substantial population of mESCs to become neuron-like after 48 hours in culture. Extensive dose-response analyses were performed as illustrated in Fig. 2. The cells treated with rmMIF developed long neurofilament-positive neurites (Fig. 2D-F). By contrast, D3 cells remained proliferative in the absence of rmMIF (Fig. 2A). Control cultures of D3 cells treated with the growth factors/neurotrophins NGF (Fig. 2B) or CNTF (Fig. 2C) similarly took on a neuron-like morphology after 48 hours, as described in our previous studies (Roth et al., 2007; Roth et al., 2008).

Our recent studies of the role of MIF in zebrafish support the idea that MIF might be involved in neuronal development outside the auditory system. Antisense oligonucleotide morpholinos (MOs) to mif and mif-like (ddt – Zebrafish Information Network) injected into either the 1- to 8-cell zebrafish embryo (Shen et al., 2012) or directly into the nascent inner ear at 20 hours postfertilization (hpf) (Holmes et al., 2011) had substantial inhibitory effects on neurite extension in other head ganglia, reducing the size of the zn-5 antibody-stained (zn-5 labels neurons in fish) SAG and other head ganglia by 50% or more (Shen et al., 2012); capped mif and mif-like RNAs rescued this loss. Thus, MIF is able to influence neural development in addition to its effects on SAG neurons.

To further explore the role of MIF in the developing inner ear, immunohistochemical analyses of inner ear tissues from embryonic stages to postnatal stages were undertaken to map protein expression patterns. At the early embryonic stage E9.5, MIF was expressed in the medial otocyst closest to the hindbrain (Fig. 3) and in the ventral half to two-thirds of the hindbrain. At later embryonic stages, MIF was broadly distributed in the epithelia of the cochlea and vestibular regions, although individual cell populations could not be discerned (supplementary material Fig. S3; E15.5, n=5).

To determine the cellular distribution of MIF at the stages of ODF activity, we used two complementary approaches. In addition to labeling sections of developing embryonic mouse otocysts and hindbrain with antibodies to MIF (Fig. 3), we also labeled cultures of re-associated embryonic inner ear cells (Bianchi et al., 2002; Bianchi et al., 2006), a model that we have found useful for distinguishing developing hair cells (HCs) from supporting cells (SCs) (supplementary material Fig. S4). In these cultures, inner ear cell ‘domes’ are produced from dissociated cells from isolated inner ears, which reaggregate with a central mass of HCs (myosin VIIa positive; supplementary material Fig. S4A) resting upon an SC layer (which is cytokeratin positive). MIF was localized to the underlying SC layer, but not the HC layer of these cultured inner ear domes. Cytokeratin expression colocalized with MIF in the SCs (supplementary material Fig. S4B), whereas the central regions of the domes (above the focal plane) containing HCs were not labeled for MIF (supplementary material Fig. S4B). Thus, MIF is localized to developing SC regions in embryonic inner ear tissues, where it might act as a general attractant for SAG neurons early in their development.

To test whether MIF expression persists in mature inner ear tissues, 4-week-old Balb/c (WT) mice (n=4) were examined. MIF was expressed in SCs that cup the inner and outer HCs (IHC and OHC, respectively) (Fig. 4). MIF expression was also detected in the stria vascularis, spiral limbus and spiral ligament (Fig. 4).

The cytokine MCP1 was detected in the IHCs and OHCs of the organ of Corti (OC), as well as in the stria vascularis (Fig. 4). Overlapping staining of MCP1 and MIF (Fig. 4, yellow) was visible in the regions where the SCs cup the HC. The expression of these two cytokines in mature inner ear tissues might suggest roles in maintaining sensory epithelia or neural innervation.

To test for a signaling pathway that might be used by inner ear-derived MIF, we assayed for the presence of the known MIF receptor CD74 in SAG or inner ear epithelial regions. RT-PCR was used to examine Cd74 mRNA expression in E14 mouse SAG neurons and otocysts, as well as in adult mouse spiral ganglia (SG) neurons and adult mouse sensory epithelium (vestibular ampulla and combined cochlear/vestibular regions) (supplementary material Fig. S5).

Cd74 mRNA was expressed in the embryonic SAG and embryonic otocysts (supplementary material Fig. S5A, lanes 2-3). Cd74 was also expressed in the adult ear (supplementary material Fig. S5A), including SG neurons (Fig. 5A), as well as the adult sensory epithelium in the cochlea (supplementary material Fig. S5B). The presence of CD74 on SAG neurons supports the existence of a signaling pathway by which MIF influences SAG survival and outgrowth. The distribution of CD74 in sensory epithelial regions and adult SG neurons further supports diverse roles for MIF in inner ear development. The zebrafish equivalents of CD74, Iclp1 and Iclp2 (now called Cd74a and Cd74b), have been found on SAG neurons in zebrafish (Shen et al., 2012).

The in vitro and immunohistochemical data strongly suggest that MIF plays an important role in the development of the SAG and, perhaps, other inner ear structures, as in the zebrafish inner ear (Holmes et al., 2011; Shen et al., 2012).

To determine if MIF is necessary for inner ear development or function in vivo, we analyzed the hearing thresholds and inner ear structures of Mif KO mice, which have previously been used to study the role of MIF in the immune system (Bozza et al., 1999; Leech et al., 2003).

Auditory brainstem response (ABR) measurements were taken at 4 weeks of age, the earliest age for ABR testing, from WT and Mif KO mice (Table 1). Mif KO mice had statistically significant elevated thresholds at all frequencies compared with WT mice. Hearing loss in Mif KO mice was clearly evident at 24 kHz and even more pronounced at 48 kHz.

To identify morphological changes in inner ear structures of Mif KO mice, neural innervation patterns and cochlear HC patterning were examined histologically in whole-mount cochlear tissues from 4-week-old WT and Mif KO mice, the age when hearing loss was documented in the same animals. Fig. 5 shows an analysis of neurofilament and myosin VIIa (a sensory HC marker) expression in WT and Mif KO mice in the basal high-frequency (48 kHz) region. Although both IHCs and OHCs were phenotypically normal and present in the OC of WT mice, with dense nerve fiber innervation to the HC regions (Fig. 5A-C), nerve fiber density was decreased throughout all cochlear regions in Mif KO mice, with some areas demonstrating dramatic decreases. Additionally, OHCs were frequently absent in the Mif KO OC, suggesting that normal expression of MIF might also be needed for normal OHC patterning (Fig. 5D-F,H).

The decreased nerve fiber density in Mif KO mice supports the hypothesis that MIF is necessary for innervation of the developing inner ear. To explore innervation patterning in WT and Mif KO mice, we examined interactions between P3 WT SG neurons (from SG explants) with either WT or Mif KO OC explanted at P3, over a 7-day period in co-culture.

Within 24 hours, neurites extended from the SG explants and advanced towards and made contact with the WT OC (n=19). In those cultures in which individual contacts could be visualized, the average number of contacts at 24 hours was 130±10 (±s.e.m., n=3). Neurite contacts persisted and grew more profuse in WT co-cultures over the 7 days of the experiment, by the end of which the contacts became too numerous to discern individually (Fig. 6).

By contrast, OC from Mif KO mice did not attract WT SG neurites. Only very short, radially distributed processes extended from these SG, but none of these processes ever made contact with the Mif KO OC, even after 7 days in culture (n=17). These results support the hypothesis that the developing inner ear sensory epithelium releases MIF to attract emerging neurites and to support the process of neural innervation. This also raises the question of whether MIF might have a therapeutic application in instigating regeneration events, such as the regrowth of SG neuronal processes to the OC in cases of injury or disease.

‘Classical’ nerve growth factors such as the neurotrophins, fibroblast growth factors (FGFs) and glial cell line derived neurotrophic factor (GDNF) have been found to play specific, stage-dependent roles in inner ear development (reviewed by Bianchi and Fuchs, 2010; Sánchez-Calderon et al., 2007). However, none of these previously identified factors was found to influence early stages of SAG development, at the time of initial neurite outgrowth (Bianchi and Cohan, 1993); nor were they found in proteomic analyses of ODF from any source (chick, mouse, IMO cell lines).

Despite numerous studies and a variety of experimental approaches in many model systems, the composition of ODF has remained unclear for many years. Only recently have our data suggested a role for immune system cytokines (Bianchi et al., 2005; Holmes et al., 2011; Shen et al., 2012).

In the present study, using various complementary methods we demonstrate that the cytokine MIF, which is expressed in inner ear SCs, is present in mammalian and chick ODF, that it influences SAG neurite outgrowth and neuronal survival at the stage of initial neurite outgrowth, and is necessary for the development of normal hearing in mice.

Proteomic analyses demonstrate that MIF is present in the CM from both IMO and otocyst CMODF. It is also presumably released by the WT mouse OC, in contrast to the Mif KO OC (Fig. 6). Although IMO cell line or otocyst CMODF was able to promote outgrowth and survival of SAG neurons, bioactivity was largely lost when ODF was treated with anti-MIF antibodies. However, the bioactivity of ODF appears to rely on the combined cytokine activities of MIF and MCP1; residual bioactivity in anti-MIF-treated ODF was further decreased – essentially removed altogether – when subsequently treated with anti-MCP1 antibody (supplementary material Fig. S1). Although there are other cytokines in ODF (Bianchi et al., 2005), including RANTES (also known as CCL5) and TNF-α, they appear to play a minimal role in its bioactivity. Future studies of synergies among various cytokines should provide additional understanding of how cytokines interact to regulate inner ear development. Several lines of evidence indicate that cytokines are potentially interactive, even synergistic at specific concentrations (Wu and Bradshaw, 1996; Kavanaugh, 2002). This is supported, for example, by our observation that MIF and MCP1 interact in a yeast two-hybrid study (data not shown).

The results of the current study are also consistent with the findings of our recent studies in zebrafish demonstrating that inactivating MIF, either by creating MIF morphants at the 1- to 8-cell stage (Shen et al., 2012), by injecting the MOs directly into the developing inner ear at 20 hpf (Holmes et al., 2011), or by chemically inhibiting the action of MIF protein with a specific inhibitor, resulted in decreased survival of SAG neurons (Holmes et al., 2011; Shen et al., 2012). Furthermore, the size of the SAG was at least partially restored when the 1- to 8-cell stage embryos were injected concomitantly with capped Mif RNA (mif and mif-like in combination – there are two Mif genes in zebrafish), further supporting a direct role for MIF in mediating SAG survival (Shen et al., 2012).

In both the current study of mouse and chick and in the previous analysis of zebrafish, the MIF receptor CD74 (Iclp1 and Iclp2 in zebrafish) was found on SAG neurons. Although it is not the only receptor for MIF (Roger et al., 2001), CD74 is the best-characterized MIF receptor in the immune system. Blocking SAG CD74 with siRNA reduced both neurite outgrowth and neuronal survival (supplementary material Fig. S2). Its presence on SAG neurons supports the involvement of a MIF signaling pathway in neuronal survival. MIF appears to support SAG survival by inhibiting p53-dependent apoptosis in the immune system (Leech et al., 2003), as we demonstrated in zebrafish SAG (Shen et al., 2012).

The present study demonstrated a role for MIF not only in vitro but also in vivo, as shown by analysis of the Mif KO mouse. Mif KO mice exhibited hearing loss across all frequencies, particularly in higher frequency (48 kHz) regions of the cochlea, providing clear evidence of the functional necessity for MIF in inner ear development. The Mif KO mice also had HC losses, particularly in the 48 kHz region of the cochlea (Fig. 5), as well as a prominent decrease in the density of nerve fiber innervation in the high-frequency region of the cochlea by 4 weeks of age.

MIF protein is expressed in both embryonic and adult inner ear, particularly in the SCs in the mature inner ear. Similar patterns of expression were noted in zebrafish (Shen et al., 2012). This localization of MIF to SCs rather than the sensory HCs suggests that MIF might play a role as a general guidance factor, attracting SAG neurites to the area of the sensory epithelium.

The attractant properties of MIF were observed in cultures of postnatal mouse SG neurons that exhibited directional neurite outgrowth towards the WT, but not the Mif KO, OC. The lack of growth towards tissue lacking MIF further supports the hypothesis that a MIF source (presumably SCs) acts to attract SG nerve fibers.

The observation that MIF is expressed in SCs that underlie sensory HCs and that significant numbers of HCs are missing in Mif KO mice, particularly in high-frequency regions, might also suggest that MIF expression is required to maintain sensory HCs or perhaps the relationship between HCs and SCs. HCs are also missing in Mcp1 KO animals (data not shown), again with the greatest loss from the high-frequency region of the cochlea.

The present study also demonstrates that, whereas MIF is expressed in SCs rather than HCs, MCP1 is localized in the sensory HCs themselves (Fig. 4). Although our previous work demonstrated that MCP1 was necessary, but not sufficient, for SAG outgrowth (Bianchi et al., 2005), its presence in HCs provides additional evidence that it too is likely to play a vital role in mediating inner ear development and innervation. Further evidence that MCP1 is an important contributor to ODF was provided by the second series of experiments using function-blocking antibodies, in which we demonstrated that residual ODF bioactivity after anti-MIF antibody treatment was lost upon addition of function-blocking antibodies to MCP1 (supplementary material Fig. S1).

Our finding that immune system ‘inflammatory’ cytokines such as MIF and MCP1 behave as neurotrophic factors in early inner ear development further suggests that, evolutionarily and developmentally, such cytokines are multifaceted and pleiotropic. It is also clear that a requirement for MIF in development is not limited to the inner ear. For example, the present study showed that mESCs take on neuronal characteristics in the presence of MIF, and that this neuronal induction is similar to the neuronal phenotypes induced by neural growth factors NGF and CNTF, as our previous studies demonstrate (Roth et al., 2007; Roth et al., 2008). Thus, MIF might play multiple roles throughout the developing nervous system, particularly at early stages of development (Suzuki et al., 2004). Such a role is supported by evidence from several earlier studies. For example, cytokines, including MIF, have been found in the developing CNS, eye and inner ear of Xenopus (Suzuki et al., 2004) and mouse (Kobayashi et al., 1999) and in the zebrafish eye (Ito et al., 2008) and ear (Holmes et al., 2011; Shen et al., 2012). MOs injected into one-cell Xenopus embryos resulted in the absence of a neuraxis (Suzuki et al., 2004), a phenotype that was rescued by concomitant injection of Mif RNAs. Although such a drastic effect is not seen in zebrafish treated simultaneously with the two (mif and mif-like) MOs (Shen et al., 2012), considerable cell death was evident in both the developing otocyst and the CNS; this phenotype could be rescued by injection of capped mif and mif-like RNAs (Shen et al., 2012). MIF also appears to be crucial in regulating cell death in the CNS (Chalimoniuk et al., 2006).

Other cytokines/chemokines appear to regulate axonal outgrowth by inhibiting the repellent effects of molecules such as semaphorins and Slit2 (Chalasani et al., 2003). Additionally, some cytokines/chemokines are essential players in neuronal repair processes after axotomy in both the peripheral nervous system and CNS (Deverman and Patterson, 2009, Bajetto et al., 2001; Koda et al., 2004). Thus, cytokines/chemokines have multiple roles throughout various developmental stages of nervous system development and repair.

Additional cytokines might be important for inner ear development. For example, RANTES/CCL1 (Bolin et al., 1998) and tumor necrosis factor (TNF) (Bianchi et al., 2005), as well as the receptors CCR1 (for RANTES) and CXCR4/2 (for MIF) (Bernhagen et al., 2007) and TNFR1/2 (also known as TNFRSF1A/B) (Zou et al., 2005), are found in the inner ear. Another MIF receptor, CD44, is expressed in pillar cells (Hertzano et al., 2010) in developing and adult inner ear tissues. The function of these various cytokines and cytokine-driven signaling systems has yet to be fully explored, and any potential interactions between them still need to be addressed. Determining the exact mechanisms by which cytokines/chemokines and their receptors regulate inner ear development and innervation might find applications in cochlear repair or regeneration (Nishihira, 2000; Morand, 2005), given that MIF receptors (CD74) are found on adult SG neurons.

We thank Kristina M. Brooks and Sarah Tomkovich for technical assistance.