A kinase inhibitor screen identifies signaling pathways regulating mucosal growth during otitis media

Otitis media, the most common disease of childhood, is characterized by extensive changes in the morphology of the middle ear cavity. This includes hyperplasia of the mucosa that lines the tympanic cavity, from a simple monolayer of squamous epithelium into a greatly thickened, respiratory-type mucosa. The processes that control this response, which is critical to otitis media pathogenesis and recovery, are incompletely understood. Given the central role of protein phosphorylation in most intracellular processes, including cell proliferation and differentiation, we screened a library of kinase inhibitors targeting members of all the major families in the kinome for their ability to influence the growth of middle ear mucosal explants in vitro. Of the 160 inhibitors, 30 were found to inhibit mucosal growth, while two inhibitors enhanced tissue proliferation. The results suggest that the regulation of infection-mediated tissue growth in the ME mucosa involves multiple cellular processes that span the kinome. While some of the pathways and processes identified have been previously implicated in mucosa hyperplasia others are novel. The results were used to generate a global model of growth regulation by kinase pathways. The potential for therapeutic applications of the results are discussed.


Introduction
Otitis media (OM) is the most common childhood disease, affecting more than 90% of children under five years of age [1]. OM causes more physician visits, prescriptions, and surgeries than any other pediatric condition. The estimated annual cost of OM in the US is $6 billion in health care and associated expenses [2,3]. The peak of OM incidence also occurs at a critical developmental period. Chronic OM has been associated with delays in speech and language and learning [4][5][6]. In many developing countries, OM is a very serious health issue. The

Animals
For in vitro studies, subjects were 60-90 day old Sprague Dawley rats (Harlan). For gene array studies, subjects were C57/WB F1 mice (Jackson Labs). Animals were housed under specific pathogen-free conditions. All experiments were performed to National Institutes of Health guidelines and approved by the Institutional Animal Care and Use Committee of the VA San Diego Medical Center.

Bacteria
Haemophilus influenzae strain 3655 (nontypeable, biotype II), which had been = isolated from the ME of an OM patient in St. Louis, was provided by Dr. Asa Melhus, Lund University, Sweden, was used at titers of 5x102-10 3 to induce infection of the ME. Inocula were prepared as described earlier [23,24].

Surgery
All surgeries were performed under standard aseptic conditions. Animals were deeply anesthetized with rodent cocktail (10 mg/kg xylazine, 1 mg/kg acepromazine and 50 mg/kg ketamine i.m.). For in vitro studies, and approximately 50 μl of NTHi inoculum was bilaterally injected into the ME of rats via the surgically exposed ME bulla, after which the incision was closed. Following inoculation, the tympanic membranes were visually confirmed to be intact. After 48 hours, rats were anesthetized and humanely sacrificed. The ME bullae were harvested and opened. The ME mucosa was carefully dissected from the hypo-tympanum as a single sheet. The tissue was placed in a sterile petri dish with a layer of black silastic rubber on the bottom, in PBS. A diamond scalpel was used to cut the mucosa into approximately 0.25 x 0.25 mm explants.

Kinase inhibitor libraries
The kinase inhibitory compounds used (EMD Calbiochem Kinase Inhibitor Libraries) were contained in two separate libraries (I and II), each consisting of 80 inhibitory compounds. The libraries broadly targeted elements of the mammalian kinome, with inhibitors against all of the major kinase families, as shown in Fig 1. The libraries were provided in two separate sets of 96-well plates (I and II), and all compounds were provided in DMSO. Each library set included both DMSO and blank wells for controls.

Tissue culture
The explants were placed individually, lumen-side up, into 48-well culture plates (Co-Star). A 150 μl of culture medium (75% DMEM, 25% Hams-F12 supplemented with 5% bovine serum, with additives: 100 IU/mL penicillin, 100 mg/mL streptomycin, 0.4 mg/mL hydrocortisone and 10 −6 M isoproterenol) was carefully added so that the fluid meniscus kept the explant flat on the well bottom. The explants were cultured for 24 hours to allow for attachment. The culture medium was then increased to 250 μl and the explants cultured for two more days to facilitate adhesion and expansion. After, the mucosal explants were randomly divided into groups of three explants per concentration, and three concentrations of each kinase inhibitor were applied. The 160 kinase inhibitors from libraries I and II were diluted in warm culture media and added to explant wells at concentrations of 0.05, 0.5, or 5.0 μM. For each plate, one group of 3-6 explants/plate was cultured without an inhibitor but with DMSO (0.1% concentration, as used for all inhibitors after dilution) to serve as negative controls. Kinase inhibitors were applied from 3-9 days in vitro (DIV) for a total of six days. Each explant was then fixed and photographed, and the 9 DIV area of mucosa growth and proliferation was measured using NIH SPOT Image software.

Data analysis
Preliminary studies were performed using different numbers of control explants and explants treated with 0.5 μM of the JNK inhibitor SP000126, which we had previously found to reduce explant growth [20]. These prior studies established that approximately 8-10 explants would be required to identify a statistically significant difference due to a given treatment. As this number was not practical for a screen, we used an N of three and applied threshold measures to select compound hits. For mucosal samples treated with 0.5 μM of an inhibitor, a criterion threshold of <10 mm 2 after 9 DIV was used. For 5.0 μM, a threshold of <5 mm 2 was applied. A criterion threshold of >45 mm 2 was used to identify compounds that increased mucosal growth in culture.

Gene arrays
To assess the expression of Hippo pathway genes during OM, we employed gene arrays in a mouse model of acute ME infection, as described previously [25]. Briefly, the MEs of C57/WB F1 mice were inoculated with nontypeable Haemophilus influenzae (NTHi) and ME tissue harvested from deeply anesthetized (rodent cocktail) animals at 0 hours (0h, without infection), 3 hours (3 h), 6 h, 1 day (1d), 2d, 3d, 5d or 7d. mRNA from 20 mice at each time point was reverse transcribed and biotinylated cRNA probes were hybridized to duplicate Affymetrix MU430 2.0 microarray sets. The arrays contain~55,000 probes against virtually all murine genes and including many splice variants. An additional 20 mice at each time point was used to generate a biological replicate. Gene expression levels were evaluated for differences due to infection using variance-modeled posterior inference (VAMPIRE), which distinguishes signal from noise by modeling error structure and identifying expression-related and expressionunrelated variance [26]. Based on the results obtained with inhibitors, expression of genes related to the Hippo pathway in uninfected MEs was compared to that in NTHi-infected MEs at each time point.

Ethical approval
All applicable international, national, and/or institutional guidelines for the care and use of animals were followed. All experiments were performed to National Institutes of Health guidelines and approved by the Institutional Animal Care and Use Committee of the VA San Diego Medical Center.

Controls
When combined across all library plates, negative control explants not treated with any inhibitors exhibited an average area of 25.42 ±0.92 mm 2 after 9 DIV. Positive control explants, treated with 0.5 μM of the JNK inhibitor SP000126 from 3 DIV to 9 DIV averaged 8.86 ±0.92 mm 2 .

Library I
Representative cultures illustrating a control explant and explants treated with two doses of compound 1-26 (an EGFR Inhibitor) from Kinase Inhibitor Library I are presented in     Table 1

Discussion
We developed an in vitro screening method to identify compounds with the potential to influenced ME mucosal hyperplasia which is induced by bacterial infection. Out of 160 kinase inhibitors screened in triplicate, 38 compounds demonstrated an inhibitory effect on mucosal explants growth, while 2 compounds enhanced growth. The large number of kinase inhibitors that influenced explant growth testify to the complex regulation of ME mucosal hyperplasia by this important group of cellular mediators [27,28].

Inhibitors reducing mucosal growth
Growth factor receptor inhibitors. Inhibitors of both epithelial growth factor (EGF) and platelet derived growth factor (PDGF) receptors reduced mucosal growth. The fact that inhibitors of growth factor receptors were identified in the screen was expected and consistent with prior data. This is particularly the case for EGFR inhibitors, given that the mucosa is an epithelial structure [29], and we have previously shown that heparin-binding EGF (HB-EGF) strongly stimulates the growth of ME epithelial cells in culture [30]. Moreover, the expression of HB-EGF in the ME increases by 26-fold and EGFR expression by 2.5-fold, during OM [25]. Similarly, stromal cells within the explants can be expected to express PDGF receptors, although it has also been found that epithelial cells can respond to injury by up-regulating PDGFRs [31]. During OM, ME PDGFA expression rises 7-fold, and PDGFR expression increases 6-fold [25]. The FBS present in the culture media would supply both EGFs and PDGFs, and production by cells of the explant is also possible. It should also be noted that both EGFRs and PDGFRs can be activated independent of ligands, by reactive oxygen species [32]. Since reactive oxygen species is often released from tissue in culture [33], this provides another potential mode of receptor activation.

PI3K/AKT/MTOR inhibitors.
Inhibitors of protein kinase B (AKT), phosphatydlinositol 3 kinase (PI3K) and mammalian target of rapamycin (MTOR) reduced explant growth, which was also an expected result. PI3K activation of AKT occurs downstream from the action of many growth factor receptors, and has long been associated with cell proliferation [28]. Substrates of PI3K/AKT signaling are known to be involved in OM. For example, the Class I PI3K catalytic MTOR subunit genes pik3cd and pik3cg are upregulated 8.5-fold and 2.5-fold, respectively, while the expression of akt1 increases 3.5-fold, and that of akt2 and akt3 increases 2-fold, during OM [25]. A different AKT inhibitor was found to reduce the growth of ME mucosal explants by Lee et al. [34]. Interestingly, these authors also found that an inhibitor of MTOR, which typically opposes PI3K, also reduced mucosal growth. This suggests an atypical role for MTOR in the ME mucosa.
Cell cycle progression inhibitors. Of course, it is obvious that cell division is involved in ME mucosal proliferation. It was, therefore, no surprise that several cyclin-dependent kinase (CDK) inhibitors of the cell cycle reduced the growth of ME explants. Similarly, aurora kinases are essential for cell division via the regulation of mitosis, especially the process of chromosomal segregation. Thus, suppression of mucosal growth by an aurora kinase A inhibitor was not unexpected.
JNK inhibitor. The mitogen-activated protein kinases (MAPKs) are often associated with tissue growth, as in cancer. Activation of Jun kinase (JNK), a cell stress protein, can have many effects, from cell proliferation to apoptosis [35]. During OM, while genes encoding JNK isoforms are only modestly (<2-fold) or not upregulated, expression of the genes encoding the JNK targets cJun and JunB increase 9-fold and 12.6-fold, respectively [25]. JNK inhibition by SP00165 has previously been shown to reduce mucosal growth both in vitro and in vivo [20], which is why it was chosen as a positive control for this screen. A study in transgenic mice found that lack of the jnk1 or jnk2 gene altered the course of bacterial OM [36].
NFκB inhibitors. Nuclear factor kappa-B (NF-κB) is a dimeric transcription factor that regulates immune and inflammatory genes in response to infection [37,38]. It also promotes cell proliferation and survival, in part by altering cell cycle regulation (e.g. [39,40]). Activation of NF-κB is tightly regulated, and inappropriate activation of the NF-κB has been linked to autoimmunity, chronic inflammation and cancer [41].
The canonical NF-κB activation pathway, which can be initiated by activated growth factor receptors via PI3K and AKT, relies upon the phosphorylation of IκBα, mediated by a heterotrimer of IκB kinases (IKKs), IKKα, IKKβ and IKKγ [42]. This initiates IκB cleavage of NF-κB1 of a RelA/ NF-κB1 dimer, releasing RelA/p50 to activate expression of genes driving inflammation and tissue growth. A non-canonical NF-κB pathway is activated primarily by members of the TNF receptor family. It activates NF-κB1 independently of the IKK complex to induce persistent activation of RelB/p52. Genes encoding NF-κB and its signaling molecules are strongly upregulated during OM, and inhibition of either the canonical or non-canonical pathways reduces ME mucosal growth in vitro [43]. Thus reduction of growth by NF-κB inhibition was expected.
PKR inhibitor. An inhibitor of protein kinase R (PKR) reduced explant growth. PKR, a multifunctional kinase best known for being activated by viral double-stranded RNA, can also be activated by the eukaryotic protein PRKRA, and by heparin [44]. Activation of PKR suppresses protein synthesis, and it is known to be involved in the regulation of tissue proliferation via complex interactions with NFκB and the JNK and p38 MAPKs [45]. The gene encoding PKR is modestly (1.5-2X) upregulated during OM, but the gene of its activator PRKRA is strongly suppressed (to 0.4X) within hours after ME infection [25]. PKR has not previously been linked to OM. CAMKII inhibitor. An inhibitor of Ca ++ /calmodulin-dependent protein kinase II (CAM-KII) reduced mucosal growth. CAMKIIA is best known for regulating memory and long-term potentiation in the brain [46]. However, CamKII members, especially CAMKIID, have been found to be upregulated in cancers [47], potentially linking them to tissue growth. During OM, while the Camk2a gene is strongly (0.15X) down-regulated, Camk2d is up-regulated 3.4X [25]. CamkIIs have not previously been linked to OM. JAK3 inhibitor. JAK3 is best known for its involvement in signaling via type 1 cytokine receptors, and for expression in leukocytes [48]. However, JAK3 is expressed in a variety of nonlymphoid tissues [49], and several cytokines can be produced by epithelial cells [50]. While it has not been linked previously to ME mucosal hyperplasia, JAK3 has been shown to regulate lung epithelial cell proliferation [51].
Inhibitors increasing mucosal growth MEK inhibitor II. Library II inhibitor 68, MEK Inhibitor II, significantly enhanced the growth of ME mucosa explants at 0.5 μM. MAPK kinase (MEK) is best known for downstream activation of the ERK MAPK, which is a potent mediator of tissue growth and cell survival [52]. While the erk1 and erk1 genes are upregulated only 2-fold during OM, the gene encoding the ERK upstream activator, TPL2, is upregulated more than 35-fold [25]. It has previously been shown that MEK/ERK inhibition by a different inhibitor, U0126, reduces the growth of ME mucosal explants [17]. Therefore, the fact that MEK Inhibitor II increased ME mucosal growth was unexpected. However, while U1026 is selective for MEK1 AND MEK2, it has been shown that MEK Inhibitor II also inhibits TPL2 [53], an upstream activator of several MAPKs and a tumor growth suppressor gene, at a lower IC50 than its inhibition of MEK, which is limited to MEK1 at an IC50 well above 0.5 μM. Inhibition of the growth suppression activity of TPL2 would be consistent with the effects of MEK Inhibitor II on explant growth. Two other MEK inhibitors had no effect on explant growth.
H-89 hydrochloride. Library II inhibitor 81, H-89 hydrochloride, a potent inhibitor of protein kinase A (PKA), also enhanced explant growth. PKA has been linked to the Hippo signaling pathway, which regulates organ size. Hippo signaling stimulates cell proliferation at low cell densities by directly activating the transcription factor YAP. However, in conditions of cell crowding, the Hippo receptor FAT4 stimulates intermediary proteins including ENT1/2, MERTK, and MST1/2 to activate LATS1/2, which phosphorylates YAP and prevents it from entering the nucleus [54]. PKA is an alternative activator of LATS1/2 [55], and thus can downregulate cell proliferation even in the absence of cell crowding. As shown on Fig 6, a number of Hippo pathway genes are regulated during OM, suggesting the involvement of this pathway in limiting mucosal growth. The Hippo pathway has not previously been identified as having a role in ME mucosal hyperplasia.

Implications of kinase screen results for mechanisms of ME mucosal hyperplasia
Comparing the results of this screen with Fig 1 and Table 1, it is clear that kinases which participate in the regulation of ME mucosal growth span the breadth of mammalian kinases. Evaluating the targets of inhibitors with the strongest effects on mucosal growth in vitro, our results identify three primary kinome groups; the AGC, CMGC, and TK families. The AGC group contains many core intracellular signaling kinases such as PKA, PKG and PKC, after which the family is named, and AKT. AGC kinases are modulated by phospholipids, cyclic nucleotides, and calcium. In various cell types they are responsible for cellular metabolism, transcription regulation, immune responses and cell growth. The CMGC group (also named after the initials of some members) regulates primarily the cell cycle, MAPK signaling, and RNA splicing. Inhibitors identified to have a significant regulatory effect on ME mucosa tended to target Cdks, as well as the JNK and p38 MAPKs from this family. The TK family (the members of which phosphorylate primarily the tyrosine residues of target proteins) are perhaps the most studied kinase group. They have been implicated in a wide variety of cellular processes including cell cycle control, mitogenesis, tissue growth, and cell proliferation. Many growth factor receptors and oncogenes are the primary targets of TK family inhibitors in the ME mucosa.
This novel in vitro screening method identified a number of protein kinases and intracellular signaling pathways which are already known to regulate ME mucosal hyperplasia, as well as additional kinases and pathways not previously associated with mucosal growth. A schematic summary of the kinase pathways regulating tissue growth that were implicated in this screen is provided in Fig 7. This global model provides a rich source of molecules to be evaluated in vivo, to confirm any role in the disease process of OM and to evaluate potential therapeutics that could reduce OM pathogenesis.
Explants from infected MEs were used for this screen. We previously noted that explants harvested 48 hours after infection of the ME with NTHi grow significantly faster than explants from naïve MEs. Unlike in vivo OM, when the mucosa returns to quiescence after bacteria are cleared, this accelerated in vitro growth continues for the duration of culture up to 10 days Gene array analysis of the ME transcriptome reveals that many genes related to Hippo signaling are significantly upregulated beginning 1 day after bacterial inoculation of the ME. � = p< .05; �� = P < .01. The fold change values, ranges and significance levels are represented in S1 Table. https://doi.org/10.1371/journal.pone.0235634.g006 PLOS ONE [18,19]. The reason for the extended growth of previously infected mucosal explants is not clear. The reduction in explant growth by a JAK3 inhibitor suggests that interleukin expression continues in culture for previously infected explants, which may contribute to continuous growth. Similarly, the effects of GF receptor inhibitors indicate that these receptors are also constitutively expressed at high levels in culture and could contribute to prolonged growth. Finally, processes that operate during OM recovery in vivo presumably act to suppress interleukin, GF, GF receptor and other growth stimulant expression in mucosal cells. In culture, these suppressive mechanisms may be less active, leading to prolonged growth.

Advantages and disadvantages of the screening assay
In previous studies, we used ME mucosal explants to evaluate specific growth pathways suspected of mediating mucosal hyperplasia. These and other studies have identified several mechanisms with the potential to influence mucosal growth, including growth factors and downstream signaling pathways. The results of the present study validate the use of ME mucosa explants as a screening tool, with which to broadly evaluate the effects of pharmacological compounds on mucosal growth.
Screening assays have the advantage of being able to test the effects of much larger numbers of compounds than more detailed in vivo studies. This allows for the evaluation of compounds without a priori assumptions regarding the mechanism underlying mucosal growth, and permits the discovery of novel cellular mechanisms not predicted from current knowledge. In the case of this screen, several such previously unsuspected mechanisms were uncovered, and several novel inhibitors were identified. A throughput screening assay also provides comparative

PLOS ONE
information on the relative ability of compounds to inhibit growth. Thus, we found that some compounds with similar kinase targets displayed quite disparate inhibitory ability.
However, the nature of screening means that the assay used is constrained by necessity. For most screens, as in our case, in vitro methods are necessary to increase throughput. Dissection and tissue culture can alter the responses of cells and tissue. In addition, rodents are separated by millions of years of evolution from humans, and may exhibit significant differences. That said, murine and human MEs exhibit more similarities than differences. There have been many studies demonstrating that the ME mucosa from both species respond similarly to bacterial infection (e.g. [16,56]), while genetic defects that influence OM in the mouse often have similar effects on human disease (e.g. [57]).
Another issue for the assay is that only a limited number of conditions can be evaluated. ME mucosal explants cannot be produced in very large numbers, as is possible with cell lines. While we used three different concentrations of the screened compounds, dosages outside of this range might also have proved effective. This study was performed as a screen, and testing a larger number of dosages would have made the screen less practical. The effects of compounds that might require a longer pre-treatment period may have also been missed.
As noted above, several of the pathways identified in our screen have previously been evaluated using inhibitors or gene deletion models. These prior studies provide retrospective validation of these assay hits. However, the novel pathways identified require confirmation by more extensive studies and evaluation in vivo.

Potential clinical implications
The results of the present study have implications for the treatment of OM in patients. As noted above, OM is characterized by extensive hyperplasia of the ME mucosa. The transformation of the mucosa from a monolayer of simple squamous epithelium to a pseudostratified, cuboidal epithelium with multiple cell layers and the characteristics of a respiratory epithelium [16] can result in many deleterious sequelae. This includes the generation of mucus from goblet cells, the secretion of additional factors from secretory cells, and the growth of a supportive vasculature. Mucus can result in reduced ability to clear fluid from the ME cavity. In chronic OM, this can result in "glue ear" with mucus and cellular debris so dense that clearance is virtually impossible [58]. Neovascularization can enhance the entry of inflammatory leukocytes into the ME, as well as the extravasation of fluid into the tympanic cavity, resulting in cellular effusion.
We have identified pathways that are likely responsible for hyperplasia of the ME mucosa, as well as inhibitors that are capable of limiting its growth. These and similarly targeted inhibitors have the potential to reduce the above damaging sequelae of OM, and reduce pathogenesis. Most uncomplicated OM resolves without treatment over the course of days [8], and would not be targets for intervention. However, chronic OM could benefit from limiting mucosal growth. Obviously, hyperplasia of the mucosa, mucus generation and leukocyte infiltration can serve a protective function against bacterial infection of the ME. However, antibiotics are prescribed for chronic OM [11], and the response of the ME mucosa can be deleterious. In such cases, mucosal growth inhibitors may have a place alongside antimicrobial therapy.
Supporting information S1 Table. Fold change averages, ranges and probabilities for gene array data presented in