Development of a central nervous system axonal myelination assay for high throughput screening

Development of the embryonic cortical cell co-culture assay

In an effort to move a low-throughput, well-established myelination cell culture technique to a format suitable for higher throughput screening applications, we miniaturized and automated a previously described RGC-OPC co-culture technique [9]. Unfortunately, the low yield of RGCs and lack of assay robustness from each preparation makes this co-culture myelination assay unsuited for high-throughput compound screening (Additional file 1: Fig. S1A). We therefore sought another source of tissue where numbers of neurons would not be limiting.

For increased yield of primary neurons, we chose the cortex of embryonic day 18 (E18) rats as an abundant source of relatively homogeneous brain cells with well-established culture methods [5, 6] (see “Methods”). From one litter, enough cells can easily be generated for high throughput drug screening applications (~30 × 10⁶ cells/cortex; Additional file 1: Fig. S1B).

Following the differentiation and growth of neurons and glia for 5 days, we determined that the differentiation and early myelination of exogenously added OPCs proceeded optimally when the growth medium was switched from NB/N21 to an OPC-supporting myelination medium (MyM). With this growth medium, we observed that it was not necessary to add exogenous OPCs as we could readily identify mature and axon ensheathing OLs that had differentiated from the embryonic cortical preparation, presumably from neural precursor cells and/or OPCs. We actually found that the differentiating OPCs already present in the cultures produced better myelination (Additional file 2: Fig. S2, see below and methods for quantification of myelination). Finally, we determined that the optimal time course for myelination to proceed was 8 days after test compound addition and 13 DIV total (Additional file 3: Fig. S3). Figure 1 depicts the flow scheme of the embryonic cortical cell assay. At this early myelination time point, we observed MBP staining aligning with SMI 31/32 axon staining, indicating that indeed OLs are contacting and aligning with axons (Fig. 2a).

We utilized the γ-secretase inhibitor, DAPT, a known enhancer of myelination [9, 11] as a positive control to test our assay system and establish an automated morphology analysis. After compound treatment, cells were stained for the OL lineage marker, Olig2, myelin basic protein (MBP) to stain mature OLs, and the nuclear dye, DAPI, and imaged. Myelination was scored by quantifying the characteristic change of morphology of OLs when ensheathing axons—from many branched, flattened, and diffusely MBP stained processes to condensed and aligned MBP-positive fibers. For each high resolution 10× image, we quantified the total length of contiguous, aligned MBP staining (fiber length)/number of Olig2-positive (Olig2⁺) nuclei, referred to as myelination). Figure 2b demonstrates the digital mask created by our protocol used in the fiber length calculation. With these methods, we determined significant dose-dependent increases in myelination with DAPT (Fig. 2c). Importantly, we were able to determine reproducible EC₅₀ values of four GSI compounds, DAPT, LY411,575, BMS 708,163, and MRK560, allowing the ranking of compounds (Fig. 3a–d, Table 1). GSI-mediated facilitation of myelination was only observed in the presence of live axons and had no effect on the differentiation of purified OPCs grown in isolation (Additional file 4: Fig. S4). We tested two other compounds identified from published high throughput library screens that promote OL differentiation in cultures containing purified OPCs, benztropine and clemastine [2, 3]. As expected, these compounds demonstrated significant OL differentiation in our acutely prepared OL differentiation assay (Additional file 4: Fig. S4). However, in our cortical myelination assay, benztropine and clemastine did not promote myelination (Additional file 5: Fig. S5). This data demonstrates that the cortical myelination assay identifies novel compounds with myelination activity distinct from compounds that solely promote OL differentiation.

Known mechanism of action	Compound	EC50 μM avg b
Kinase inhibitor	Imatinib mesylate	1.4
Anti-cholinergic	Atracurium Besylate	5.3
Mitotic inhibitors	Docetaxel	0.1
	Methotrexate	0.1
	Tegafur	2.7
	Artesunate	3.3
Ion channels	Zu-capsaicin	4.7
	Amiloride	8.9
	Oxcarbazepine	14.7
Na2+/K+ ATPase inhibitor	Digoxin	11.3
γ-secretase inhibitors	LY 411, 575a	0.00053
	BMS 708, 163a	0.067
	MRK 560a	0.082
	DAPTa	0.55

^aCompounds not part of the NCC library screen

^bN = ≥ 2

Long-term characterization of cortical myelination cultures

To test whether the enhanced early myelination effects of GSIs had longer lasting effects with the single 8 day drug treatment course, cortical cultures were treated as described followed by 2 weeks of half medium changes with fresh MyM without GSI compound. These GSI-treated cultures demonstrated robust MBP alignment compared to DMSO vehicle controls (Fig. 4a, b). Additionally, we tested these cortical cultures for their ability to initiate the formation of axonal nodes of Ranvier, essential to action potential propagation in functionally myelinated axons. As an indicator of early node formation, we immunostained these longer term cultures with antibodies to the paranode-localized protein, Caspr [12] along with anti-MBP antibodies. Figure 4c demonstrates the accumulation of Caspr protein at the edges of myelinated axon segments indicating that the initiation of node formation was induced by contact with OL myelin. These data demonstrate the ability of cortical cultures to form robustly myelinated axon segments and initiate node formation which is enhanced by an early, single dose treatment with GSIs.

Cellular composition of cortical myelination cultures

To determine the composition of these cortical cultures, we used well-established cell type marker antibodies to identify different cell populations. Cultures were grown using the culture conditions described above, and fixed on DIV13. All cultures were stained with the nuclear marker DAPI to identify the total population of all the cells in culture. To identify the neuronal population, we used anti-NeuN antibodies to identify neuronal nuclei, as well as anti-MAP2 and anti-SMI 31/32 neurofilament antibodies to assess the health and extent of dendrite and axon formation, respectively. Imaging of neurons in these cultures demonstrated mature cortical neurons with well-developed dendrites and a dense bed of axons (Additional file 6: Fig. S6). In addition to NeuN for the identification of neurons, we also used anti-Olig2 antibodies to identify OPC/OLs and anti-GFAP antibodies to identify astrocytes. We then quantified the percentage of each of these cell types in this cortical co-culture preparation as a percentage of the total cell population identified with DAPI nuclear staining of all cells. We found that the cell composition under these culture conditions was 23 % neurons, 46 % astrocytes, 22 % OPCs/OLs, and 9 % unidentified cells.

In order to better understand how OLs differentiate and develop in these cultures, we stained and imaged DIV5 cultures to assess the cell composition of our cultures on the day of test compound addition. At this stage, the cultures contained ~50 % neurons, having already generated an axon network (Additional file 7: Fig. S7). Since our cultures were derived from embryonic cortex, we used the bi-potent O2A glial progenitor antibody marker A2B5 [13, 14], to identify glial progenitors still capable of differentiating. DIV5 cultures contained abundant A2B5 positive cells which were not observed at DIV13 (Additional file 8: Fig. S8). There were relatively few astrocytes (positively staining for GFAP) and differentiated OLs (positively staining for MBP, CNP, O4 or MOG) at this stage, indicating that a majority of the OL differentiation and myelination occurred during the test compound treatment window (Additional file 9: Fig. S9). Using the microglial marker Iba1, we detected <1 % microglial cells at DIV5, and undetectable microglia at DIV13.

NIH clinical collection library screening for compounds that promote myelination

To demonstrate that the assay conditions we developed were robust enough to support drug discovery screening efforts, we screened a small library of compounds. We selected the NIH Clinical Collection (NCC) library for screening which contains Food and Drug Administration (FDA)-approved off-patent drugs. Therefore, hits retrieved from this collection could lead to potential drug candidates for further development and rapid repositioning as therapeutics for MS. The NCC library consists of 727 biologically active compounds that have been through phase I–III clinical trials. This collection is additionally attractive because of the wide variety of cellular targets that are represented. Because this focused FDA-approved compound collection is small and the drug structures diverse, we decided to screen at two concentrations (5 and 1 μM) to reduce the possibility of missing hits due to false negatives. Each plate contained eight wells treated with DMSO or DAPT controls and each test compound concentration was screened in duplicate.

Automated image acquisition was performed from four randomized fields from each well, representing a total of eight data points per test concentration. The data was analyzed to find active compounds that increase myelin formation above a pre-defined threshold (>50 % of DAPT pro-myelinating activity). Though the threshold is somewhat arbitrary, we included in our analysis the delineation of three SDs above the mean signal of DMSO-treated well (Fig. 5a). While not a criterion in our assay for hit selection, it provided a statistical assurance that we were well out of a false-positive hit rate (0.15 %) range. Control DAPT versus DMSO values from the entire myelination screen were highly statistically significant (Fig. 5b) indicating an acceptable screen window.

We were concerned about the possibility of inter-preparation variability and differences in the behaviors of dissociated primary neurons and glia in culture, so we considered how to evaluate the consistency of responses to treatment in addition to assessing the consistency of the plate controls. As a measure of assay quality, we considered the Z-factor; however, since our assay is based on multiple readouts, we incorporated an additional parameter for assay window and robustness, a specific morphological measurement of early myelination. This criterion is a quotient of two morphological measurements—fiber length intensity score and MBP intensity score which adjusts for the contribution of OL differentiation (MBP expression) (Additional file 10: Fig. S10). A ratio of one indicates that an observed increase in myelination may almost be entirely accounted for by an increase in the extent of OL differentiation, whereas a value significantly greater than one indicates that there is an observed increase in myelination above and beyond what would be expected by an increase in OL differentiation alone, i.e. specific induction of myelination. We primarily used this to measure the quality of each plate in the screen. We typically observed DAPT treated cultures to have fiber score/MBP scores >1.5. Figure 5c depicts the DAPT/DMSO fiber score/MBP scores of each plate from the entire library screen which generated an average fiber score/MBP score of 1.61. For each plate in the screen, the acceptable fiber score/MBP score cutoff was in the range of 1.3 ± 0.2. In addition, the fiber score/MBP score was incorporated into the criteria for assay hits to potentially distinguish between active compounds with distinct mechanisms of action (see below). Additional file 12: Table S1 contains the data set for the entire library screen, showing the % DAPT normalized scores for myelination for each library compound.

Selection criteria for myelination assay positive hit compounds

To select candidate compounds with substantial activity in our screen, compounds that had scores >50 % of DAPT were included as secondary hits (Fig. 5d). In addition, to further delineate the myelination effect enhanced by compounds, we implemented a second criterion of fiber score/MBP score (Additional file 10: Fig. S10) including any compounds that had scores >1.5 (fiber score/MBP scores for entire data set are included in Additional file 12: Table S1). Three compounds passed this criterion alone and were included in further hit compound refinement. We observed that some compounds displayed unusually large myelination scores but had very low overall Olig2⁺ nuclei expression, which could either reflect an inhibition of OL proliferation or increased OL cell death (e.g. methotrexate, Additional file 11: Fig. S11B7). Not surprisingly, visual assessment of images from these cultures revealed a low overall number of myelinating OLs. We therefore incorporated a third criterion to eliminate compounds that primarily act as anti-proliferative compounds and thus greatly reduced Olig2 expression: the ratio of total nuclei (DAPI staining)/Olig2⁺ nuclei. Large DAPI/Olig2⁺ nuclei numbers (>40) were an indicator that the test compound severely depleted OPCs and OLs, undesirable in a screen for compounds that promote myelination. DAPT reduces the number of Olig2⁺ cells by ~50 %, most likely by promoting OPC differentiation and reducing OPC proliferation [9]. Therefore, we implemented a criterion of >25 % of the DAPT Olig2⁺ cell count which also effectively eliminated compounds that severely reduced the number of Olig2⁺ cells (Fig. 5d). A fourth criterion was the qualitative assessment of OL MBP staining, taking into account the number of OLs/image field and OL morphology. Compounds that dramatically changed OL morphology (e.g. greatly enlarging the cell) while reducing the number of OLs/field were eliminated. Active compounds that passed all of these criteria were referred to as refined hits (Fig. 5d). A fifth criterion was to confirm activity and potency of refined hit compounds with full dose–response curve experiments of at least two replicates using reordered or resynthesized material. Actives that met this criterion were referred to as our confirmed hits (Fig. 5d) and our hit rate is based on this number. In a screen of 727 FDA-approved drugs, our screen identified 53 primary hits, 33 refined secondary hits and ten confirmed and reproducible hits (Table 1). The resulting hit rate for the entire screen was ~1.7 %. Additional file 11: Figure S11 shows the chemical structure of each hit compound, screening image of MBP/Olig2/DAPI staining, and the EC₅₀ curves for myelination. Table 1 shows the calculated myelination EC₅₀ values for the top hits from our cortical myelination screen. Based on the available literature on these previously characterized compounds, we grouped the hits based on the known mechanisms of action. These compounds fell into many different classes, grouped in Table 1, and are distinct from compounds previously identified by other library screens that have used OL differentiation assays in the absence of axons [2–4]. These data represent an initial screen to identify myelination-promoting compounds. Future studies will include an investigation of compound mechanisms as well as a thorough characterization of how these compounds effect cellular composition of the cultures.