I, along with Nina, performed our first high-throughput screen – well really medium-throughput but high-throughput for us – on Niemann-Pick Type C (NPC) patient-derived fibroblasts a couple of months back. We wanted to do a pilot screen with the Microsource bioactives library that we purchased recently. This was to find an alternative hit for NPC, a backup for our lead compound PERL101. The new library consists of around 2500 compounds and has a collection of FDA approved drugs as well as known biologically active compounds. In addition, we wanted to collect drug screening data on diseased cells to compare to data collected from the other NPC model organism screens. It will help us determine which compounds are toxic which compounds have NPC-relevant phenotypes in mammalian cells, or which compounds cause false positive phenotypes in patient cells, and more.
We set up the assay to work in 384-well plates. We currently do not have the equipment to properly fix and stain 1536-well plates, so for now, 384-well plates are optimal for our screens.
Figure 1. Cell density gradient in a sample 384-well plate
Since this was our first time working with 384-well plates in a mammalian cell screen, we started with optimizing cell density in the plate. I seeded a gradient of cell densities varying from 125 cells/well to 3000 cells/ well. Figure 1 shows an image of the density gradient used in a Greiner plate. I used NPC I1061T fibroblasts and stained them with filipin (which binds to unesterified cholesterol, as described here in a previous post by Nina) to assess the cell density to be used for 24-hour as well as 48-hour incubation time points. Our assay was to run for 48 hours, so we narrowed down the appropriate cell density to 675 cells/well or 750 cells/well and went ahead with the higher of the two densities.
Now that the cell density was worked out for a 48-hour incubation, we moved on to incubating cells with compound. The Echo 550 was used for dispensing the compounds into destination plates. Since the compounds are at a 10mM stock concentration and the total volume in the wells needs to be 50uL, we used the Echo to dispense 50 nL of each compound into 384-well black-walled glass plates. The first two and last two columns of each plate contained positive and negative controls, respectively. To know more about Echo 550, its uses and applications check out my first post here!
After compound was dispensed, the destination plates were ready to be seeded with the cell suspension. We maintained NPC I1061T fibroblasts culture in 10 cm plates. The cells were trypsinized, counted using a hemocytometer and then diluted into a cell suspension of the appropriate density. The cells were seeded into 384-well plates using a multichannel pipette. The plates were then incubated for 48 hours. However we are now using the BioTek Multiflo to seed 384-well plates since it is significantly faster and more consistent than manual seeding.
After the incubation period of 48 hours, the plates were fixed and stained. The stains used were filipin and sytox green, a commonly used nuclear stain. The fixing and staining were done with the help of a BioTek Precision. The process could be done only in batches of four plates so we staggered the protocol. We fixed the cells in 4% paraformaldehyde for 10 minutess and washed three times with 1X PBS. The cells were stained with filipin and incubated for 1 hour during which the other batch of plates were fixed, washed and stained with filipin. After 1 hour, the plates were washed twice with a cholesterol assay wash buffer and then stained with sytox green. The plates were then sealed with a clear cover and were ready to be imaged.
The plates were imaged with ImageXpress Micro XLS System from Molecular Devices using a 10x objective. We took multiple images per well. The plates were then analyzed using MetaXpress Custom Module Editor (CME). The area of the puncta, i.e., filipin total area sum, was calculated using appropriate cell masks. An example is shown below in Figure 2. We can clearly see that in case of Fig 2A (a compound that increases filipin staining), there are more filipin puncta (yellow dots in blue mask) than in the case of Fig 2B, which is a SAHA-treated control. (More on SAHA in a bit). This confirms that the parameters set for the analysis were effective. Z-scores were then calculated using these filipin area sums. Figure 3 shows an example of our analyzed plate helping you visualize that we had strong positive (Column 2) and negative (Column 23) controls.
Figure 2. Cell masks for a) positive control and b) negative control
Figure 3. Figure showing a screenshot of an analyzed plate
This screen resulted in hit compounds that cleared filipin (SAHA-like) and those that had intense filipin fluorescence (PERL101-like). Along with the range of clearance to increased cholesterol accumulation, there are also many compounds that do not do anything to the baseline filipin staining phenotype and look similar to DMSO-treated controls.
Figure 4. a) DMSO-treated control b) SAHA-treated control c) SAHA-like hit compound
Figure 5. a) DMSO-treated control b) PERL101-treated control c) PERL101-like hit compound
Figure 4 shows the comparison of DMSO control to SAHA treatment, with SAHA clearing almost all the filipin while Fig 4C shows a compound that also clears filipin. Similarly, Figure 5 shows images of a DMSO-treated control, PERL101 increasing filipin and Fig 5C shows a PERL101-like hit compound which also increases filipin. Figure 6 show images of compounds that resulted in interesting phenotypes and a variation in filipin distribution, or toxicity.
Figure 6. (a), (b) Interesting phenotypes (c), (d) Toxic compounds
Only ~1% of the library turned up as SAHA-like hits, whereas ~8% turned up as PERL-101 like hits. Once we compiled the ID numbers of SAHA-like hits, PERL101-like hits and toxic compounds, we went back into our Collaborative Drug Discovery portal to make a list of all the compounds. We observed a lot of compounds that increased the amount of free cholesterol in the cell, and it was only when we looked at the list that we realized most of those compounds were large lipophilic compounds that would lead to this phenotype and might not rescue disease in the model organisms. This pilot test was telling because the clearance phenotype that SAHA induces can also be caused by compounds that inhibit cholesterol synthesis, uptake or other affect other vital cellular processes. It is clear that cell data has to be paired with model organism data, and cannot alone be an indicator of disease rescue. The data need to be cross-referenced with the model organism data.
The next step would be testing the library on the worm and fly NPC platforms and consequently, look for overlaps to better characterize the hits that come out of our screens, collect more data on false positives and negatives, and compile a list of toxic compounds and modifiers.