A phenotypic screening platform utilising human spermatozoa identifies new compounds with contraceptive activity

There is an urgent need to develop new methods for male contraception, however a major barrier to drug discovery has been the lack of validated targets and the absence of an effective high-throughput phenotypic screening system. To address this deficit, we developed a fully-automated robotic screening platform that provided quantitative evaluation of compound activity against two key attributes of human sperm function: motility and acrosome reaction. In order to accelerate contraceptive development, we screened the comprehensive collection of 12,000 molecules that make up the ReFRAME repurposing library, comprising nearly all the small molecules that have been approved or have undergone clinical development, or have significant preclinical profiling. We identified several compounds that potently inhibit motility representing either novel drug candidates or routes to target identification. This platform will now allow for major drug discovery programmes that address the critical gap in the contraceptive portfolio as well as uncover novel human sperm biology.

produced in large numbers. Moreover, as the production of fertilization competent cells is necessarily different between 53 species, it is essential to use a human system early in the drug discovery cascade to increase the chances of translational 54 success.

56
Though some potential sperm-specific contraceptive drug targets have been identified (Lishko and Mannowetz, 2018), 57 our relatively poor understanding of the spermatozoon makes target-agnostic strategies more attractive (Barratt et al, 58 2017). Phenotypic drug discovery is undergoing a renaissance in a number of therapeutic areas because it can uncover 59 novel biology in an unbiased way. Retrospective analysis of drug approvals shows the approach to be more successful 60 in first-in-class medicine discovery than previously appreciated (Swinney and Anthony, 2011;Moffat et al., 2017). For 61 male contraceptive drug discovery, a limiting barrier has been the absence of a scalable drug screening system using 62 spermatozoa.

63
To address this unmet need we developed the first high-throughput phenotypic sperm screening platform where an 64 image-based kinetic analysis module, tracking movement, is coupled to a flow cytometry module that detects the systems (Supplementary Figure 1B) and a comparable distribution was observed between experimental days, plates and

82
Importantly, spermatozoa were found to be tolerant to DMSO (~4%), far above typical screening concentration of 0.1% 83 (Supplementary Figure 1C). To determine the optimal screening batch size spermatozoa were dispensed every 30 84 minutes for 150 minutes observing only a small decrease in motility (~10%) after 4 plates were screened (Supplementary 85 Figure 1D). A flow cytometry-based assay measuring AR ( Figure 1D) was run directly afterwards ( Figure 1E, F & 86 Supplementary Figure 1E).

88
Screening of the ReFRAME library, confirmation of hits by dose response.

89
The ReFRAME library (11,968 compounds) was supplied in "assay-ready" imaging plates and was solubilised prior to  Table 1). Amongst the confirmed hits were the aldehyde dehydrogenase inhibitor, Disulfiram (70% maximum reduction 96 at 10 µM) and a putative platelet aggregation inhibitor, KF-4939 (100% max. reduction; EC50=0.49µM), and a range of  Table 1). In the AR screen, 9 of the resupplied compounds had a dose-dependent effect with EC50 values as low as 0.4 99 µM (See Supplementary Figure 3, Supplementary Table 2). However, following the orthogonal assay triaging none of scalable to allow throughput at sufficient speed. Moreover, concomitant assessment of a second functional attribute,

111
AR, was developed in order to maximise use of the biological material. Assessment of motility required novel 112 implementation of tracking algorithms. Importantly, the kinetic outputs were equivalent to those from the "industry-113 standard" CASA, which is very low throughput and unsuitable for screening more than a few compounds at one time.

114
Significant workflow optimisation resulted in good assay robustness and a high hit confirmation rate following the 115 primary screen. For AR detection, a flow cytometry approach was taken based on established methods (Mortimer et al., 116 1987). However, the presence of assay-interfering compounds in screening libraries (potential fluorescent compounds 117 including DNA intercalators) necessitated further triaging steps (see Materials and Methods). This combination of 118 triaging approaches meant that, in this screen, no AR compounds were suitable for progression. Overall, although this 119 was a complex biological assay, it achieved acceptable throughput with the potential to increase batch size for larger 120 screening campaigns.

122
The advantage of using the ReFRAME collection was the potential to repurpose drugs that are already approved for 123 other indications or are at earlier stages of clinical progression. The associated compound annotations and safety data 124 offers the potential to help accelerate the development of an effective male contraceptive. In the motility assay the hit 125 rate of ~0.2% is typical for a cell-based assay, however surprisingly a relatively high number of hits were not deemed 126 suitable for compound progression. Examples included potentially toxic mercury-containing compounds 127 (phenylmercuric borate and mercufenol chloride); antiseptics and antibiotics including tyrothricin, or compounds with 128 impacts on fundamental biology e.g. the microtubule stabiliser and chemotherapeutic, docetaxel. However, several hits 129 are potentially of interest. One is Disulfiram (70% maximum reduction at 10 µM), a drug that is well tolerated, has long 130 been used for treating alcohol dependency and has previously been shown to inhibit sperm motility (Lal et al., 2016).
Gruber et al., 10 th October 2019 SHORT REPORT 6 141 Moreover, the Toll-like receptor 7/8 ligand (Resquimod; R848) recently shown to preferentially reduce the motility of X 142 chromosome-bearing mouse sperm by suppressing ATP production (Umehara et al, 2019) was a hit in our screen, 143 demonstrating that this phenotypic screening approach has the potential to uncover new aspects of sperm biology.

145
In summary, our high-throughput phenotypic platform allows for the screening of bioactives including natural products, 146 focussed small molecule libraries and chemical diversity sets, which can provide novel start-points for male 147 contraceptive drug development.

157
A pre-existing automated cell-based phenotypic screening platform was adapted that utilises a Yokogawa CV7000 Cell 158 Voyager high-throughput microscope able to image 384 multiwell plates under full environmental control at high speed.

159
For high throughput screening with human spermatozoa to be effective and yield meaningful hits, the system needed 160 to replicate, as close as possible, the manual workflow currently used in an andrology lab, so that the effect of large 161 numbers of compounds on viable human sperm could be performed. In order to achieve this goal, we investigated a 162 number of factors that could impact the quality of the data such as: plate type; concentration and number of 163 spermatozoa per well; logistics of spermatozoa preparation and handling; dispense speed for getting spermatozoa into 164 384 well plates; timing of compound addition; temperature control; image acquisition parameters; algorithm choice for 165 motility assessment and data management and, evaluating the overall speed of data analysis. A flow cytometry-based 166 assay and data analysis workflow was developed to measure acrosomal status. The goal was to be able to screen a 384-167 well plate in both assays consecutively within 90 minutes of sperm dispensing using a fully automatic robotic system.

168
Data was analysed for inter-and intra-plate variation (both technical and biological replicates) and signal-to-background 169 and variance measured to demonstrate the assay was reproducible and robust. The assay was shown to be scalable 170 and all the workflows optimal. Once established, the system was validated by screening the ReFRAME library (Janes et Inc.) and maintained at 37°C with gentle stirring (100 rpm) in a water bath on top of a magnetic stirrer. Approximately 186 10,000 spermatozoa were dispensed per well using a MultiDrop Combi (ThermoFisher) into pre-warmed assay-ready 187 plates. Spermatozoa were incubated with the compounds for 10 mins (thus favouring fast acting compounds) at 37°C 188 in the CV7000 microscope prior to the commencement of imaging.

191
A CV7000 Cell Voyager high-content imaging system (Yokogawa) was used as it allowed full environmental control, 192 sufficient contrast using simple brightfield optics and a fast acquisition rate (up to 45 frames per second). Using a 20x 193 lens with bright field illumination (0.11 ms exposure, 3% lamp power), time-lapse image series were acquired (24 frame 194 in 0.5 sec) at two positions per well (with a 400 μm gap between positions to eliminate double counting of sperm).

195
Image acquisition across a 384-well plate with such settings takes ~17 minutes.

198
A Python implementation of the particle tracking algorithm originally developed by others (Crocker and Grier, 1996)      Compounds were spotted (12.5 nL, final assay concentration ~6 µM) into 384-well black-sided optical imaging plates 252 (CellCarrier, PerkinElmer) at CALIBR and were shipped on dry ice to the screening centre in Dundee and stored at -20 ₒ C 253 until required. Immediately prior to screening plates were thawed, controls were added using the acoustic dispenser 254 (Echo 555, Labcyte), and 10 µl warm media added followed by shaking on a plate shaker for 10 seconds. Plates were 255 then incubated for 30 min at 37°C to solubilise the compounds and prewarm the plates prior to the addition of live 256 sperm.
All experiments were performed with pools of donors (3-5 per batch of 4 screening plates) and results normalised to 259 DMSO-treated wells (16 wells per plate).

261
The ReFrame library has been screened for cytotox (using CellTiterGlo) in HEK293 and HepG2 cells (see 262 www.reframedb.org) with some of the hit compounds displaying activity in one or both assays (Supplementary Figure   263 2), information which helps inform compound progression decisions.

270
Triaging of AR hits.

271
Given the prevalence of a range of interfering compounds amongst the library hits, two approaches were used for