All experimental procedures were carried out with approval of the Monash University Biological Sciences Animal Experimentation Ethics Committee and conducted in accordance with Australian National Health and Medical Research Council (NHMRC) Guidelines on Ethics in Animal Experimentation. Crisp1, 2, 4, and Crisp1/4 double knockout mouse models were produced as previously described (Hu et al., 2018; Lim et al., 2019). Prior to use, genotype of all animals was confirmed by PCR analysis of tail biopsies as previously described (Hu et al., 2018; Lim et al., 2019).

Sperm were collected from adult male mice (10–12 weeks old) using the back flushing technique and stored at 37°C in dark until used as described previously (Lim et al., 2019). Sperm were diluted to 1 × 10⁵ sperm/mL into modified Toyoda, Yokoyama and Hosi (TYH) media (135 mM NaCl, 4.8 mM KCl, 2 mM CaCl₂, 1.2 mM KH₂PO₄, 1 mM MgSO₄, 5.6 mM glucose, 0.5 mM Na-pyruvate, 10 mM L-lactate, and 10 mM HEPES, pH 7.4) containing 0.3 mg/ml BSA to facilitate the binding of sperm heads to the microscope slides (Toyoda et al., 1971; Lim et al., 2019). For imaging, a chamber was constructed on a glass slide with two parallel strips of double-sided tape of 90-micron nominal thickness spaced 16 mm apart, and 40 μl of the sperm suspensions was added and the chamber capped using a 17 mm² No. 1.5 glass coverslip (Thermo Fisher Scientific). Head tethered sperm with free beating flagellum were imaged at room temperature (25°C) within 10–15 min of collection from the cauda epididymis at 400 fps for 10 s of which the first 1,000 frames were used in the analysis as described previously (Nandagiri et al., 2021). N = 25–30 sperm/genotype from at least 5–6 mice were imaged and analyzed.

The imaging system was constructed around an Olympus AX-70 upright microscope equipped with a U-DFA 18mm I.D. dark-field annulus, an UPlanAPO 20x 0.7 NA objective (Olympus, Japan) and incandescent illumination. In order to maximize system light efficiency, all extraneous optical elements were removed from the light paths. Images were captured on an ORCA-Flash4.0 v2+ sCMOS camera (Hamamatsu, Japan) streaming to a dedicated Firebird PCIe3 bus 1xCLD Camera Link frame grabber card (Active Silicon, United Kingdom). The image capture computer system utilized a Xeon E5-2667 12 core CPU running at 2.9 GHz with 64 Gb of DDR3 RAM and 1 TB of SSD hard drive space in a RAID0 configuration. Camera software control was affected by Fiji (ImageJ)¹ (Schindelin et al., 2012) running the Multi-Dimensional Acquisition plugin under the Micro Manger Studio plugin v1.4.23² (Edelstein et al., 2014). Acquisition was set to 4,000 time points with no time point interval and a 2 ms exposure time at 400 frames per second (fps) with the data saved as an image stack. The optical lateral resolution, at a 550 nm reference wavelength was 0.479 microns. The best-case optical lateral resolution, at Nyquist-Shannon sampling, was 0.650 microns (0.325 microns per pixel). Imaging at 512 × 512 pixels resulted in a sample field of view (FOV) of 166.4 × 166.4 microns. Mutant sperm which had a more rigid linear tail morphology necessitated an FOV increase, with a consequently minor reduction in the frame capture rate in order to capture the motion of the entire sperm. The best-case blur free motion capture of the system using 400 fps allowed element point velocities of 130 microns per second at Nyquist-Shannon frequency sampling to be measured. Mouse sperm are 110 μm in length and were divided into three parts: mid-piece (24 μm), principal piece (80 μm) and end piece (∼6 μm) for analysis (Nandagiri et al., 2021).

Sperm viability over time was assessed using the LIVE/DEAD^TM Sperm viability kit as recommended by the manufacturer (Thermo Fisher Scientific).

Image-analysis protocols are used to extract smooth centerlines of sperm in every frame of the recorded image sequence (Nandagiri et al., 2021). The tangent angle at any point on the centerline is the angle made by the tangent at that point with the horizontal axis. Henceforth, s denotes the arc-length coordinate along the centerline and τ denotes time. The tangent-angle data is obtained at discrete locations, s₁, s₂, …, s_M, and at times, τ₁, τ₂, …, τ_N. The proper orthogonal decomposition of the tangent angle profile has been described previously by Werner et al. (2014) and Nandagiri et al. (2021). With this technique, the tangent angle data is resolved into a weighted sum of time-independent tangent angle profiles known as the shape modes:

Here, ψ₀ is the time average of the tangent angle observed at each s_m and B_k is the relative weight of the k-th shape mode, ψ_k. The mean shape ψ₀ and the shape modes ψ_k represent the entire set of shapes in the flagellar beating pattern. They have the units of radians. The dimensionless weight, B_k, of the k-th shape mode is referred to as a shape coefficient.

The procedure to find the shape modes and the coefficients is summarized here. The values of the deviation X_nm of ψ from the mean ψ₀, X_nm = ψ - ψ₀, is an N × M matrix. In matrix form, we can therefore write X = B ⋅ ψ, where the dot signifies matrix multiplication, the k -th column of ψ corresponds to the M values of k -th shape mode vector, _k, and the (k, m) -th element of B is B_k (τ_n). The values of B and can be uniquely obtained by singular-value decomposition (SVD) of X which is calculated as follows (Ma et al., 2014; Werner et al., 2014). The M × M covariance matrix C_X = X^T ⋅ X was first calculated. Its eigenvectors arranged as columns gives ψ. The matrix X ⋅ Ψ^T.

From the smooth tangent angle profiles obtained by image-analysis and the proper orthogonal decomposition, we can calculate the instantaneous tangent and normal vectors, t→ and n→, and the velocity profile, u→, as functions of the local arc-length coordinate along the flagellar centerline, s, and the time, τ. The sperm flagellum operates in the low Reynolds number regime of fluid flow where the inertial forces are very small relative to the viscous forces. In this regime, the hydrodynamic dissipation can be calculated from the flagellar velocities using local Resistive Force Theory [RFT (Katz et al., 1975; Nandagiri et al., 2021)]. The hydrodynamic force per unit length at each point on the sperm body are directly related to the local velocities:

where, ζ_t and ζ_n are the local friction coefficients in the tangential and normal directions, and ut→ and un→ are the instantaneous local tangential and normal velocities obtained from the image-analysis algorithm.

From RFT, local friction coefficients that consider the presence of the nearby surface are used with a linear taper in the flagellar diameter a:

where, the flagellar diameter a has a linear taper:

with a_h = 0.57μm and a_t = 0.18μm as the radii at the head and tail end, respectively, and L = 110 μm. Since the head is tethered and the tail is nearly parallel to the surface, the distance of the tail from the surface h is approximated as h = a_h, which is the radius at the head (Nandagiri et al., 2021).

The instantaneous contribution of the motion of each point of the sperm body to the power dissipated in the ambient fluid due to viscous friction is then:

We refer to this as the hydrodynamic dissipation. The time-average of the power dissipation per unit length at each location along the tail over several beat cycles is then given by:

where, T is the total time duration. The rate of hydrodynamic power dissipated by the whole cell is:

The integrals are evaluated as discrete sums along the discrete centreline and the time points.

The hydrodynamic force distribution, f→(s,τ), was used to compute a hypothetical free-swimming trajectory for the sperm. This is the path that would be followed by the sperm if the tether was removed. In these calculations, it is assumed that the shape of the flagellar beat does not change after untethering.

We first assumed that the unknown free-swimming velocity is U→. The total hydrodynamic force on the hypothetical free-swimmer must be zero since the inertia is negligible. Each segment of the free-swimmer than moves at a velocity u→+U→ i.e., the sum of the undulatory velocity and the velocity of translation. Therefore, the net hydrodynamic force:

where, ζ→→=ζtt→t→+ζnn→n→ is the hydrodynamic friction tensor. Re-arranging, we get the following expression for the unknown velocity of translation:

where,

The projected trajectory R→(τ) of the hypothetical free-swimming sperm can then be computed by integrating the velocity of translation U→(τ):

The projected straight-line velocity (VSL) is then calculated from the net displacement per beat cycle:

Recombinant mouse CRISP4 protein was expressed and purified as described previously and the same batch was used in this study (Gaikwad et al., 2020b). The expression of recombinant mouse CRISP1 protein was carried out by transiently expressing a pcDNA^TM 5/FRT/TO construct containing Crisp1 (ENSMUSG00000025431) cDNA in Expi293F^TM cells (Thermo Fisher Scientific) as described in Volpert et al. (2014). The transient transfection was carried out using cells at a density of 3 × 10⁶ cells/ml, with 1 mg/ml polyethylenimine (PEI) (Polysciences, Inc., United States) at a ratio of 1:3 (DNA:PEI) and incubated at 37°C, 5% CO₂ in suspension culture for 5 days on an orbital shaker rotating at 110–140 rpm. After a further 72 h of incubation, cells were harvested by centrifuged at 1,000 rpm at 4°C for 10–15 min. The supernatant was retained for protein purification and cells pellets were discarded. The cell supernatant was dialyzed against 10mM HEPES buffer, pH 6.5, using 10kDa cut-off dialysis tubing overnight. Post-dialysis, supernatant proteins were separated using ion exchange chromatography using a carboxymethyl cellulose (CM) resin column (GE Healthcare Life Sciences). Bound protein was eluted from the CM resin in 50, 100, and 150 mM NaCl elution buffer (Supplementary Figure 1). To assess the effects of recombinant mouse CRISP1 and CRISP4 proteins on sperm motility, the proteins were buffer exchanged into standard TYH buffer using Amicon Ultra-15 centrifugal filter concentrators. Denatured recombinant CRISP1 and CRISP4 were obtained by heat-treating the purified proteins at 95°C for 10 min and used as negative controls for functional assays.

For functional assays, individual recombinant CRISP proteins at a final concentration of 1 μM were added to 1 × 10⁵ sperm/ml in TYH media as described above.

The degree of difference between a group of representative beat patterns was quantified by the following Procrustes measure:

where, x and y are the positional coordinates of group (1) and (2) at a phase-time τ_i and at a location corresponding to the arc-length variable s_j along the flagellar centreline. Procrustes measures were compared pair-wise across the different genotypes. Larger values of d imply larger differences between the flagellar waveforms.

Data were analyzed using Systat Version 13. The effect of gene deletion on aspects of sperm function were evaluated with ANOVA. Data for individual mice was averaged and values were compared between mice. Where significant differences were detected among groups, post-hoc t-tests were used to evaluate which treatment combinations drove the interactions and main effects. Significant differences were indicated with ^∗P < 0.05, ^∗∗P < 0.01, ^∗∗∗P < 0.001, and ^****P < 0.0001.

To quantify the kinematics of sperm motility, we used the high-resolution high-speed dark-field imaging system described in Nandagiri et al. (2021). In brief, sperm were collected from the cauda epididymis and imaged within 10–15 min, based on experiments revealing that motility parameter were unchanged up to 60 min post-collection (Supplementary Figure 2). Sperm were tethered to glass microscope slides and head-tethered cells were imaged over 10.0 s at 400 fps. As illustrated in the Supplementary Videos 1–8, wild type sperm from each of the C57BL/6J × C57BL/6N (Crisp1^–/– colony), C57BL/6N (Crisp2^–/– and Crisp4^–/–– colonies), and C57BL/6J × C57BL/6N (Crisp1/4^–/– colony) genetic backgrounds were visually comparable and executed traveling-wave beating patterns. 25–30 sperm from 5 to 6 mice per mouse strain were analyzed for the parameters described below. Raw image data (25–30 sperm/genotype from 5 to 6 mice/strain) used in this study can be found at³.

In order to precisely define the shape of the sperm flagella throughout the beating cycle, and over multiple cycles (the first 1,000 frames), centrelines were extracted from the image sequences and proper orthogonal decomposition (POD) (Werner et al., 2014) was used to resolve the two dimensional shape of each tail centreline, in each frame, into a weighted sum of distinct shape modes, The weighted combination of the first four of these modes (∑i=14Biψi) accounted for >98% of the total flagellum movement (Supplementary Figures 3A,B), with the weights (B₁-B₄) varying within a beat cycle and between cycles [i.e., each weight is a time-series B_i(t)]. Among these four modes, Mode 1 (B₁) and Mode 2 (B₂) were dominant, accounting for 96% of the shapes over time. To a high degree of accuracy, therefore, every pair of B₁ and B₂ values, represents a unique tangent-angle profile. If the sequence of shapes is periodic in time, it will result in a cyclic trajectory when B₁ is plotted B₂. As shown by Nandagiri et al., such cycles can be used to systematically compare beating patterns of different sperm in a statistically meaningful manner. These coefficients were thus plotted against each other to aid comparison of beat cycles (Figure 1A). These data reveal that wild type sperm display repetitive and periodic flagellar oscillation that were indistinguishable between the four tested mouse strains.

To allow a quantitative assessment of variability between mouse strains, we used the extracted centreline data to construct a mean beating pattern for each sperm. Individual sperm waveform data, within a strain (Figure 1B), were then compared in terms of a Procrustes distance. The Procrustes distance quantifies the differences between waveforms and is defined as the root-mean-squared deviation between the flagellar shapes compared over a full representative beat cycle (intra-strain comparison, Supplementary Table 1). The beat cycles were then averaged to yield a representative beat pattern in the wild type population for each strain (Figure 1B), and a comparison of pairwise deviations made between each combination of wild type strain data to yield an inter-strain Procrustes distance (Table 1A). No significant differences in flagellar waveform were observed within or between wild type mouse strains (Supplementary Table 1).

Subsequently, Fourier spectral analysis of the shape mode co-efficient (B₁) and (B₂) was undertaken to estimate the primary oscillating frequency (POF), a measure of the sperm tail beat frequency. The average POF wild type mouse sperm was 7.35 Hz and did not differ significantly between strains (Figure 2 and Supplementary Table 2).

To evaluate swimming efficiency, the hydrodynamic power dissipation was calculated from the representative beat cycle of individual sperm using resistive force theory (Nandagiri et al., 2021). With regards to sperm, power dissipation is a measure of average power dissipated into the surrounding fluid at each point on the sperm flagellum. These data were comparable between sperm from the same strain, and between strains, and revealed that the power dissipation progressively increased along the tail (Figure 1D). We calculate that wild type sperm in a low viscosity medium (1 cP) dissipate 151 ± 5.05 fW across a beat cycle (Figure 3A). To complement these analyses, we measured the amplitude of the beating along the tail, including at set points corresponding to the mid-piece (24 μm), principal piece (80 μm), and end piece (∼6 μm) (Figure 1C). Flagellar amplitude was statistically comparable within and between mouse strains (Supplementary Tables 3–3.1).

In summary, sperm flagella waveform was highly repetitive, and predictable, between sperm within and between the mouse strains tested here. Regardless, for the analyses described below, sperm from each genetically modified mouse strain were compared to data from wild type littermates. For simplicity, Crisp1 strain wild type data is shown in Figures 3A–D. All data can be found in Supplementary Figures.

In order to test the hypothesis that individual CRISPs play distinct roles in regulating sperm motility, sperm from each of Crisp1, Crisp2, Crisp4, and Crisp1/4 double knockout, and wild type littermates were analyzed using the pipeline described above. This was initially done for the intracellular CRISP, CRISP2. As shown in Figure 2, the loss of CRISP2 reduced the POF of spermatozoa by 20% (Crisp2^+/+ 7.4 Hz vs. Crisp2^–/– 5.9 Hz, P < 0.05, Supplementary Table 2). The shape cycle for sperm from Crisp2^–/– mice had significant variability in the flagellar shapes between beat cycles (Figure 3A) (Procrustes distance Crisp2^+/+ c.f. Crisp2^–/– mice, 4.5 ± 1.43, P < 0.001, Table 1A). The horizontal axis on the reconstructed mean beat cycle graph (Figure 3B), which corresponds to the length along the flagella, was used to determine the extent of restrained mobility in the flagella. Consistent with our prior work, the reconstructed mean beat cycle revealed that loss of CRISP2 lead to sperm with a stiff mid-piece (Figures 3B,C) (Lim et al., 2019), notably the first 36 μm (0.3 x-position on the X-axis) of the flagellum (Figure 3B). Sperm from Crisp2^–/– mice also displayed a significantly lower flagellar amplitude along all regions of the tail compared to wild type counterparts (Figures 3C, 4A–C and Supplementary Table 3) and had a lower hydrodynamic power dissipation along the entire length of the flagellum (Table 1B, Figure 4B, and Supplementary Table 4). Overall, the loss of CRISP2 resulted in a 55% reduction in sperm power dissipation compared to wild type (Crisp2^+/+ 151 fW vs. Crisp2^–/– 68 fW; P < 0.0001, Figure 5A and Supplementary Table 5).

Collectively, these data as reflected in the primary imaging data, reveal that CRISP2 plays a role in establishing flexibility principally in the sperm mid-piece, and in doing so, allows for an increase in flagella POF and amplitude. Nandagiri et al. (2021) have shown that the power dissipated hydrodynamically in the ambient fluid is proportional to the mechanical power delivered by the dynein motors in the axoneme. Therefore, the data effectively shows that the presence of CRISP2 boosts dynein power input.

The loss of the epididymal CRISP, CRISP1, also resulted in compromised flagellar function and was distinctly different to that seen in Crisp2 null sperm. Crisp1 deletion resulted in a significantly reduced POF compared to sperm from wild type mice (Crisp1^+/+ 7.3 Hz vs. Crisp1^–/– 5.8 Hz, P < 0.05, Figure 2). The corresponding shape cycle revealed significant variability in the flagellar waveforms between beat cycles (Figure 3A), and thus a relatively large Procrustes distance between genotypes (7.9 ± 1.8, Table 1A). The mean beat cycle revealed that sperm from Crisp1^–/– mice displayed a highly restrained flagella amplitude, wherein the region of “stiffness” extended through the mid-piece and into the principal piece up to a distance of 60 μm (Figures 3B, 4B and Supplementary Table 3). The loss of CRISP1 resulted in a 50 and 64% reduction in power dissipation from the mid-piece and principal piece of the flagella, respectively, when compared to wild type sperm (Table 1B and Figures 3D, 4B), and collectively a 65% reduction in power dissipation per cycle (Crisp1^+/+ 147 fW compared to Crisp1^–/– 51 fW; P < 0.0001, Figure 5A and Supplementary Table 5). These data reveal that CRISP1 plays a role in defining the energetics of the mid-piece and proximal principal piece of the sperm tail and suggests that the “receptors” mediating CRISP1 function are located along the proximal region of the tail plasma membrane.

Sperm from Crisp4^–/– mice also had a significantly decreased POF compared to controls (Crisp4^+/+ 7.5 Hz vs. Crisp4^–/–4.1 Hz, P < 0.0001, Figure 2 and Supplementary Table 2), and their flagellar waveform were distinctly different to sperm from Crisp1^–/– or Crisp2^–/– mice. The shape cycle displayed irregular circular loops (Figure 3A) and the reconstructed mean beat cycle, as reflected in the original videos, revealed that sperm from Crisp4^–/– mice were restrained along the entire length of the flagella compared to sperm from wild type littermates (Figure 3B). Of the sperm genotypes analyzed in this study, Crisp4^–/– mouse sperm had the highest variation in flagellar kinetics (Procrustes distance value of 7.69 ± 1.5 compared to wild type littermates, Supplementary Table 1). The power dissipation for sperm from the Crisp4^–/– was 86, 56, and 53% lower along the mid-piece, principal piece and end piece of the flagella, respectively, compared to sperm from wild type littermates (Table 1B and Figure 4B). Overall, the loss of CRISP4 resulted in a 65.5% reduction in power dissipated per sperm per cycle (Crisp4^+/+ 154 fW vs. Crisp4^–/– 53 fW, P < 0.0001, Figure 5A and Supplementary Table 5). These data suggest that the molecules that mediate CRISP4 function on the sperm plasma membrane are distributed along the entire tail and that they ultimately feed into the energy generating machinery of the axoneme.

Our previous study showed that while sperm from Crisp1^–/– mice had a compromised ability for progressive motility (Hu et al., 2018), somewhat surprisingly, sperm from Crisp4^–/– mice displayed a significantly increased progressive motility compared to those from wild type littermates in aqueous media (Hu et al., 2018). To address this apparent conundrum, using the high-resolution data obtained in this study, we calculated the net displacement of individual sperm from each of the four genetically modified mouse lines by mathematically modeling individual sperm movement under untethered, free swimming conditions. The estimated progressive motility of sperm from Crisp4^–/– mice in a low viscosity media was 36% higher than that of sperm from control mice (Crisp4^+/+ 3.3 μm/s vs. Crisp4^–/– 4.5 μm/s, P < 0.01, Figure 5B and Supplementary Table 6) and thus qualitatively consistent with the computer assisted sperm analyser (CASA) measurements. Also, in accordance with the CASA data, the loss of CRISP1 or CRISP2 resulted in a reduced projected net progressive motility compared to control: by 33% for CRISP1 (Crisp1^+/+ 3.3 μm/s vs. Crisp1^–/– 2.2 μm/s, P < 0.01) and 13% for CRISP2 (Crisp2^+/+ 3.1 μm/s compared to Crisp2^–/– 2.7 μm/s, P < 0.05) (Figure 5B and Supplementary Table 6).

Depending on the species, sperm encounter CRISP1, and/or CRISP4 during epididymal maturation. The removal of all epididymal CRISPs from the mouse leads to compromised sperm function and fertilization potential (Hu et al., 2018), and as we have hypothesized previously, is likely to mirror the roles of the single CRISP expressed in the human epididymis (Hu et al., 2018). As revealed here, the beating pattern of sperm from Crisp1/4 double knockout mice were superficially similar to sperm from wild type controls. There was a low Procrustes distance value between genotypes (2.5 ± 1.1, Table 1A) and flagellar amplitudes were comparable (Supplementary Tables 3–3.1). The combined deletion of Crisp1 and Crisp4 did not, however, rescue sperm functional competence. Sperm from Crisp1/4^–/– mice had significantly reduced POF (Crisp1/4^+/+ 7.2 Hz vs. Crisp1/4^–/– 4.8 Hz, P < 0.0001, Figure 3C and Supplementary Table 2) and a 59% reduction in power dissipation along the flagellum compared to wild type controls (Crisp1/4^+/+ 147 fW compared to Crisp1/4^–/– 60 fW, P < 0.0001, Figure 5A and Supplementary Table 5). The loss of epididymis CRISPs also resulted in reduced estimated velocity (Crisp1/4^+/+ 3.5 μm/s vs. Crisp1/4^–/– 2.9 μm/s, P < 0.05, Figure 5B and Supplementary Table 6) consistent with CASA data (Hu et al., 2018).

Collectively, these data reveal that in the mouse, CRISP1 and CRISP4 function to define different, potentially opposing, aspects of sperm flagella waveform, but collectively function to amplify the axonemal power generation. Further, and as a consequence of the observation that sperm from Crisp1/4^–/– appeared to lose motility more quickly than sperm from wild type of single deletion mice, we undertook a temporal analysis of sperm motility. Paired LIVE/DEAD staining and motility analysis revealed that this was not due to elevated rates of death (Supplementary Figure 4), but rather sperm from Crisp1/4^–/– mice rapidly became immotile after removal from the cauda epididymis.

The removal of any form of CRISP compromised flagellar waveform, power dissipation and POF (Figures 2–4). While the combined removal of CRISP1 and CRISP4 returned sperm flagellum shape to close to that seen in sperm from wild type mice, it did not rescue power. Each of the three CRISPs play a role in maximizing power output and each act autonomously (Figure 5A). Each of CRISP1, CRISP2 and CRISP4 have roles in defining POF, however, CRISP4 appears to have the dominant role and, as suggested by the non-statistically significant difference between Crisp4 and Crisp1/4 knock out mouse values, it appears to function up-stream of CRISP1 in the mouse (Figure 2).

To test if the phenotypic deficits induced by the loss of epididymal CRISPs can be rescued by in vitro exposure to the corresponding protein we produced recombinant CRISP1 and CRISP4 and exposed sperm to previously published physiologically relevant concentration of CRISPs in vitro (Ernesto et al., 2015). As illustrated in Supplementary Videos 9, 10 the addition of 1 μM of either recombinant CRISP1 or CRISP4 to sperm from wild type mice had no discernible effects on motility parameters (Figures 6A–F). By contrast, the motility defects seen in sperm from Crisp1^–/– and Crisp4^–/– mice were qualitatively and quantitatively rescued by exposure to 1 μM recombinant mouse CRISP1 (Supplementary Video 11) and CRISP4 (Supplementary Video 12), respectively (Figure 6). Raw image data used in this study can be found at⁴.

Specifically, incubation of sperm from Crisp1^–/– and Crisp4^–/– with the respective recombinant protein recovered the POF, flagellar amplitude, and power dissipation resulting in a mean beat cycle comparable to the sperm from wild type mice (Figure 6, Supplementary Figure 6A, and Supplementary Tables 7, 8). To test for functional redundancy between recombinant CRISPs, we further tested the effects of recombinant CRISP1 on sperm from Crisp4^–/– mice and vice versa. No recovery of motility parameters was observed indicating that mouse CRISP1 and CRISP4 function non-redundantly in regulating ex vivo sperm motility (Supplementary Videos 1, 2 and Supplementary Figures 5B, 7).

By contrast, sperm from Crisp1/4^–/– mice when exposed to individual recombinant CRISP1, CRISP4 or combined CRISP1 and 4, did not rescue the deficit in power dissipation (Supplementary Figures 6C, 8 and Supplementary Tables 10.1, 10.2), noting the waveform of sperm from wild type and Crisp1/4^–/– was already comparable (as shown in Figure 3 and Supplementary Table 10.1). These data suggest CRISP1 and 4 play a temporally restrained role for epididymal CRISPs in epididymal sperm maturation and that this role can be mediated by either CRISP1 or CRISP4 i.e., they act redundantly.