Human spermatozoa are the archetype of long-term self-organizing transport in nature and are critical for reproductive success. They utilize coordinated head and flagellar movements to swim long distances within the female reproductive tract in order to find and fertilize the egg. However, to date, long-term analysis of the sperm head–flagellar movements, or indeed those of other flagellated microorganisms, remains elusive due to limitations in microscopy and flagellar-tracking techniques. Here, we present a novel methodology based on local orientation and isotropy of bio-images to obtain long-term kinematic and physiological parameters of individual free-swimming spermatozoa without requiring image segmentation (thresholding). This computer-assisted segmentation-free method evaluates, for the first time, characteristics of the head movement and flagellar beating for up to 9.2 min. We demonstrate its powerful use by showing how releasing Ca2+ from internal stores significantly alters long-term sperm behavior. The method allows for straightforward generalization to other bio-imaging applications, such as studies of bull sperm and Trypanosoma, or indeed of other flagellated microorganisms – appealing to communities other than those investigating sperm biology.

The main objective of a spermatozoon is to fertilize the female gamete. To achieve this, the mammalian sperm, which measures ∼50 micrometers, navigates through an approximately 10 cm-long female reproductive tract to find the egg. With an average swimming speed ranging between 35 and 50 µm/s in viscous physiological media (Milligan et al., 1980; Smith et al., 2009), human spermatozoa would take ∼100 min to cover this distance. During this journey, a spermatozoon will undergo several physiological changes due to alterations in media ionic composition, flows within the female genital tract and signals from the surrounding cells, some of which have been argued to induce chemotaxis (reviewed in Darszon et al., 2011, 2020). It is thus natural to expect that during this relatively long time, spermatozoa need to adapt their behavior to overcome the ever-changing conditions and obstacles of the reproductive tract (Gaffney et al., 2011). Globally, the biochemical changes that occur in the female tract that are essential for the maturational process that allows mammalian sperm to fertilize the egg are known as capacitation. They include flagellar beat alterations, which when measured in aqueous media, change from high frequency, low amplitude and symmetrical to a vigorous, highly asymmetric mode displaying deep flagellar bends in the midpiece and principal piece, and pronounced head lateral movements (head yawing; reviewed in Stival et al., 2016). This latter mode of flagellar beating is named hyperactivation and is necessary for fertilization (Chang and Suarez, 2011). Its characteristics can vary among species and even within the same cell population (Drobnis et al., 1988; Ravaux et al., 2016), indicating possible functional switching between different sperm behaviors (Achikanu et al., 2019).

Considering the time it takes a human spermatozoon to reach the site of fertilization, it seems worthwhile to develop efficient analytical tools that allow for very long recordings of freely swimming spermatozoa. However, the flagellar beating is fast, ranging from 10 to 25 Hz (Gaffney et al., 2011), thus requiring high-speed cameras to resolve the rapid movement of the tail (>100 frames per second, fps). Current computer-assisted semen analysis (CASA) systems are inadequate to this task (Mortimer et al., 2015). CASA systems can only capture sperm head trajectories, and as such, no flagellar beating information is accessible. Furthermore, the large data recordings resulting from such long periods of fast spermatozoa swimming are awkward to handle. To surmount these limitations, the majority of studies have focused on simplistic sperm head trajectories for short periods of time, which are of much lower frequencies. Even so, if a better temporal resolution is available, it is still a difficult task to measure flagellar parameters (Gadêlha et al., 2020).

Commonly, automated analyses of spermatozoa motility in clinics and for research are performed using CASA (Davis and Katz, 1996), and thus are limited to characteristics of sperm head movement only. Furthermore, CASA systems employ segmentation procedures and require sperm samples to be constrained to 2D surfaces (20 μm height), characterized only during very short periods of time (e.g. 1 s) and in small fields of view under the microscope. Other recently reported systems, such as SpermQ (Hansen et al., 2019) and FAST (Gallagher et al., 2019), have attempted to capture flagellar beating characteristics from sperm video-microscopy data, but these are also segmentation-based procedures. The overall accuracy of any segmentation-based method is, however, highly dependent on the characteristics and quality of the raw image. For example, FAST (Gallagher et al., 2019) can only be applied to images acquired with specific microscopy (negative phase contrast), which may not be widely available across different laboratories. Recently, a mathematical image-analysis development allowed accurate characterization of detailed flagellar kinematics of human sperm flagella in 3D, but only for few beat cycles (Gadêlha et al., 2020). In contrast, Achikanu et al. (2019), looking at long trajectories, showed that human spermatozoa incubated under capacitating conditions are able to modify their swimming mode 6 times/min on average and at least once every 3.5 min. This switching of the swimming gaits over long periods has probably evolved to contend with the changes occurring along the female genital tract.

In this work, we endeavored to solve the above-mentioned bottlenecks by devising a segmentation-free image analysis method that allows, for the first time, inspection of the dynamics of the head movement and flagellar beat over very long periods (9.2 min), corresponding to hundreds of thousands of flagellar beat cycles. This provides a robust alternative to segmentation-based methods bridging the gap between CASA and flagellar analysis systems. The strategy is based on local orientation and isotropy features of an image (Püspöki et al., 2016) enclosing the sperm head and flagellum of a swimming spermatozoon captured with high-speed video microscopy. This yields key head and flagella kinematic characteristics of a single spermatozoon swimming for long periods of time without the need for segmentation and post data analysis. These include (1) sperm head kinematics: head yawing amplitude, namely amplitude of lateral head displacement (ALHCASA) and the associated head-yawing frequency, traditionally referred to as ‘beat cross frequency’ (BCFCASA) in CASA standards; and (2) flagellar beating: flagellar beat amplitude (FBA) and associated frequency, referred to here as flagellar beat frequency (FBF). The head and flagellar characteristics obtained with this methodology were also compared with those estimated using segmentation-based systems, revealing high correlations.

We demonstrate the powerful use of our segmentation-free technique by examining how the release of Ca2+ from internal stores alters flagellar beating in the long term. Intracellular Ca2+ and its concentration changes have been implicated in the regulation of flagellar properties (Darszon et al., 2011; Strünker et al., 2015; Lishko and Mannowetz, 2018). In many cell systems it is known that the Ca2+ released from internal stores contributes significantly to the increase of intracellular Ca2+ concentration ([Ca2+]i) that regulates many cell signaling responses (Clapham, 2007; Putney, 2013). In this context, this work shows how head–flagellar characteristics are modified by emptying Ca2+ stores using two well-known Ca2+ regulators of these stores that inhibit their Ca2+-ATPase, thapsigargin (Andersen et al., 2015) and cyclopiazonic acid (Chang et al., 2009). The strategy permitted recording short intercalated intervals (avoiding cell damage) of a [Ca2+]i-sensitive dye-emitted fluorescence while measuring long-term swimming and flagellar characteristics, before and after applying these drugs to the same spermatozoon. Increasing [Ca2+]i in this manner affected by several fold the head yawing and flagellar beating frequencies and relative amplitude of the head (ALHCASA) in a convoluted manner.

The main advantages of the proposed methodology are that it avoids the substantial effort needed to segment and track flagella, showing high robustness to contend with different sources of microscope imaging, noise, light heterogeneity insensitivity and debris, as well as fast processing time of large datasets with tens of thousands of image frames. This makes the method easily applicable to an umbrella of bio-imaging applications, such as in CASA systems for sperm analysis, as well as studies of slender-bodied organisms and of flagellated microbes and cells, including, but not limited to, spermatozoa of any species. We have demonstrated the generality of the segmentation-free method by applying it to video data with varying microscope imaging quality from different laboratories using a range of mono-flagellates, such as bull sperm and Trypanosoma. Thus, this method may appeal to communities beyond the field of sperm motility research.

We implemented a computer-assisted segmentation-free method to obtain long-term kinematic and physiological parameters for up to 9.2 min from individual free-swimming spermatozoa that does not require flagella image segmentation, which is cumbersome in nature (see Fig. 1). As in many cellular systems, the release of Ca2+ from internal stores is well known to contribute significantly to [Ca2+]i homeostasis, thus driving specific signaling responses (Clapham, 2007; Putney, 2013). We demonstrated the powerful use of this technique by examining how spermatozoa swimming characteristics (head and flagellum movement) were modified by emptying the Ca2+ stores using two different store Ca2+ regulators: thapsigargin and cyclopiazonic acid.

Fig. 1.

Extraction of sperm flagellum kinematics without segmentation. (A) A free-swimming spermatozoon in 3D not confined in space. (B) Local orientation of the image extracted from grayscale gradients (pseudo-color corresponds to the local orientation angle relative to the x-axis; see Materials and Methods section: Segmentation-free spermatozoa orientation detection and tracking). (C) Measuring the spermatozoon head and flagella orientation based on the local average orientation θ(t) of the image at time frame t. The square is used to follow the head in consecutive time points, while the circle and rectangle are used to compute the sperm orientation. The red line represents the dominant orientation of the structure within the rectangle. (D) The segmentation-free tracking automatically ignores other cells and debris (see the Modified OrientationJ plugin section of Materials and Methods) only the square and rectangle ROIs are considered to obtain the motility and fluorescence measurements reported. Scale bar: 50 μm.

Fig. 1.

Extraction of sperm flagellum kinematics without segmentation. (A) A free-swimming spermatozoon in 3D not confined in space. (B) Local orientation of the image extracted from grayscale gradients (pseudo-color corresponds to the local orientation angle relative to the x-axis; see Materials and Methods section: Segmentation-free spermatozoa orientation detection and tracking). (C) Measuring the spermatozoon head and flagella orientation based on the local average orientation θ(t) of the image at time frame t. The square is used to follow the head in consecutive time points, while the circle and rectangle are used to compute the sperm orientation. The red line represents the dominant orientation of the structure within the rectangle. (D) The segmentation-free tracking automatically ignores other cells and debris (see the Modified OrientationJ plugin section of Materials and Methods) only the square and rectangle ROIs are considered to obtain the motility and fluorescence measurements reported. Scale bar: 50 μm.

Fig. 2 shows, for two swimming spermatozoa (control, red; pharmacologically perturbed case, black), the four extracted kinematic parameters and how [Ca2+]i behaves before and after applying thapsigargin (see the Computer-assisted segmentation-free measurements section of Materials and Methods for parameter definitions). Fig. 2A shows the angle θ(t) rescaled by 2π, as depicted in Fig. 1C, corresponding to the dominant orientation of the sperm as it swam. Here, head yawing and flagellar beat characteristics were extracted directly from θ(t) using Fourier transform analysis. A close-up of θ(t) for the small region marked with a black rectangle in Fig. 2A is shown in Fig. 2B. This unveils the characteristic two frequency peaks of θ(t) associated with head yawing and flagellar beating. Figs 2A and C show complex behavioral changes for both head and flagella oscillations before and after the administration of thapsigargin. The sperm flagellum hyperactivated as [Ca2+]i increased and then stalled when levels reached a plateau, as indicated by yellow and blue markers in Fig. 2F and Movie 1. The positive slope of the sperm orientation angle relative to the x-axis (Fig. 2A, θ) indicates anti-clockwise rotations of the sperm flagellum as it swam (Fig. 2F). In Fig. 2A, the orientation angle is divided by 2π, thus capturing the number of turns as the orientation of the swimming direction changed during the course of the experiment. The sperm associated with the black curve in Fig. 2A started roughly aligned with the x-axis at the beginning of the experiment and slowly rotated its swimming direction anti-clockwise, as also depicted in Fig. 2F for its swimming trajectory. This sperm completed almost two full rotations in 4.6 min (anti-clockwise, see Fig. 2F). This highlights how swimming sperm orientation varies tremendously during long time scales, thus correlating weakly with past swimming directions. This is indicative of intermittent dynamical beating asymmetries that are taking place over long swimming periods, thus facilitating sperm transport and increasing sperm diffusion in this timescale.

Fig. 2.

Motility parameters and [Ca2+]i changes for two free-swimming spermatozoa extracted with the segmentation-free method. (A) Head–flagellum orientation time series θ(t) rescaled by 2π for a control sperm (red) and a thapsigargin-treated sperm (black), illustrating several stages before and after thapsigargin addition (Thaps. addition instant; vertical dashed lines). After a significant delay following thapsigargin addition (∼75 s), the treated cell briefly displays a hyperactivated-type mode before stopping, as the rate of [Ca2+]i increase slows (see E) before approaching the plateau. Thereafter, when [Ca2+]i is no longer increasing during the plateau, the sperm re-initiates motility, displaying a hyperactivated-type mode of behavior that becomes more exacerbated as [Ca2+]i continues increasing. Box indicates region shown expanded in B. (B) Close-up of the head–flagellum orientation θ(t) trace (top), and the Fast Fourier transform (FFT; bottom) showing two principal peaks corresponding to head yawing and flagellar beating frequencies, defining BCFCASA and FBF, respectively, as depicted in C. The amplitude of the FFT spectrum providing ALHCASA and FBA are shown in D. (C) Head (dashed line) and flagellar (continuous line) beating frequencies for control (red) and thapsigargin-treated (black) spermatozoa as a function of time (both decreasing up to 200 s). (D) Head (dashed line) and flagellar (continuous line) beating amplitudes for control (red) and thapsigargin-treated (black) spermatozoa as a function of time (ALHCASA increasing up to 220 s while FBA remains near-constant). (E) [Ca2+]i followed by Fluo-8 fluorescence intensity as a function of time (increasing for thapsigargin-treated sperm, black; control sperm, red). (F) Trajectory of the spermatozoon treated with thapsigargin. Red point, tracking begins; green point, addition of thapsigargin; yellow point, when the flagellum stalls; blue point, the exact moment when the swimming gait changes moving away from the starting focal plane in 3D (see Movie 3). AU, arbitrary units.

Fig. 2.

Motility parameters and [Ca2+]i changes for two free-swimming spermatozoa extracted with the segmentation-free method. (A) Head–flagellum orientation time series θ(t) rescaled by 2π for a control sperm (red) and a thapsigargin-treated sperm (black), illustrating several stages before and after thapsigargin addition (Thaps. addition instant; vertical dashed lines). After a significant delay following thapsigargin addition (∼75 s), the treated cell briefly displays a hyperactivated-type mode before stopping, as the rate of [Ca2+]i increase slows (see E) before approaching the plateau. Thereafter, when [Ca2+]i is no longer increasing during the plateau, the sperm re-initiates motility, displaying a hyperactivated-type mode of behavior that becomes more exacerbated as [Ca2+]i continues increasing. Box indicates region shown expanded in B. (B) Close-up of the head–flagellum orientation θ(t) trace (top), and the Fast Fourier transform (FFT; bottom) showing two principal peaks corresponding to head yawing and flagellar beating frequencies, defining BCFCASA and FBF, respectively, as depicted in C. The amplitude of the FFT spectrum providing ALHCASA and FBA are shown in D. (C) Head (dashed line) and flagellar (continuous line) beating frequencies for control (red) and thapsigargin-treated (black) spermatozoa as a function of time (both decreasing up to 200 s). (D) Head (dashed line) and flagellar (continuous line) beating amplitudes for control (red) and thapsigargin-treated (black) spermatozoa as a function of time (ALHCASA increasing up to 220 s while FBA remains near-constant). (E) [Ca2+]i followed by Fluo-8 fluorescence intensity as a function of time (increasing for thapsigargin-treated sperm, black; control sperm, red). (F) Trajectory of the spermatozoon treated with thapsigargin. Red point, tracking begins; green point, addition of thapsigargin; yellow point, when the flagellum stalls; blue point, the exact moment when the swimming gait changes moving away from the starting focal plane in 3D (see Movie 3). AU, arbitrary units.

Fig. 2C shows how the head yawing and flagellum beating frequencies decreased after injection of thapsigargin (up to 64% and 19% for the head yawing and flagellum frequency, respectively). However, this change was not immediate, it occured after a significant time delay (∼75 s). In contrast, the control sperm (depicted in red) remained stable and only showed a large increase in beating frequency towards the end of the imaging period, demonstrating the potential for spontaneous self-switching, in agreement with previous observations (Achikanu et al., 2019). Fig. 2D shows that after decreasing to zero, when the flagellum stalled, the amplitude of the thapsigargin-treated sperm head lateral motion (arbitrary units) increased dramatically and very quickly by 242%, whereas that of the flagellum also increased, though mildly (22%), as Ca2+ plateaued (note that this captures the amplitude of the flagellar beat relative to the head, as detailed in the Computer-assisted segmentation-free measurements section of the Materials and Methods). Fig. 2E depicts [Ca2+]i and shows an increase of 16% by 45 s after administration of thapsigargin. This cell roughly displayed a hyperactive-type behavior, because both head yawing and flagellum beating frequencies decreased, whereas their relative amplitudes increased, at the same time that Ca2+ increased and plateaued (Chang and Suarez, 2011; Ooi et al., 2014). The red curve for the control sperm in Fig. 2E shows no significant change in [Ca2+]i nor in swimming behavior until ∼250 s, when only the beating frequency decreased (Fig. 2C). Fig. 2F illustrates the trajectory in the xy plane of the pharmacologically treated sperm shown in Fig. 2A (black curves), where a red dot indicates the beginning of the journey and a green dot marks inhibitor addition. Approximately one minute after the thapsigargin addition, the trajectory underwent significant changes. The sperm swam close to two full anti-clockwise circles during the 4.6 min, performing a total of 3850 beat cycles. This is depicted in Movie 3, which includes the evolving sperm trajectory. The sperm swimming behavior can be divided into three swimming gaits: (1) 0–120 s, the sperm swam in an activated progressive manner covering a long distance; (2) 120–140 s, after a short period of hyperactivated-type mode (6 s) the sperm abruptly stopped and thereafter initiated rapid flagellar shape changes with high curvature (yellow point in Fig. 2F and Movie 3), and (3) 159–277 s, the sperm clearly switched to a hyperactivated gait (blue point in Fig. 2F and Movie 3), characterized by a decrease in flagellar beating frequency and progressivity, and an increase of its amplitude (Fig. 2C,D). At this stage the sperm swam more randomly, oscillating between straight-swimming and turning periods due to vigorous whip-like flagellar movements. This is apparent in the jittery trajectory after the blue mark in Fig. 2F and at 160 s in Movie 3. It is worth noting that this sperm was actually swimming in three dimensions, moving away from the starting focal plane. For xy recording purposes, this was compensated by refocusing the microscope as required during the experiment. Fig. 2F thus constitutes an xy projection of the 3D sperm movement.

Figs 3 and 4 summarize the effects of emptying Ca2+ stores via the application of thapsigargin and cyclopiazonic acid. A total of 88 spermatozoa were analyzed for periods of up to 9.2 min (4.6 min/9.2 min, bright field/intercalated fluorescence; see the Bright field and intercalated fluorescence image acquisition section of Materials and Methods). In order to contrast the behavior before and after applying the Ca2+ modulators, we employed two controls (see the Spermatozoa samples and statistical analysis section of Materials and Methods). The first one was provided by the same sperm 45 s before adding the inhibitors, and its basal properties were compared to the rest of the time after inhibitor addition. Twenty-two spermatozoa were treated with thapsigargin and twenty-two with cyclopiazonic acid. The second control consisted of recordings without applying Ca2+ modulators to document the effect of long-term swimming itself, where we corroborated that time itself does not produce significant statistical changes in frequency or beating amplitude (26 cells). In our experimental data set we found head and flagellar frequency variations of 0.7–7 Hz and 6–27 Hz, respectively, with relative amplitudes increasing up to 4-fold from their control value.

Fig. 3.

Effect on emptying Ca2+ stores on sperm kinematic parameters. (A–D) Emptying calcium stores using thapsigargin (Thaps., 22 cells) or cyclopiazonic acid (Cyclo., 22 cells) produces statistically significant changes in three of the four kinematic parameters analyzed, compared with control untreated conditions, as shown by Cohen's d effect size and Kruskal–Wallis test (see Eqns 1 and 2). Only the relative amplitude of the flagellar beat (FBA) did not show a statistically significant difference. (A) Head yawing frequency (beat cross frequency, BCFCASA). (B) Flagellar beat frequency (FBF). (C) Head yawing amplitude (relative amplitude of lateral head displacement, ALHCASA). (D) Relative flagellar beat amplitude (FBA). Boxes show the interquartile range (IQR) with the median indicated by a notch of width ±1.58×IQR/√n. Whiskers show the maximum and minimum values.

Fig. 3.

Effect on emptying Ca2+ stores on sperm kinematic parameters. (A–D) Emptying calcium stores using thapsigargin (Thaps., 22 cells) or cyclopiazonic acid (Cyclo., 22 cells) produces statistically significant changes in three of the four kinematic parameters analyzed, compared with control untreated conditions, as shown by Cohen's d effect size and Kruskal–Wallis test (see Eqns 1 and 2). Only the relative amplitude of the flagellar beat (FBA) did not show a statistically significant difference. (A) Head yawing frequency (beat cross frequency, BCFCASA). (B) Flagellar beat frequency (FBF). (C) Head yawing amplitude (relative amplitude of lateral head displacement, ALHCASA). (D) Relative flagellar beat amplitude (FBA). Boxes show the interquartile range (IQR) with the median indicated by a notch of width ±1.58×IQR/√n. Whiskers show the maximum and minimum values.

Fig. 4.

Percentage of total time that free-swimming spermatozoa spent in three categories. (A–D) Values for each of four evaluated kinematic parameters – BCFCASA, FBF, ALHCASA and FBA – were obtained for the control (TIME CONTROL), cyclopiazonic acid-treated (CYCLO) and thapsigargin-treated (THAPS.) spermatozoa and placed into three determined categories: (1) in the control range (mean±2 s.d.; orange), (2) greater than 2 s.d. above the mean (yellow) and (3) greater than 2 s.d. below the mean (blue). The percentage of the total time that each parameter spent in each of the three categories is plotted before and after treatment. The left bars for CYCLO and THAPS pairs correspond to the first 45 s before application of the calcium modulators (control) and the right bars for each pair correspond to the rest of the time (up to 4.6 min) after their application. TIME CONTROL bars correspond to cells where no drugs were administrated. *P<0.05 (Kruskal–Wallis test). (A) Head yawing frequency (beat cross frequency, BCFCASA). (B) Flagellar beat frequency (FBF). (C) Head yawing amplitude (relative amplitude of lateral head displacement, ALHCASA). (D) Flagellar beat amplitude (FBA).

Fig. 4.

Percentage of total time that free-swimming spermatozoa spent in three categories. (A–D) Values for each of four evaluated kinematic parameters – BCFCASA, FBF, ALHCASA and FBA – were obtained for the control (TIME CONTROL), cyclopiazonic acid-treated (CYCLO) and thapsigargin-treated (THAPS.) spermatozoa and placed into three determined categories: (1) in the control range (mean±2 s.d.; orange), (2) greater than 2 s.d. above the mean (yellow) and (3) greater than 2 s.d. below the mean (blue). The percentage of the total time that each parameter spent in each of the three categories is plotted before and after treatment. The left bars for CYCLO and THAPS pairs correspond to the first 45 s before application of the calcium modulators (control) and the right bars for each pair correspond to the rest of the time (up to 4.6 min) after their application. TIME CONTROL bars correspond to cells where no drugs were administrated. *P<0.05 (Kruskal–Wallis test). (A) Head yawing frequency (beat cross frequency, BCFCASA). (B) Flagellar beat frequency (FBF). (C) Head yawing amplitude (relative amplitude of lateral head displacement, ALHCASA). (D) Flagellar beat amplitude (FBA).

To contrast the differences before and after applying the Ca2+ modulators, we determined the Cohen's d effect size and applied the Kruskall–Wallis test for distributions that were not normal (see Materials and Methods, Eqns 1 and 2). Fig. 3 shows that emptying the Ca2+ stores had a significant effect in three of the four evaluated motility parameters, that is BCFCASA, FBF and ALHCASA (see the Computer-assisted segmentation-free measurements section of Materials and Methods for parameter definitions). Although increases were observed in the FBA after administration of thapsigargin and cyclopiazonic acid, they were not statistically significant.

There was a large variance in spermatozoa swimming behavior, thus in Fig. 4 we show the relative total time when the cell swimming behavior was above or under two standard deviations (s.d.) of the control conditions. The yellow intervals for each condition individually (pre-addition or post-addition) correspond to the total percentage of time where the kinematic cell parameters were more than two times the s.d. greater than their mean, and the blue intervals show the percentage of time that the parameters were more than two times the s.d. below their mean. Time spent within two s.d. for each condition individually is shown in orange. The left column from each pair corresponds to the control condition before the application of cyclopiazonic acid or thapsigargin, while the right column from each pair depicts the behavior after application of treatment – except for the time control columns.

It can be observed that the behavior during the control periods before treatment (Fig. 4, left columns of each pair) was quite symmetric, from the point of view that the time spent above and below two s.d. were similar (blue and yellow regions in Fig. 4). In contrast, when cyclopiazonic acid or thapsigargin were applied, the total time for both head yawing (BCFCASA) and flagellar beating frequencies (FBF) decreased, as compared with control conditions (blue regions increase in Fig. 4A,B, right columns for treated cells). Inversely, the total time of the relative amplitudes of the head yawing (ALHCASA) and flagellar beating (FBA) increased (yellow areas increase in Fig. 4C,D right columns for treated cells). Significant statistical differences (using Wilcoxon test) for different ranges are marked with an asterisk in Fig. 4. Notably, in this analysis, the total time of the amplitude of the flagellar beating (FBA) increased in a statistically significant manner.

Fig. 5 shows that after applying thapsigargin to empty the Ca2+ stores, sperm [Ca2+]i increased. This is in agreement with previous reports on [Ca2+]i (Meizel and Turner, 1993; Williams and Ford, 2003), and illustrates the capability of our methodology to record and accumulate statistics of [Ca2+]i during very long swimming periods.

Fig. 5.

After applying thapsigargin to empty Ca2+ stores, sperm [Ca2+]i increases. Box plots showing Cohen's d effect size for changes in Fluo-8 fluorescence intensity for individual cells (untreated control, first 45 s) and thapsigargin-treated (after 46 s) spermatozoa (n=18), illustrating the capability of our methodology to record and accumulate statistics of [Ca2+]i during very long swimming periods. Boxes show the interquartile range (IQR) with the median indicated by a notch of width ±1.58×IQR/√n. Whiskers show the maximum and minimum values.

Fig. 5.

After applying thapsigargin to empty Ca2+ stores, sperm [Ca2+]i increases. Box plots showing Cohen's d effect size for changes in Fluo-8 fluorescence intensity for individual cells (untreated control, first 45 s) and thapsigargin-treated (after 46 s) spermatozoa (n=18), illustrating the capability of our methodology to record and accumulate statistics of [Ca2+]i during very long swimming periods. Boxes show the interquartile range (IQR) with the median indicated by a notch of width ±1.58×IQR/√n. Whiskers show the maximum and minimum values.

Comparison with segmentation-based methods and application to other flagellated microorganisms

We compared our results with those obtained using available segmentation-based systems, such as SpermQ (Hansen et al., 2019) for flagellar and head kinematics, and OpenCASA (Alquézar-Baeta et al., 2019) for head kinematics. FAST (Gallagher et al., 2019) can only be applied to videos acquired with negative phase-contrast microscopy, thus it was unable to process our video data. We note that the segmentation-free method described here does not require a specific type of microscopy imaging. Due to uneven contrast, illumination, noise and existing debris in the background of our images, both SpermQ and OpenCASA presented difficulties while tracking the sperm motion under these conditions, which in turn dramatically affected the processing time. SpermQ required more than 40 h to process a single video-microscopy recording containing 28,000 images, while our algorithm processed the same video data in less than 2 min. This demonstrates an important advantage of the developed segmentation-free procedure, which is able to robustly process non-ideal and noisy images with a fast processing time, especially relevant when working with videos with many thousands of image frames. Movie 4 illustrates how background noise and debris affects the segmentation output in comparison with our segmentation-free algorithm.

The correlations of kinematic parameters using SpermQ and OpenCASA with the parameters extracted with the segmentation-free method are displayed in Fig. 6. Fig. 6A and B show a high correlation for both BCFCASA and FBF, with correlation values of 0.99 and 1, respectively, for the 15 cells evaluated. Regarding ALHCASA, Mortimer et al. (2015) pointed out existing challenges in the different ways to evaluate this parameter, making it difficult to standardize between distinct algorithms in the literature. We compared ALHCASA values obtained using the segmentation-free method with those obtained using OpenCASA. Fig. 6C shows a strong correlation of 0.84 with OpenCASA for ALHCASA. We also note that previous studies using OpenCASA found a correlation of 0.88 with respect to a commercial CASA system (Alquézar-Baeta et al., 2019).

Fig. 6.

Comparison between the segmentation-free and segmentation-based methods, showing strong correlations for both head and flagellar kinematic parameters for sixteen spermatozoa. (A) Head yawing frequency (beat cross frequency, BCFCASA), compared with values obtained using SpermQ (Hansen et al., 2019). Pearson's r=0.98. (B) Flagellar beat frequency (FBF), compared with values obtained using SpermQ. Pearson's r=0.99. (C) Head yawing amplitude (amplitude of lateral head displacement, ALHCASA), compared with OpenCASA measurements (Alquézar-Baeta et al., 2019). Pearson's r=0.84. (D) Flagellar beat amplitude (FBA), compared with direct measurement from segmented waveform using SpermQ. Pearson's r=0.84. The numbers labeling points in A and B indicate the number of cells with the same value. AU, arbitrary units.

Fig. 6.

Comparison between the segmentation-free and segmentation-based methods, showing strong correlations for both head and flagellar kinematic parameters for sixteen spermatozoa. (A) Head yawing frequency (beat cross frequency, BCFCASA), compared with values obtained using SpermQ (Hansen et al., 2019). Pearson's r=0.98. (B) Flagellar beat frequency (FBF), compared with values obtained using SpermQ. Pearson's r=0.99. (C) Head yawing amplitude (amplitude of lateral head displacement, ALHCASA), compared with OpenCASA measurements (Alquézar-Baeta et al., 2019). Pearson's r=0.84. (D) Flagellar beat amplitude (FBA), compared with direct measurement from segmented waveform using SpermQ. Pearson's r=0.84. The numbers labeling points in A and B indicate the number of cells with the same value. AU, arbitrary units.

OpenCASA (Alquézar-Baeta et al., 2019) provides sperm head trajectories and associated statistics, but no flagellar beating information. SpermQ (Hansen et al., 2019) extracts the flagellar waveform via segmentation procedures, but the FBA is not included as an output. For this, we measured the FBA directly from the flagellar waveform via the flagellar ‘head centerline deviation’, (Smith et al., 2009). This captures the wave amplitude relative to the sperm head orientation. Fig. 6D shows a strong correlation of 0.84 between the flagellar wave amplitude segmented directly from the flagellar waveform and the image orientation signal θ(t) derived from the segmentation-free method.

To demonstrate the wider applicability of the segmentation-free method to other flagellated microorganisms, we successfully processed and captured the orientation signal to track both body and flagellum kinematics from different laboratory video data available in the literature for Trypanosoma (Gadêlha et al., 2007) (Movie 5) and bull spermatozoa (Kantsler et al., 2014) (Movie 6). Movies 5 and 6 show that the segmentation-free approach does not require specific microscopy type, and the segmentation-free principle bypasses challenges associated with imaging quality, background noise, varying illumination, cell morphology and swimming direction. Indeed, Trypanosoma possess a larger body and swim with the flagellum pulling the body (Movie 5), in contrast with human sperm, in which the flagellum pushes a small head to swim. Movie 6 shows the segmentation-free method applied to bull sperm, which share similar morphological and kinematic traits with human sperm. Thus, we foresee that the method presented can be applied and generalized to spermatozoa from other species.

The mechano-chemical micro-environment of the reproductive tract is complex. Spermatozoa must contend with dramatic physiological alterations in the ionic composition (Ng et al., 2018), temperature (Bedford, 2015), viscosity (Smith et al., 2009) and fluid flow conditions (Gaffney et al., 2011; Ishimoto et al., 2017, 2018) as they undergo capacitation (Gervasi and Visconti, 2016) and may, in certain regions of the female reproductive tract, undergo rheotaxis (Miki and Clapham, 2013) and chemotaxis (Eisenbach and Giojalas, 2006) to be able to reach the site of fertilization (reviewed in Darszon et al., 2011, 2020). The ability of spermatozoa to coordinate head yawing and flagellar beating over large distances and long swimming periods is thus critical for a successful fertilization. Against this background, the large majority of studies to date devoted to understanding sperm motility are critically limited to very short periods of observation. In this context, developing novel tools to record and process complex video microscopy imaging of sperm swimming for long time periods is essential.

CASA derives primitive statistics of head trajectories only by imaging samples over short periods of time (e.g. 1 s) with a small field of view, and only for spermatozoa constrained to move in two dimensions (CASA chambers have 10–20 micrometers depth; Davis and Katz, 1996; Mortimer et al., 2015; Gallagher et al., 2018). No direct flagellar beating parameter is assessed by CASA systems. On the other hand, high-precision spatio-temporal flagellar tracking is equally constrained to short-term analysis (Smith et al., 2009; Hansen et al., 2019; Gallagher et al., 2019; Ishimoto et al., 2017). One of the difficulties is that the flagellar beating occurs at a much faster rate (10–20 Hz) than sperm head lateral movement, requiring high-speed digital cameras (>100 fps) to resolve and track the flagellar beat. This requirement limits the total recording period to typically very few swimming strokes (Smith et al., 2009; Ishimoto et al., 2017). Even when flagellar waveform data is available, extraction of key flagellar kinematic parameters, such as amplitude and frequency of the beat, are cumbersome in 2D (Hansen et al., 2019; Gallagher et al., 2019) and especially so in 3D (Gadêlha et al., 2020).

Swimming patterns and long-term behavior of freely swimming human spermatozoa remain elusive in the literature (Achikanu et al., 2019). Are, for example, swimming patterns observed over short periods representative of subpopulations with distinct motility traits, or instead, are individual sperm capable of generating multiple swimming gaits when observed over long periods? Furthermore, what is the relationship between the motility information measured in short-term experiments and that from their long period counterparts? All these questions are now receiving more attention as the multifaceted ability of spermatozoa to adapt the flagellar behavior over the course of long periods of time is unveiled (Achikanu et al., 2019).

Bio-imaging segmentation is at the heart of many automated computer-assisted image analysis systems. The most recent systems implementing a range of image segmentation strategies include SpermQ (Hansen et al., 2019) and FAST (Gallagher et al., 2019) for flagellar waveform analysis and OpenCASA (Alquézar-Baeta et al., 2019) for sperm head trajectory analysis. However, segmentation-based methods require thresholding of the pixel intensity of the image, and thus are sensitive to the microscopy used, imaging quality, background noise, defocusing and illumination effects, among other factors. This sensitivity limits the general applicability of thresholding methods, often requiring a high level of algorithmic customization to fit the specific needs of each laboratory.

The kinematic relation between head movement and tail bending is subtle and cannot be used interchangeably – that is, one cannot measure flagellar amplitude and/or frequency from head kinematics or vice-versa (David et al., 1981; Mortimer, 1997). Here, we solved a critical bottleneck in sperm motility and flagellar analysis. We developed a segmentation-free analysis system that allows the inspection of sperm head movement, flagellar beating and intracellular calcium levels of individual cells over long periods of time (up to 9.2 min). Sperm beating can thus be analyzed over tens of thousands of flagellar beat cycles, as required during long swimming distances within the female reproductive tract. The segmentation-free analysis is based on the local orientation and anisotropic features of an image (Püspöki et al., 2016), and can be easily extended to general video-microscopy imaging data of any duration from other mono-flagellated microorganisms, or indeed slender-bodied organisms. This generality was demonstrated by applying the segmentation-free method to video data from other laboratories for Trypanosoma (Gadêlha et al., 2007; see Movie 5) and bull sperm (Kantsler et al., 2014; see Movie 6) with varying image quality, cell morphology, microscopy type and acquisition, among other factors.

Our segmentation-free method provides a temporal series representing the orientation of the swimming sperm at each timeframe. As such, the orientation reduces the complexity and dimensionality of the data to a one-dimensional signal (a function of time only) and embodies both head and flagellum kinematic parameters, avoiding complicated post-processing to extract kinematic features. Both head and flagellar kinematic parameters can be directly disentangled from the one-dimensional orientation signal using simple Fourier decomposition. This bypasses convoluted spatial-temporal flagellar tracking and analysis (Gadêlha et al., 2020; Werner et al., 2014). Thus, in addition to the head-trajectories and swimming directionality, the frequency and amplitude of both head lateral oscillation and flagellar beating movement are also obtained with this strategy.

We investigated the role of emptying calcium stores during long swimming periods in human spermatozoa. In many cellular systems, the release of Ca2+ from internal stores is known to contribute significantly to the overall [Ca2+]i and its changes, thus driving specific signaling responses (Clapham, 2007; Putney, 2013). Changes in the sperm [Ca2+]i are implicated in the regulation of flagellar beating (Darszon et al., 2011; Strünker et al., 2015; Lishko and Mannowetz, 2018), and they are influenced by Ca2+ release and re-uptake from internal stores such as the acrosome and the residual nuclear vesicles (Chang and Suarez, 2011; Correia et al., 2015). Ion channels, for instance inositol triphosphate (IP3) receptors (Prole and Taylor, 2019) and ryanodine receptors (Ogawa et al., 2020), able to release Ca2+ from internal stores, have been located to both of these sperm internal stores (reviewed in Correia et al., 2015). In many somatic cell types, store Ca2+ mobilization is triggered by intracellular messengers like cADPR, NAADP and IP3 (Galione and Chuang, 2020). This has been documented to occur in mammalian sperm (Correia et al., 2015). Ca2+ release can also be induced by inhibiting the Ca2+-ATPases of these stores, as we have done in the present work, or by Ca2+ waves triggered by external ligands, such as progesterone in human sperm (Blackmore, 1999; Garcia and Meizel, 1999; Harper et al., 2004). As [Ca2+]i increases in the vicinity of these stores, it triggers what is known as Ca2+-induced Ca2+ increases by activating ryanodine receptors and IP3 receptors (Parkash and Asotra, 2012).

It had been shown previously that releasing Ca2+ from internal stores using thimerosal, an activator of IP3 receptor and ryanodine receptor channels (Elferink, 1999; Chang and Suarez, 2011), induces hyperactivation in mouse and human sperm. Notably, thimerosal can induce hyperactivation in the absence of external Ca2+, clearly indicating the effect is due to Ca2+ release from internal stores (Marquez et al., 2007; Alasmari et al., 2013). Thapsigargin has been used in several mammalian sperm to induce Ca2+ release from internal stores; in many of them it is able to induce hyperactivation, and in some it can induce the acrosome reaction (reviewed in Correia et al., 2015).

Here, we found that elevating [Ca2+]i by emptying the human sperm internal Ca2+ stores using two different store Ca2+-ATPase inhibitors, thapsigargin and cyclopiazonic acid, significantly decreased head yawing and flagella beating frequencies and increased the head yawing amplitude in a statistically significant manner. As would be expected for hyperactivation inducers, the amplitude of the flagellar beat also increased, but without reaching significance. However, this latter parameter did significantly increase when the relative total time that this parameter spent above two standard deviations of the control condition was determined. These findings are in agreement with most previous reports of hyperactivation stimulatory conditions mentioned earlier and reported in the literature (Ho and Suarez, 2001, 2003; Chang and Suarez, 2011; Correia et al., 2015). There are published exceptions, such as Rossato et al. (2001), where 10 µM thapsigargin and up to 100 µM cyclopiazonic do not affect motility (see also Vijayaraghavan et al., 1994; Williams and Ford, 2003). It is worth pointing out that few studies that use cyclopiazonic acid have been reported. Our findings highlight the usefulness of our new segmentation-free strategy to study and correlate sperm swimming properties and [Ca2+]i, a matter that is fundamental and still requires further research and analysis.

In summary, CASA kinematic parameters have been typically derived from short ∼1 s intervals, assuming that sperm behavior is stable for long periods; however, recent evidence indicates kinematic parameters are continuously changing and dynamically adapting. The novel methodology presented here based on the local orientation and isotropy of individual free-swimming spermatozoa allows extraction of long-term kinematic and physiological parameters (up to 9.2 min) without the need of bio-image segmentation. We have demonstrated that the sperm head yawing and flagellar beat may have important variations in frequency and amplitude over time. Furthermore, our strategy is compatible with intercalated fluorescence recordings able to follow associated [Ca2+]i changes that occur upon sperm swimming mode switching. We detected significant alterations in human sperm head yawing and flagellar beat characteristics upon emptying their Ca2+ stores. This was possible because the strategy permitted, while recording long-term head yawing and flagellar beat characteristics before and after applying internal store [Ca2+]i-releasing drugs to the same spermatozoon, intercalating short fluorescence measurements (avoiding cell damage) from a [Ca2+]i-sensitive dye.

Our segmentation-free procedure presents significant advantages as compared with segmentation-based protocols. Besides avoiding the substantial effort needed to segment the flagella, the strategy is insensitive to noisy background and artifacts in the images, as well as being significantly tolerant to defocusing and heterogeneous background illumination, and still allowing fast processing times of large data (less than 2 min). The parameters obtained with this methodology were compared with those estimated using other systems, observing high correlations. Notably, this new strategy may be applied to spermatozoa from other species and other swimming microorganisms, such as mono-flagellated microbes, for instance Trypanosoma (Gadêlha et al., 2007), and even slender-bodied organisms, for example Caenorhabditis elegans (Ding et al., 2019) by exploiting its overall orientation during motion. Thus, this method may appeal to research communities away from sperm physiology and reproductive biology.

Although the segmentation-free method is able to estimate both head movement and flagellar beating parameters without post-processing and convoluted analysis, detailed spatial-temporal flagellar shape information is not available through this method. Robust characterization of flagellar wave-number parameters from long video-microscopy data remains an open challenge, as it would require precise waveform information at all times, only possible with segmentation-based methods. We thus hope that this study will motivate future work on the cross-fertilizing interactions between segmentation-free and segmentation-based methods for robust and effective multiparameter characterization of sperm beating from head to tail.

Ethical approval for the spermatozoa samples

The Bioethics committee of the Institute of Biotechnology, UNAM approved the proposed protocols for the human spermatozoa sample handling. The donors were properly informed regarding the experiments that would be performed and a consent form was signed and agreed by each donor. All World Health Organization requirements were fulfilled in this study.

Biological preparations and dye loading

After a minimum period of 48 h of sexual abstinence, healthy donors masturbated and human spermatozoa samples were collected. Highly motile spermatozoa were selected by a 1 h swim-up protocol [Ham's F-10 medium (Sigma Aldrich, USA) at 37°C in an atmosphere of 5% of CO2 and 95% air]. The collected cells were centrifuged at 470 g for 5 min and resuspended at a concentration of 107 cells/ml in a physiological solution that contained (mM): 120 NaCl, 4 KCl, 2 CaCl2, 1 MgCl2, 25 NaHCO3, 5 glucose, 30 HEPES and 10 lactate (pH 7.4). To measure [Ca2+]i, cells were incubated in a medium containing the Ca2+-sensitive Fluo-8-AM dye at 10 µM for 60 min and then washed with the same medium without dye once by centrifugation for 5 min at 470 g to remove the remaining external dye. Motility controls were performed using CASA, comparing non-loaded and loaded non-capacitated human spermatozoa. After testing four independent donors, we found there was no significant difference in the main motility parameters. Thapsigargin (5 μM) and cyclopiazonic acid (5 μM) were administered to examine how the release of Ca2+ from internal stores affected the long-term motility of individual spermatozoon.

Spermatozoa samples and statistical analysis

A total of 88 spermatozoa were analyzed for periods of up to 9.2 min [4.6 min for bright field (BF) or 9.2 min for intercalated fluorescence]. Twenty-two cells were recorded before and after thapsigargin treatment, and 22 were recorded before and after cyclopiazonic acid treatment, for kinematic analysis. Eighteen cells were used for fluorescence analysis. We performed two types of controls: (1) for the same individual spermatozoon, the first 45 s without the application of thapsigargin or cyclopiazonic acid was compared with the recorded behavior after being exposed to the Ca2+ modulators; (2) twenty-six spermatozoa were recorded without applying Ca2+ modulators (long term time control) for 4.6 min (BF). To determine the effect size and its statistical significance between spermatozoa characteristics before and after applying the Ca2+ modulators, we measured the Cohen's d effect size (Eqns 1 and 2) and Kruskal–Wallis tests for non-normal populations. Cohen's d effect size can be used to compare two means and is defined as:
formula
(1)
formula
(2)

being the ratio between the difference of means before (M1) and after (M2) Ca2+ modulator application divided by their corresponding pooled standard deviation (SDpooled) calculated from the standard deviations of means before and after, respectively, (SD1) and (SD2). For d=1 or 0.5, the two group means differ by one or by a half standard deviation, respectively. Cohen proposed that a small, medium or large effect size, corresponds roughly to d= 0.2, 0.5 or 0.8, respectively (Nakagawa and Cuthill, 2007).

Experimental setup

Experiments were performed with an Olympus IX71 (Olympus America Inc., USA) inverted optical microscope configured for bright field illumination (for head and flagellum kinematic analysis) with a LUCPlanFLN 40×/0.6 NA objective (Olympus America Inc., USA). A UIS-2 LUMPLFLN 60×/1.00 NA water immersion objective (Olympus America Inc., USA) was used for fluorescence analysis. A 49011 Fluo cube filter (Chroma Technology Corporation, USA) and a high intensity excitation LED M490D2 (Thorlabs, USA) were employed in these experiments. To keep individual spermatozoon in the field of view swimming freely in 3D we utilized a motorized xy stage (Märzhäuser Wetzlar GmbH & Co. KG, Germany) driven by a LUDL 5000 controller (Ludl Electronic Products, Ltd., USA) in combination with the mouse pointer, using a bespoke Arduino UNO platform (Evans, 2008) and Java algorithms (Oracle, USA), from which the xy stage position coordinates corresponding to the spermatozoa displacement were recorded. The motorized stage is not essential to image spermatozoa swimming. A manual stage can be used instead to keep an individual spermatozoon in the field of view. We used a high-speed camera NAC Q1v (Nac Americas, Inc., USA) with 8 GB RAM (recording up to 4.6 min at 100 fps with a spatial resolution of 640×480 pixels for bright field experiments and 9.2 min at 50 fps for intercalated fluorescence experiments). A TMC optical table (GMP SA, Switzerland) shielded the optical system from external vibration. A temperature of 37°C was kept constant with a thermal controller TCM/CL-100 (Warner Instruments LLC, USA). Data acquisition and image analysis was conducted with an Intel Core i7-6700 CPU at 3.4 GHz, 32 GB RAM processor (Intel Corporation, USA).

Image acquisition and segmentation-free analysis

Bright field and intercalated fluorescence image acquisition

Spermatozoa at a low density (∼102 cells/ml) were placed in a Chamlide CMB chamber having an 18 mm diameter coverslip at the bottom, which was mounted on the microscope stage, and temperature controlled at 37°C. Individual spermatozoa swimming at the focal plane of the coverslip were randomly selected for analysis after preparation and dye loading. Once a spermatozoon was selected, the microscope stage was moved to keep the spermatozoon in the field of view for the duration of the experiment. This was achieved by following the cell with a mouse pointer that controlled the microscope stage and a focus motor for 4.6 min, acquiring images with the digital camera at 100 fps (for bright field) and for 9.2 min at 50 fps (for intercalated bright field–fluorescence experiments to have a higher camera light integration time). A total of 28,000 images were acquired in this fashion independently of the illumination mode.

For intercalated fluorescence image analysis, we implemented an electronic switch based on an Arduino UNO platform (Evans, 2008) alternating between bright field illumination and fluorescence. To acquire fluorescence information during 9.2 min without bleaching or damaging spermatozoa, fluorescence was sampled for 0.2 s every 4.8 s of bright field illumination (Fig. 7). For both bright field and fluorescence imaging, the first 45 s were recorded as a control without the Ca2+ regulatory compounds. After 45 s, thapsigargin or cyclopiazonic acid were administrated with a pipette, while recording continued for a given individual spermatozoon. Control experiments were also conducted without any drug administration for 4.6 min for bright field experiments. A total of 88 free-swimming spermatozoa were recorded and tracked, 26 for time control (no drugs), 22 for thapsigargin, 22 for cyclopiazonic acid and 18 for fluorescence.

Fig. 7.

Fluorescence versus bright field sampling intervals. Each 4.8 s, bright field (BF) capture of the flagellar movement is switched to epifluorescence (Fluo) for 0.2 s to acquire fluorescence intensity information – namely, intercalated fluorescence. These imaging intervals were used for up to 9 min (at 50 fps for a higher cameras light integration time), recording 28,000 images with the interlaced information.

Fig. 7.

Fluorescence versus bright field sampling intervals. Each 4.8 s, bright field (BF) capture of the flagellar movement is switched to epifluorescence (Fluo) for 0.2 s to acquire fluorescence intensity information – namely, intercalated fluorescence. These imaging intervals were used for up to 9 min (at 50 fps for a higher cameras light integration time), recording 28,000 images with the interlaced information.

Image stack pre-processing

The image sequences (bright field microscopy) firstly needed to be pre-processed. This was done with standard FIJI (Schindelin et al., 2012) commands as follows: (1) Project the original image stack along the axis perpendicular to image plane (z-project with average intensity). The resulting projection contains non-moving artifacts and the background (even or uneven) illumination. (2) Invert both the original stack and its corresponding averaged z projection. (3) Subtract from original stack the averaged z projection to eliminate non-moving artifacts and uneven background illumination. (4) Enhance the contrast by adjusting the brightness/contrast (use the automatic mode) and apply the improved look-up tables. (5) Apply a median filter with a radius kernel of 2 or 4 pixels to eliminate impulsive noise and artifacts. At this point, the stack is ready to be processed by the modified OrientationJ plugin (Püspöki et al., 2016). It is worth mentioning that the pre-processed image stack is composed of gray-level 8-bit images and that no thresholding nor segmentation procedures were applied.

Segmentation-free fluorescence measurements to monitor Ca2+ over long periods.

[Ca2+]i was monitored via Fluo-8 fluorescence by acquiring 0.2 s samples every 4.8 s. This sampling rate allowed recording [Ca2+]i for long periods before and after the application of Ca2+-regulatory compounds without bleaching or damaging the spermatozoon. At the same time, this sampling rate was sufficiently short for long periods of Ca2+ evaluations, with a time resolution of 5 s, sufficient to obtain quantitative kinetic parameters from Fourier transform analysis. Fig. 7 illustrates the sampling periods and the switch between bright field and epifluorescence illumination.

Computer-assisted segmentation-free measurements

The critical advantage of the implemented method is the segmentation-free principle. This means that it is not necessary to segment the flagella in the video microscopy files to extract kinematic information. Instead, it is obtained directly from raw imaging data. Additional important advantages are the insensitivity to defocusing and inhomogeneous illumination. Our method is based on the local orientation and isotropy features of an image (Schindelin et al., 2012; Püspöki et al., 2016), which can be used to extract the ‘dominant orientation’ of a single swimming spermatozoon in the image (see details in next section: Segmentation-free spermatozoa orientation detection and tracking). The image orientation effectively combines the information from the angle of sperm head and the sperm flagellum. As a result, round objects and debris are automatically discarded due to their mostly isotropic aspect ratios with no dominant direction. Fig. 1 shows a spermatozoon in the microscope field of view and its flagella pseudo-colored according to the local image orientation (in degrees) relative to the x axis. The dominant direction of one image frame corresponds to the average direction of all orientations detected from the grayscale gradients in each pixel.

While the method works satisfactorily with one single cell in the field of view (for low density samples), we have modified and extended the functionality of OrientationJ (Püspöki et al., 2016) for multiple cells in the same field of view. This provides a robust dominant orientation detection avoiding the influence of debris and other cells when working with higher cell densities. This was achieved using the sperm orientation in an image frame to predict the next region of interest where the sperm has a greater probability to be found. By selecting the region of interest with a mouse in the first frame of the video, and continuous iteration of the procedure above, our method allows the continuous tracking of individual spermatozoa by only measuring the dominant orientation locally (see details in next section: Segmentation-free spermatozoa orientation detection and tracking). This is achieved, once again, without requiring segmentation or thresholding procedures so common for tracking moving-object video images. In this way, the dominant orientation of the spermatozoon image, including its flagellum, is obtained by removing any image orientation bias caused by nearby swimming cells (see Fig. 1D and Movie 1 for over 250 s of duration). Movie 2 shows the robustness of the algorithm for a very noisy background in the image and scattered debris (see t=180 s and 260 s). With this procedure, we extracted swimming and flagellar kinematics directly from each raw video recording with durations of 4.6 and 9.2 min for bright field and intercalated fluorescence, respectively. This is described at the end of this section and also depicted in Figs 1 and 7. Frequency and amplitude measurements were calculated directly with windowed Fast Fourier transforms of the image orientation signal measured as time advanced (see details in next section: Segmentation-free spermatozoa orientation detection and tracking).

The segmentation-free method provides a direct measurement of the principal orientation of the image containing the spermatozoon as a whole for each timeframe. As such, the orientation angle captures the combined effect of both head lateral movement and the shape of the flagellum. Thus, the method provides a time series of the instantaneous orientation angle θ(t) of the spermatozoon image at time t, relative to the x-axis along the microscope image. Head yawing and flagellar beating are extracted directly from θ(t) using Fourier Transform decomposition. The Fourier transform of the orientation signal θ(t) shows typical two-frequency peaks in the spectrum (Fig. 2B). The lower frequency peak defines the frequency of the head yawing that occurs at a slower rate than flagellar beating frequency. The second frequency peak in the Fourier spectrum of θ(t) captures the flagellar beating frequency. The associated amplitudes of the first and second frequency peaks of the Fourier spectrum thus provide a measure of the amplitude of the oscillations of the orientation θ(t), respectively, associated with the head lateral movement and flagellar beating amplitude (see details in next section: Segmentation-free spermatozoa orientation detection and tracking). Note that the flagellar wave amplitude defined here is relative to the mean sperm orientation for a given instant due to the nature of the orientation signal of a sperm image. Two sperm head and two flagellum kinematic parameters were extracted directly from the one-dimensional signal θ(t), highlighting the simplicity, robustness and reduction of complexity of the reported method. These are (1) sperm head kinematics, the frequency and amplitude of the sperm head yawing; and (2) flagellar beating parameters, the frequency and amplitude of flagellar waves. The parameters obtained with the proposed segmentation-free methodology are described as follows: (1) Beat cross frequency (BCFCASA), defined as the number of times the spermatozoon head crosses the average direction of movement, i.e. sperm head yawing (see OpenCASA: Alquézar-Baeta et al., 2019). Here we estimate BCFCASA from the first frequency peak of the Fourier transform of the orientation signal θ(t) (see Fig. 2A–C). (2) Flagellar beat frequency (FBF), estimated from the second frequency peak of the orientation signal θ(t) in the Fourier transform (see Fig. 2A–C), which captures the flagellar waving/beating frequency. (3) Amplitude of lateral head displacement (ALHCASA), defined by CASA as the value of the head displacement with respect to the mean trajectory (Katz, 1991), i.e. head yawing movement. Here we estimate ALHCASA from the relative amplitude of the first frequency peak of the Fourier transform of the orientation signal θ(t) (see Fig. 2A,B,D). (4) Flagellar beat amplitude (FBA), defined as the ‘head centerline deviation’ D (see Smith et al., 2009) to extract the amplitude of the flagellar wave relative to the sperm head. Here we estimate FBA from the relative amplitude of the second frequency peak of the Fourier transform of the orientation signal θ(t) (see Fig. 2A,B,D).

These parameters obtained via our segmentation-free method were compared with direct measurements of both head and flagellar oscillations using segmentation-based techniques reported in the literature.

Segmentation-free spermatozoa orientation detection and tracking

Orientation detection from an image is important for a wide range of applications. A popular method to extract orientation relies on gradient information at each position x=(x0,y0) in the image, where the first-order directional derivatives are computed. The directional derivatives become low when the direction is similar to the orientation at point x. Orientation calculation using this approach is popular because it is easy to implement and provides good approximations. OrientationJ (Püspöki et al., 2016; Rezakhaniha et al., 2012; Fonck et al., 2019) is a FIJI plugin (Schindelin et al., 2012) developed to more efficiently compute the average orientation of a region of interest (ROI) in an image. This software employs a more robust approach (less sensitive to noise) based on the use of a tensor structure given by:
formula
(3)
where the image, I, is convolved with a Gaussian filter to remove noise, Ix is the partial derivative of I with respect to x evaluated at x, Iy is the partial derivative of I with respect to y evaluated at x, the eigenvalues of matrix J(x) provide local shape information (Fonck et al., 2009; Harris and Stephens, 1988) and cubic B-spline interpolation is used to compute the continuous spatial derivatives. The two eigenvalues of J(x) give local shape information in the neighborhood of point x. If the two eigenvalues are similar and close to zero then the region is homogeneous (noisy structure), if the two eigenvalues are similar and larger than zero then the region is rotationally symmetric (blob structure), if one is positive and large and the other is close to zero, then the eigenvalue is aligned with the gradient direction (line structure). The eigenvector associated to the largest eigenvalue gives the dominant direction. It can be shown that the orientation of the dominant direction at a given position x can be computed by:
formula
(4)
To compute the orientation over a ROI, the derivatives are summed over the ROI:
formula
(5)

This approach to compute the dominant direction on the ROI performs very well and therefore it is feasible to apply it to obtain the flagellum orientation in an image. Nevertheless, for free-moving cells, it is difficult to have clear images with only one single cell in the field of view, thus the computation of the flagellum orientation (the derivatives are summed over the ROI, usually the full image) gets incorrectly affected by other cells or artifacts. For that reason, we modified and extended the functionality of OrientationJ to a more robust dominant spermatozoa flagellum orientation detector avoiding adding orientations from spurious objects. This was achieved by tracking the spermatozoa and measuring locally its main dominant orientation, as described below.

Modified OrientationJ plugin

The user has to identify the spermatozoon of interest by clicking over the head to determine the initial position p0 for tracking. Then, the dominant direction (Eqn 5) is calculated using three different ROIs: (1) a squared ROI with center at p0 is employed to calculate the spermatozoon orientation, Osquare. The square region must be big enough to contain the spermatozoon head and flagellum information. Note that the larger the square, the more plausible the orientation can be affected by external objects; (2) a circular ROI with center at p0 and radius r0 is employed to compute the spermatozoon head orientation, Ocircle. The radius r0 must be big enough to cover the spermatozoon head without including the spermatozoon flagellum; and (3) a rectangular ROI with height h, width w and rectangle width parallel to Osquare is constructed to compute the spermatozoon flagellum orientation, the points belonging to circular ROI are excluded from the analysis. Hence, this region should include orientations only from the spermatozoon flagellum, eliminating the influence of orientations coming from the spermatozoon head. Note that the larger the width, the more flagellum information can be included to compute the flagellum orientation. Fig. 1C,D depict the three different ROIs employed to compute the dominant direction. Rectangular ROIs should be used for analysis; however, this is sometimes unfeasible because the flagellum gets out of focus and there is no dominant direction. In our analysis, we use square ROIs because they are more stable. In addition to measuring the dominant direction, we also keep track of three parameters: mean intensity, coherence (Püspöki et al., 2016) and magnitude of dominant direction for the square ROI. These measures are useful to identify whether the spermatozoon head is still being tracked (we may have cases where the track does not follow the spermatozoon head because it is out of the field of view in the xy plane or it is moving too much in the z direction). Once the spermatozoon orientation has been computed for time t, we proceed to the following timepoint by automatically updating the spermatozoon position pt+1. To achieve this, a weighted sum over the positions inside the circle ROI and image intensity is calculated by:
formula
(6)
where x is a point (two coordinates) in the image and I(x) is the image intensity at position x. This approach places pt+1 near the spermatozoon head center at time t+1. However, this process can be affected if the spermatozoon moved a larger distance from the circle ROI at time t+1 or if it disappears from the image, or there is another spermatozoon crossing the circle ROI. In such cases, the spermatozoon track can be recovered at the current time t+1 or a posterior time if it has disappeared. We used two conditions to verify that a square ROI contains the spermatozoon head: (1) the first condition takes into account the coherency, it must be larger than 0.1 and; (2) the second condition takes into account the square mean intensity over the ROI, it must not change by a large amount from previously identified spermatozoa. This implies that the mean intensity over the square ROI for time t+1 must be larger than 0.4×μI, where μI is the mean of the measure mean intensity over previous times that were identified as correctly having a spermatozoon head. The value 0.4 was found experimentally. If these two conditions are satisfied, a spermatozoon exists in the current square ROI, otherwise it is inexistent. In the case that there is no spermatozoon, then the full image is covered with not overlapping square ROIs. The square ROI with the lowest cost is identified as the best ROI to compute the orientation at the current time. The cost for each square ROI is given by:
formula
(7)

where Ri is the current region to analyze, , , are the mean intensity, coherency and magnitude over each ROI used, circle, square and rectangles respectively. μI, μC and μM are the mean intensity, mean coherency and mean magnitude of the measures from regions of previous times that have been identified as correctly including a spermatozoon head. σI, σC and σM, are the standard deviations from the Gaussian distribution and were set experimentally to 10, 0.5 and 10, respectively. The cost is designed to give highest priority to regions having similar measures than the previous detected regions. Once that the best square ROI is selected for time t+1, the algorithm computes the three orientation measures presented previously and continues with time t+2, and the process is repeated until no images are left.

Our approach to compute the orientation has three main advantages over the original OrientationJ version (Schindelin et al., 2012): (1) it is less sensitive to external objects because it is computed over a neighborhood from the spermatozoon; (2) it is fast, because the tensor structure J(x) has to be computed only on a small squared region (except for the few cases that the algorithm missed the spermatozoon and has to re-compute the best ROI) instead of computing the tensor over the full image; and (3) our approach is customized to obtain three different orientations, which can be better suited for the user needs than a single orientation. As in the previous section, it is worth mentioning that no thresholding nor segmentation procedures are applied in the original and modified OrientationJ plugins.

We thank Paulina Torres for her helpful assistance with biological experimental procedures and Juan Manuel Hurtado for computational support. We also thank the editor and anonymous referees for the suggested improvements to the manuscript.

Author contributions

Conceptualization: G.C., P.H.-H., F.M., H.G., A.D.; Methodology: G.C., P.H.-H., F.M., A.D.; Software: G.C., P.H.-H., F.M.; Validation: G.C., P.H.-H., F.M.; Formal analysis: G.C., P.H.-H., F.M., H.G., A.D.; Investigation: G.C., F.M.; Resources: G.C., A.D.; Data curation: G.C., P.H.-H., F.M.; Writing - original draft: G.C., P.H.-H., H.G., A.D.; Writing - review & editing: G.C., H.G., A.D.; Visualization: G.C., P.H.-H., F.M., H.G., A.D.; Supervision: G.C., A.D.; Project administration: G.C., A.D.; Funding acquisition: G.C., A.D.

Funding

We acknowledge support from Dirección General de Asuntos del Personal Académico, Universidad Nacional Autónoma de México (DGAPA) scholarship CJIC/CTIC/0961/2019 to P.H. and grant to A.D. (IN200919). H.G. acknowledges support from DTP Engineering and Physical Sciences Research Council. G.C. and A.D. gratefully acknowledge financial support from Consejo Nacional de Ciencia y Tecnología (Conacyt) 253952, 255914 and Fronteras 71.

Achikanu
,
C.
,
Correia
,
J.
,
Guidobaldi
,
H. A.
,
Giojalas
,
L. C.
,
Barratt
,
C. L. R.
,
Da Silva
,
S. M.
and
Publicover
,
S.
(
2019
).
Continuous behavioural ‘switching’ in human spermatozoa and its regulation by Ca2+-mobilising stimuli
.
Mol. Hum. Reprod.
25
,
423
-
432
.
Alasmari
,
W.
,
Costello
,
S.
,
Correia
,
J.
,
Oxenham
,
S. K.
,
Morris
,
J.
,
Fernandes
,
L.
,
Ramalho-Santos
,
J.
,
Kirkman-Brown
,
J.
,
Michelangeli
,
F.
,
Publicover
,
S.
et al.  (
2013
).
Ca2+ signals generated by CatSper and Ca2+ stores regulate different behaviors in human sperm
.
J. Biol. Chem.
288
,
6248
-
6258
.
Alquézar-Baeta
,
C.
,
Gimeno-Martos
,
S.
,
Miguel-Jiménez
,
S.
,
Santolaria
,
P.
,
Yániz
,
J.
,
Palacín
,
I.
,
Casao
,
A.
,
Cebrián-Pérez
,
,
Muiño-Blanco
,
T.
and
Pérez-Pé
,
R.
(
2019
).
OpenCASA: A new open-source and scalable tool for sperm quality analysis
.
PloS Comput. Biol.
15
,
e1006691
.
Andersen
,
T. B.
,
López
,
C. Q.
,
Manczak
,
T.
,
Martinez
,
K.
and
Simonsen
,
H. T.
(
2015
).
Thapsigargin--from Thapsia L. to mipsagargin
.
Molecules
20
,
6113
-
6127
.
Bedford
,
J. M.
(
2015
).
Human spermatozoa and temperature: the elephant in the room
.
Biol. Reprod.
93
,
97
.
Blackmore
,
P. F.
(
1999
).
Extragenomic actions of progesterone in human sperm and progesterone metabolites in human platelets
.
Steroids
64
,
149
-
156
.
Chang
,
H.
and
Suarez
,
S. S.
(
2011
).
Two distinct Ca2+ signaling pathways modulate sperm flagellar beating patterns in mice
.
Biol. Reprod.
85
,
296
-
305
.
Chang
,
P. K.
,
Ehrlich
,
K. C.
and
Fujii
,
I.
(
2009
).
Cyclopiazonic acid biosynthesis of Aspergillus flavus and Aspergillus oryzae
.
Toxins (Basel)
1
,
74
-
99
.
Clapham
,
D. E.
(
2007
).
Calcium signaling
.
Cell
131
,
1047
-
1058
.
Correia
,
J.
,
Michelangeli
,
F.
and
Publicover
,
S.
(
2015
).
Regulation and roles of Ca2+ stores in human sperm
.
Reproduction
150
,
R65
-
R76
.
Darszon
,
A.
,
Beltran
,
C.
,
Felix
,
R.
,
Nishigaki
,
T.
and
Trevino
,
C. L.
(
2011
).
Calcium channels in the development, maturation and function of spermatozoa
.
Physiol. Revs.
91
,
1305
-
1355
.
Darszon
,
A.
,
Nishigaki
,
T.
,
López-González
,
I.
,
Visconti
,
P. E.
and
Treviño
,
C. L.
(
2020
).
Differences and similarities: the richness of comparative sperm physiology
.
Physiology (Bethesda)
35
,
196
-
208
.
Davis
,
R. O.
and
Katz
,
D. F.
(
1996
).
Computer-aided sperm analysis: a critical review
. In
Evaluation and Treatment of the Infertile Male
(ed.
G. M.
Centola
and
K. A.
Ginsburg
), pp.
30
-
41
.
New York
:
Cambridge University Press
.
David
,
G.
,
Serres
,
C.
and
Jouannet
,
P.
(
1981
).
Kinematics of human spermatozoa
.
Gamete Res.
4
,
83
-
95
.
Ding
,
S. S.
,
Schumacher
,
L. J.
,
Javer
,
A. E.
,
Endres
,
R. G.
and
Brown
,
A. E. X.
(
2019
).
Shared behavioral mechanisms underlie C. elegans aggregation and swarming
.
Elife
8
,
e43318
.
Drobnis
,
E. Z.
,
Yudin
,
A. I.
,
Cherr
,
G. N.
and
Katz
,
D. F.
(
1988
).
Hamster sperm penetration of the zona pellucida. Kinematic analysis and mechanical implications
.
Dev. Biol.
130
,
311
-
323
.
Eisenbach
,
M.
and
Giojalas
,
L. C.
(
2006
).
Sperm guidance in mammals - an unpaved road to the egg
.
Nat. Rev. Mol. Cell Biol.
7
,
276
-
285
.
Elferink
,
J. G.
(
1999
).
Thimerosal: a versatile sulfhydryl reagent, calcium mobilizer, and cell function-modulating agent
.
Gen. Pharmacol.
33
,
1
-
6
. .
Evans
,
B. W.
(
2008
).
Arduino Programming Notebook
, 2nd ed.
Lulu Press, Inc
.
Fonck
,
E.
,
Feigl
,
G. G.
,
Fasel
,
J.
,
Sage
,
D.
,
Unser
,
M.
,
Rüfenacht
,
D. A.
and
Stergiopulos
,
N.
(
2009
).
Effect of aging on elastin functionality in human cerebral arteries
.
Stroke
40
,
2552
-
2556
.
Gadêlha
,
C.
,
Bill
,
W.
and
Keith
,
G.
(
2007
).
Flagellar and ciliary beating in trypanosome motility
.
Cell Motil. Cytoskel.
64.8
,
629
-
643
.
Gadêlha
,
H.
,
Hernández-Herrera
,
P.
,
Montoya
,
F.
,
Darszon
,
A.
and
Corkidi
,
G.
(
2020
).
Human sperm uses asymmetric and anisotropic flagellar controls to regulate swimming symmetry and cell steering
.
Science Adv.
6
,
eaba5158
.
Gaffney
,
E. A.
,
Gadelha
,
H.
,
Smith
,
D. J.
,
Blake
,
J. R.
and
Kirkman-Brown
,
J. C.
(
2011
).
Mammalian sperm motility: observation and theory
.
Annu. Rev. Fluid Mech.
43
,
501
-
528
.
Galione
,
A.
and
Chuang
,
K. T.
(
2020
).
Pyridine nucleotide metabolites and calcium release from intracellular stores
.
Adv. Exp. Med. Biol.
1131
,
371
-
394
.
Gallagher
,
M. T.
,
Smith
,
D. J.
and
Kirkman-Brown
,
J. C.
(
2018
).
CASA: tracking the past and plotting the future
.
Reprod. Fert. Develop.
30.6
,
867
-
874
.
Gallagher
,
M. T.
,
Cupples
,
G.
,
Ooi
,
E. H.
,
Kirkman-Brown
,
J. C.
and
Smith
,
D. J.
(
2019
).
Rapid sperm capture: high-throughput flagellar waveform analysis
.
Hum. Reprod.
34
,
1173
-
1185
.
Garcia
,
M. A.
and
Meizel
,
S.
(
1999
).
Progesterone-mediated calcium influx and acrosome reaction of human spermatozoa: pharmacological investigation of T-type calcium channels
.
Biol. Reprod.
60
,
102
-
109
.
Gervasi
,
M. G.
and
Visconti
,
P. E.
(
2016
).
Chang's meaning of capacitation: a molecular perspective
.
Mol. Reprod. Dev.
83
,
860
-
874
.
Hansen
,
J. N.
,
Rassmann
,
S.
,
Jikeli
,
J. F.
and
Wachten
,
D.
(
2019
).
SpermQ – a simple analysis software to comprehensively study flagellar beating and sperm steering
.
Cells
8
,
10
.
Harper
,
C. V.
,
Barratt
,
C. L.
and
Publicover
,
S. J.
(
2004
).
Stimulation of human spermatozoa with progesterone gradients to simulate approach to the oocyte. Induction of [Ca2+]i oscillations and cyclical transitions in flagellar beating
.
J. Biol. Chem.
279
,
46315
-
46325
.
Harris
,
C.
and
Stephens
,
M.
(
1988
).
A combined corner and edge detector
. In
Proceedings of the fourth Alvey Vision Conference
, pp.
147
-
151
.
Ho
,
H. C.
and
Suarez
,
S. S.
(
2001
).
An inositol 1,4,5-trisphosphate receptor-gatedintracellular Ca2+ store is involved in regulating sperm hyperactivated motility
.
Biol. Reprod.
65
,
1606
-
1615
.
Ho
,
H. C.
and
Suarez
,
S. S.
(
2003
).
Characterization of the intracellular calcium store at the base of the sperm flagellum that regulates hyperactivated motility
.
Biol. Reprod.
68
,
1590
-
1596
.
Ishimoto
,
K.
,
Gadêlha
,
H.
,
Gaffney
,
E. A.
,
Smith
,
D. J.
and
Kirkman-Brown
,
J.
(
2017
).
Coarse-graining the fluid flow around a human sperm
.
Phys. Rev. Lett.
118
,
124501
.
Ishimoto
,
K.
,
Gadêlha
,
H.
,
Gaffney
,
E. A.
,
Smith
,
D. J.
and
Kirkman-Brown
,
J.
(
2018
).
Human sperm swimming in a high viscosity mucus analogue
.
J. Theor. Biol.
446
,
1
-
10
.
Kantsler
,
V.
,
Dunkel
,
J.
,
Blayney
,
M.
and
Goldstein
,
R. E.
(
2014
).
Rheotaxis facilitates upstream navigation of mammalian sperm cells
.
eLife
3
,
e02403
.
Katz
,
D. F.
(
1991
).
Characteristics of sperm motility
.
Ann. N. Y. Acad. Sci.
637
,
409
-
423
.
Lishko
,
P. V.
and
Mannowetz
,
N.
(
2018
).
CatSper: a unique calcium channel of the sperm flagellum
.
Curr. Opin. Physiol.
2
,
109
-
113
.
Marquez
,
B.
,
Ignotz
,
G.
and
Suarez
,
S. S.
(
2007
).
Contributions of extracellular and intracellular Ca2+ to regulation of sperm motility: Release of intracellular stores can hyperactivate CatSper1 and CatSper2 null sperm
.
Dev. Biol.
303
,
214
-
221
.
Meizel
,
S.
and
Turner
,
K. O.
(
1993
).
Initiation of the human sperm acrosome reaction by thapsigargin
.
J. Exp. Zool.
267
,
350
-
355
.
Miki
,
K.
and
Clapham
,
D. E.
(
2013
).
Rheotaxis guides mammalian sperm
.
Curr. Biol.
23
,
443
-
452
.
Milligan
,
M. P.
,
Harris
,
S.
and
Dennis
,
K. J.
(
1980
).
Comparison of sperm velocity in fertile and infertile groups as measured by time-lapse photography
.
Fertil. Steril.
34
,
509
-
511
.
Mortimer
,
S. T.
(
1997
).
A critical review of the physiological importance and analysis of sperm movement in mammals
.
Hum. Reprod. Update
3
,
403
-
439
.
Mortimer
,
S. T.
,
Van Der Horst
,
G.
and
Mortimer
,
D.
(
2015
).
The future of computer-aided sperm analysis
.
Asian J. Androl.
17
,
545
-
553
.
Nakagawa
,
S.
and
Cuthill
,
I. C.
(
2007
).
Effect size, confidence interval and statistical significance: a practical guide for biologists (review)
.
Biol. Rev.
82
,
591
-
605
.
Ng
,
K. Y. B.
,
Mingels
,
R.
,
Morgan
,
H.
,
Macklon
,
N.
and
Cheong
,
Y.
(
2018
).
In vivo oxygen, temperature and pH dynamics in the female reproductive tract and their importance in human conception: a systematic review
.
Hum. Reprod. Update
24
,
15
-
34
.
Ogawa
,
H.
,
Kurebayashi
,
N.
,
Yamazawa
,
T.
and
Murayama
,
T.
(
2020
).
Regulatory mechanisms of ryanodine receptor/Ca2+ release channel revealed by recent advancements in structural studies
.
J. Muscle Res. Cell Motil
. [Epub]
Ooi
,
E. H.
,
Smith
,
D. J.
,
Gadêlha
,
H.
,
Gaffney
,
E. A.
and
Kirkman-Brown
,
J.
(
2014
).
The mechanics of hyperactivation in adhered human sperm
.
R. Soc. Open Sci.
1
,
140230
.
Parkash
,
J.
and
Asotra
,
K.
(
2012
).
Calcium oscillations and waves in cells
.
Adv. Exp. Med. Biol.
740
,
521
-
529
.
Prole
,
D. L.
and
Taylor
,
C. W.
(
2019
).
Structure and Function of IP3 Receptors
.
ColdSpring Harb Perspect. Biol.
11
,
a035063
.
Püspöki
,
Z.
,
Storath
,
M.
,
Sage
,
D.
and
Unser
,
M.
(
2016
).
Transforms and operators for directional bioimage analysis: a survey
. In
Focus on Bio-Image Informatics: Advances in Anatomy, Embryol. and Cell Biol
, Vol.
219
(ed.
W.
De Vos
,
S.
Munck
and
J. P.
Timmermans
), pp.
69
-
93
.
Cham
:
Springer
.
Putney
,
J. W.
(
2013
).
Alternative forms of the store-operated calcium entry mediators, STIM1 and Orai1
.
Curr. Top. Membr.
71
,
109
-
123
.
Ravaux
,
B.
,
Garroum
,
N.
,
Perez
,
E.
,
Willaime
,
H.
and
Gourier
,
C.
(
2016
).
A specific flagellum beating mode for inducing fusion in mammalian fertilization and kinetics of sperm internalization
.
Sci. Rep.
19
,
31886
.
Rezakhaniha
,
R.
,
Agianniotis
,
A.
,
Schrauwen
,
J. T. C.
,
Griffa
,
A.
,
Sage
,
D.
,
Bouten
,
C. V. C.
,
van de Vosse
,
F. N.
,
Unser
,
M.
and
Stergiopulos
,
N.
(
2012
).
Stergiopulos experimental investigation of collagen waviness and orientation in the arterial adventitia using confocal laser scanning microscopy
.
Biomech. Model. Mechanobiol, vol
11
,
461
-
473
.
Rossato
,
M.
,
Di Virgilio
,
F.
,
Rizzuto
,
R.
,
Galeazzi
,
C.
and
Foresta
,
C.
(
2001
).
Intracellular calcium store depletion and acrosome reaction in human spermatozoa: role of calcium and plasma membrane potential
.
Mol. Hum. Reprod.
7
,
119
-
128
.
Schindelin
,
J.
,
Arganda-Carreras
,
I.
,
Frise
,
E.
,
Kaynig
,
V.
,
Longair
,
M.
,
Pietzsch
,
T.
,
Preibisch
,
S.
,
Rueden
,
C.
,
Saalfeld
,
S.
,
Schmid
,
B.
, et al. 
(
2012
).
Fiji: an open-source platform for biological-image analysis
.
Nat. Meth.
9
,
676
.
Smith
,
D. J.
,
Gaffney
,
E. A.
,
Gadêlha
,
H.
,
Kapur
,
N.
and
Kirkman-Brown
,
J. C.
(
2009
).
Bend propagation in the flagella of migrating human sperm, and its modulation by viscosity
.
Cell Motil. Cytoskeleton
66
,
220
-
236
.
Stival
,
C.
,
Puga Molina
,
L. C.
,
Paudel
,
B.
,
Buffone
,
M. G.
,
Visconti
,
P. E.
and
Krapf
,
D.
(
2016
).
Sperm Capacitation and Acrosome Reaction in Mammalian Sperm
.
Adv. Anat. Embryol. Cell Biol.
220
,
93
-
106
.
Strünker
,
T.
,
Alvarez
,
L.
and
Kaupp
,
U. B.
(
2015
).
At the physical limit - chemosensation in sperm
.
Curr. Opin. Neurobiol.
34
,
110
-
116
.
Vijayaraghavan
,
S.
,
Trautman
,
K.
,
Mishra
,
S. K.
and
Hermsmeyer
,
K.
(
1994
).
Evidence against a functional ATP-dependent calcium extrusion mechanism in bovine epididymal sperm
.
Mol. Reprod Dev.
38
,
326
-
333
.
Werner
,
S.
,
Rink
,
J. C.
,
Riedel-Kruse
,
I. H.
and
Friedrich
,
B. M.
(
2014
).
Shape mode analysis exposes movement patterns in biology: flagella and flatworms as case studies
.
PloS one
9
,
e113083
.
Williams
,
K. M.
and
Ford
,
W. C.
(
2003
).
Effects of Ca-ATPase inhibitors on the intracellular calcium activity and motility of human spermatozoa
.
Int. J. Androl.
26
,
366
-
375
.

Competing interests

The authors declare no competing or financial interests.

Supplementary information