The long and the short of sperm selection in vitro and in vivo: swim-up techniques select for the longer and faster swimming mammalian sperm


William V. Holt, Institute of Zoology, Zoological Society of London, Regent’s Park, London NW1 4RY, UK.
Tel.: +44 207 449 6630; fax: +44 207 586 2870; e-mail:


Sperm competition and sexual selection outcomes are sometimes reported as depending on sperm velocity and flagellar length, suggesting that sperm shape may be optimized for maximum efficiency. This is a largely unexamined assumption regarding sperm performance. Here, we examine this idea using a ‘swim-up’ selection technique as a proxy for sperm transport within the female tract, testing the hypothesis that variation in sperm tail length should be reduced by this procedure. We detected small but significant (< 0.001) increases in mean flagellar length in brown hare, pig and bull spermatozoa without reduction in variance. Applying the swim-up technique to boar ejaculates confirmed that the selected populations were enriched for fast motile spermatozoa. These effects were also reflected in vivo where boar spermatozoa with both short and long flagellae were able to reach and colonize the oviductal sperm reservoir. The benefits of possessing a longer flagellum thus appear to be marginal, suggesting that sperm selection in vivo is based on more complex criteria.


The sexually selected sperm hypothesis (Harvey & Bennett, 1985; Keller & Reeve, 1995; Pizzari & Birkhead, 2002) provides a theoretical framework for investigating the processes involved in post-copulatory sperm competition and cryptic female choice. One prediction arising from that hypothesis is that females are able to select for male traits that increase fertilization efficiency. Similarly, theoretical models of sperm size evolution (Parker, 1982, 1993; Parker & Begon, 1993) predict that sexual selection and sperm competition tend to favour convergence towards optimally effective sperm traits.

Experimental studies in different taxonomic groups have confirmed that when females are inseminated by spermatozoa from more than one male, the offspring are not equally derived from all of the males. Experiments in mammals involving artificial inseminations with heterospermic mixtures, in which single females are inseminated with spermatozoa from two or more males mixed in equal numbers, have repeatedly shown that fertilization success is skewed in favour of some individuals at the expense of others (for review, see Dziuk, 1996). This effect is commonly attributed to the combined influences of sperm competition and sperm selection effects within the female reproductive tract and the phenomenon itself is thought to have far reaching evolutionary consequences as it provides opportunities for biased post-copulatory breeding outcomes. The underlying mechanisms involved in these effects are complex and probably involve intrinsic sperm factors such as physical size, shape, motility, longevity within the female reproductive tract and metabolic efficiency, as well as the ability to interact appropriately with the female reproductive tract and negotiate the various physical and physiological obstacles that the spermatozoa encounter on their journey towards the oocyte. Some researchers have argued that sperm swimming speed, flagellar length and middle-piece volume are important determinants of relative fertilization success under conditions of sperm competition (Gomendio & Roldan, 1991, 2008; Anderson & Dixson, 2002). These arguments have been derived both from physical principles, which can be used to demonstrate that flagellar length and middle-piece volumes are related to flagellar beat frequency, power output and swimming speed (Dresdner & Katz, 1981) and from comparative phylogenetic studies of sperm dimensions in various taxonomic groups. The outcomes of these studies remain contradictory and equivocal. Some studies in primates and other mammals have failed to confirm these findings (Gage & Freckleton, 2003), whereas other recent evidence has supported the view that aspects of sperm morphology are correlated with both swimming speed (Malo et al., 2006) and in vivo fertility. Artificial insemination (AI) experiments in red deer (Malo et al., 2005) suggested that the fertility of different individuals was correlated with mean sperm velocity, and a recent analysis of sperm dimensions across 25 snake species concluded that sperm midpiece length is proportional to the risk of sperm competition (Tourmente et al., 2009). Confirmation of the expected positive relationship between sperm velocity and flagellum length in birds was also recently demonstrated by Lupold et al. (2009), although these authors were unable to find direct relationships between sperm velocity and the risk of sperm competition.

The concept that sperm flagellar length is a determinant of in vivo fertility leads to the logical conclusion that under conditions involving heterospermic insemination, as would happen frequently in species whose social systems involve significant multiple mating, there should be a strong tendency towards achieving an optimal sperm shape. Recent studies of sperm size variation in birds support this view (Calhim et al., 2007; Immler et al., 2008; Kleven et al., 2008), suggesting that greater variation in sperm length is correlated with lower incidence of sperm competition.

If species-dependent optimal flagellar lengths exist, then we might expect to find very little intra-specific, or indeed intra-individual, variation in flagellar length in species with high levels of sperm competition, as departures from the optimal value would tend to disappear by selection. However, several researchers have shown that this is not the case (Ward, 1998; Morrow & Gage, 2001) and that significant intra-individual variations in sperm dimensions are widely observed. The significance of such variation has remained difficult to interpret. With this in mind, we have analysed spermatozoa from three mammalian species (boar, bull and brown hare) to evaluate flagellar length variation within and between individuals, and have also investigated directly the hypothesis that spermatozoa with longer flagellae should possess significant advantage when placed under selective conditions that are believed to act upon motility. If this hypothesis is upheld, we would expect to see spermatozoa with shorter flagellae eliminated from the selected populations. The flagellar measurements were therefore combined with a simple sperm selection procedure, the swim-up technique, which operates by isolating the more motile sperm population (Esteves et al., 2000). This is similar to, but not precisely the same as, the estimation of sperm ‘mobility’ that has been used as an objective method of measuring sperm activity in chicken semen (Birkhead et al., 1999). In addition, using boar spermatozoa as a model system, we evaluated the effectiveness of the swim-up technique at producing enrichment for more motile sperm populations.

As part of this study, we also wished to examine and measure spermatozoa within the porcine female reproductive tract after in vivo insemination to see whether there was evidence of size-biased selective sperm transport from the uterus into the oviduct. Although we prepared and inseminated 12 sows for this experiment, it was ultimately only possible to examine and measure oviductal spermatozoa in two of the females. Although this precluded statistically valid direct comparisons of sperm length between the utero-tubal junction (UTJ) and oviductal isthmus, it was nevertheless possible to investigate the hypothesis that the UTJ might selectively impede the passage of unusually short or long spermatozoa.

Materials and methods

Experiment 1: sperm length variation in the brown hare, domestic boar and domestic bull

Sources of spermatozoa

Three different sources of spermatozoa were involved in these experiments, and the swim-up procedures had to be modified to suit the species and the situation.

Brown hare (Lepus europaeus: Pallas 1778) spermatozoa were obtained post-mortem from the caudae epididymides of animals culled in and around Western Plains Zoo, Dubbo, NSW, Australia. This wild hare population descends from animals introduced into Australia around 1860 by English settlers (Richardson & Walton, 1989). Testes and epididymides were washed in phosphate-buffered saline (PBS) and the caudae epididymides were removed and cut into several pieces. The pieces were placed on dental wax, covered with 2 mL Ham’s F10 culture media and incubated at 38 °C inside a 35 mm Petri dish for approximately 10 min to obtain epididymal spermatozoa. After incubation, the pieces of epididymal tissue were removed; drops of the resultant sperm suspension were used for the preparation of smears. The smears were air-dried and fixed in ethanol.

A ‘swim-up’ suspension of epididymal spermatozoa was prepared as follows. The epididymal sperm suspension (0.5 mL) was diluted (1 : 1 v/v) with Ham’s F10 medium and centrifuged at 500 g for 5 min in a 1.5 mL Eppendorf tube. The supernatant was removed without disturbing the pellet, replaced with the same volume of fresh media, and the samples incubated at 38 °C for 1 h. The supernatant media containing the swim-up sperm suspension was then carefully aspirated and used for the preparation of smears. These smears were also air dried and fixed in ethanol.

Bovine (Bos Taurus; Linnaeus) semen samples (N = 8) were obtained from the Genus Plc Freezing Unit, Ruthin, Wales. The raw semen samples were diluted 1 : 10 v/v in Eqcellsire (IMV, L’Aigle; 209, France) storage diluent shortly after collection and then sent to London overnight for use the following day. Smears of this unselected suspension were prepared on glass slides and air-dried. To undertake the ‘swim-up’ procedure, the freshly diluted semen was first concentrated by centrifugation (600 g for 5 min) and the sperm pellet overlaid with 2 mL of Eqcellsire diluent in tubes. The tubes were incubated at 39 °C for 40 min, after which 100 μL of the ‘swim-up’ fraction was retrieved and used for the preparation of smears.

Sperm-rich fractions from commercial boars (Sus scrofa; N = 18), diluted to approximately 3 × 107 cells mL−1 in Beltsville thawing solution (BTS), an extender widely used for commercial insemination (Pursel & Park, 1985), were obtained from fertile boars maintained by JSR-Genetics Ltd (Thorpe Willoughby, Yorkshire, UK). The diluted samples were sent overnight to the laboratory for use the following day. Smears of unselected spermatozoa from nine boars were prepared on glass slides and allowed to dry in air. The swim-up procedure was carried out by first centrifuging the sperm suspensions (40 mL at 1000 g for 10 min), so that the supernatant containing BTS and seminal plasma could be removed. The soft sperm pellet was resuspended in phosphate-buffered saline (PBS, pH 7.4), and centrifuged as before. The supernatant was then removed and the pellet resuspended in 3 mL of a modified Tyrode’s medium (TALP) supplemented with lactate and pyruvate. This consisted of 116 mm NaCl, 3.1 mm KCl, 0.4 mm MgSO4, 0.3 mm NaH2PO4, 5 mm glucose, 21.7 mm sodium lactate, 1 mm sodium pyruvate, 5 mm NaHCO3, 1 mm EGTA, 20 mm Hepes (adjusted with NaOH to pH 7.6 at 20 °C), 3 mg BSA mL−1, 100 μg kanamycin mL−1, and 20 μg phenol red mL−1; its final pH at 38 °C was 7.4 and its osmolality 300 mOsmol kg−1. The swim-up procedure was undertaken by placing 1 mL of the sperm suspension in a 15 mL Falcon tube, gently overlaying it with 500 μL TALP and placing it at an angle of 45° for 1 h at 38 °C in a 5% CO2 atmosphere. 100 μL of the upper layer were recovered after 1 h and used for the preparation of smears.

Staining procedure, light microscopy and measurement technique

All slides were stained using Diffquick (Raymond A. Lamb, London, UK) according to the manufacturer’s instructions. Slides were air dried after staining, mounted in DPX and examined using a manual measurement facility included in the computer-assisted Hobson Sperm Morphology assessment system (Hobson Tracking Systems Ltd, Sheffield, UK). This consisted of an Olympus BH-2 microscope equipped with a ×40 bright field objective (numerical aperture 0.70) and a Sony SPT-M108CE monochrome video camera (Sony, Tokyo, Japan) linked via a X 3.3 projection eyepiece. Tail measurements were obtained by tracing the image of the tail on the computer screen: the tail length measurement represented the distance from the neck of the spermatozoon to the end-piece. 100–200 sperm tail measurements were obtained for each animal/treatment combination.

Statistical analyses

Comparisons of flagellar length data between ‘swim-up’ and control samples were undertaken using paired t-tests and Wilcoxon matched pairs test for contrasting treatment outcomes across several individuals. The Mann–Whitney U-test was used when comparing data within individuals. Tail length frequency distributions were examined for normality using the Shapiro–Wilks test. Coefficients of variation within and between individual animals and treatments were also calculated. Homogeneity of variance was calculated using the Levene test.

Experiment 2: effects of swim-up on boar sperm motility

Samples from nine boars (not those used for flagellar length estimation) were used for objective comparisons of sperm motility, before and after swim-up, using the Hobson Sperm Tracker (for technical details see Holt & Harrison, 2002). In brief, the sperm populations were diluted to approximately 5 × 106 per milliliter and were examined at 38 °C using a microscope fitted with a ×10 negative-high phase contrast objective and video camera. Sperm motility parameters were measured using the Hobson Sperm Tracker (Hobson Tracking Systems, Sheffield, UK). Unselected sperm populations in TALP, diluted to 5 × 106 sperm mL−1, were compared with the swim-up population, which was adjusted to the same sperm concentration.

The effects of swim-up on the motility of boar spermatozoa were determined by first combining all the CASA-derived data as a single dataset, and then using a hierarchical multivariate cluster analysis technique (PATN; Abaigar et al., 1999) to categorize each spermatozoon into different groups on the basis of four motility parameters (curvilinear velocity (VCL), velocity along the average path (VAP), straight line velocity (VSL) and amplitude of lateral head displacement (ALH). Treatment effects were examined by calculating the proportion (%) of spermatozoa in each resultant PATN group, for each boar × treatment combination, and then examining the ‘unselected vs selected’ contrasts using paired t-tests. Individual CASA parameters were also compared as means and medians using the ‘boar × treatment’ interactions to test for treatment effects.

Experiment 3: porcine in vivo insemination

Experimental animals were obtained from an environmentally controlled, experimental pig farm at the University of Murcia (Murcia, Spain). Semen samples were obtained from boars of high fertility that were routinely used for artificial insemination studies.

Animal preparation

A total of 12 sows were randomly selected and their piglets were weaned 24 h prior to induction of oestrus. The sows were injected with 1250 IU per sow of Serum Gonadotrophin PhEur (PMSG; Folligon, Intervet, the Netherlands). Then, 48 h after PMSG injection, the sows were injected with 500 IU per sow hCG (Veterin Corion, Divasa Farmavic S.A., Barcelona, Spain). The ovulatory states were determined with transrectal ultrasonography (Pie Medical SC100 Scanner, Maastrich, the Netherlands) and sows were split into two groups; (i) surgery 4 h post AI and (ii) 48 h post AI depending on whether sows were either at the preovulatory, periovulatory or post-ovulatory stages. Because of uterine infections, two sows were excluded from the experimental analysis.

Experimental design

Semen samples from three boars of proven fertility were collected by the gloved hand technique, mixed into a heterospermic sample for insemination, and diluted in BTS. AI was carried out with 3 × 109 spermatozoa per 85 mL dose. Sows were tranquilized with azaperone (2 mg Kg−1 body weight). Then, general anaesthesia was induced with intravenous sodium thiopental (7 mg kg−1 body weight) and maintained using 3.5–5% Halothane. The sows were prepared for surgery individually.

Oviduct removal was performed surgically either 4 or 48 h post AI. Oviducts were flushed through from the infundibulum with 10–20 mL of PBS, and the effluent fluid was collected by cannula inserted into an incision in the uterus. Any oocytes and embryos collected were transferred into M199 (Gibco) medium and transported back to the laboratory in a portable incubator for analysis at 39 °C. The PBS flushings were also collected and transported back to the laboratory for the analysis of sperm content. The oviducts were trimmed free of surrounding connective tissue and the uterine horns were ligated 1.5 cm from the UTJ. The oviducts were split open longitudinally to expose the lumen, opened flat and pinned to a cork-board. Tissues were immersed in fixative (4% formaldehyde and 2% glutaraldehyde in 100 mm phosphate buffer, pH 7.4) for 30 min to 1 h and then washed and stored in PBS at 4 °C until required for further processing by scanning electron microscopy (SEM).

The fixed oviducts were divided into UTJ, isthmus and ampulla regions; each anatomical region was cut into pieces of approximately 5 mm2. The pieces were rinsed and soaked in 200 mm cacodylate buffer and post-fixed with 2% osmium tetroxide in this buffer for 1hr. Samples were then rinsed in cacodylate buffer and dehydrated in increasing concentrations of ethanol (50%, 70%, 90% and 100%) for 15 min each, then stored in 100% ethanol. Before SEM analysis, the samples were dried with 1,1,1,3,3,3-hexamethyldisilazan. The dried samples were mounted onto stubs with carbon tabs and sputter-coated with gold to obtain a conductive layer. The sample pieces were then subjected to scanning electron microscopy analysis using a JOEL JSM-5000LV SEM (JOEL, Tokyo, Japan). Approximately 70–80% of each section of each oviduct was carefully examined to estimate the number of spermatozoa bound to the epithelial surface; sperm tail lengths, sperm head lengths and widths were measured from SEM images using Image Pro plus v6.5 (Media Cybernetics, Marlow, UK).

Limited statistical analysis of this experiment was undertaken as data were only obtained from two of the inseminations (the two sows that were preovulatory at the time of insemination), and it was therefore not possible to compare mean sperm flagellar lengths with confidence. However, we were still able to examine the distribution, range and variances of flagellar length values to see whether they were normally distributed and whether there was any significant exclusion of exceptionally short or long spermatozoa during their passage from the UTJ to the oviductal isthmus.


Experiment 1: sperm length variation in the brown hare, domestic boar and domestic bull

Comparison of total flagellar length distributions for unselected spermatozoa from the three species examined here revealed that the hares showed the widest range of values, spanning approximately 60 μm. More than expected spermatozoa were evident at both the high and low extremes of the distribution (Fig. 1a). Boar sperm lengths were less variable, with the range spanning about 30 μm, but the distribution also showed departure from normality with more than expected short spermatozoa (Fig. 1b). Bull spermatozoa ranged from 40 to 70 μm, and also exhibited significant departure from normality (Fig. 1c).

Figure 1.

 Normal plots of sperm tail lengths in three mammalian species. Panels (a) data from European brown hares (N = 15); (b) data from domestic boar (N = 9); (c) data from domestic bulls (N = 8). The fitted line represents an expected normal distribution.

Frequency distribution analysis revealed that hare epididymal sperm tail lengths were highly variable within individuals. Most tail length distributions (13/15) showed highly significant negative skews and departure from statistical normality (Shapiro–Wilks test; < 0.0001); one showed less significant departure from normality (= 0.024) and one was normally distributed. The ‘within-animal’ coefficients of variation (CV) for the unselected, epididymal, sperm populations ranged from 3.5–11.1%; (mean ± SD) CV = 6.78 ± 2.25%); this was higher than the corresponding ‘between-animals’ CV (1.96%).

After the swim-up treatment, most (9/13) of the negatively skewed distributions remained negatively skewed (< 0.001), showing that many of the shorter sperm tails had been retained during the swim-up process. The swim-up technique produced significant enrichment (Mann–Whitney: < 0.05) for longer sperm tails in 13/15 individual hare samples (Fig. 2a). This resulted in a highly significant increase in flagellar length with the paired t-test (= 0.00024) as shown in Table 1 and Figs 2 and 3.

Figure 2.

 Plots showing the effects of selection by the swim-up technique upon tail lengths (mean ± SEM) in sperm samples from individual boars (Panel a) and hares (Panel b). (○ Nonswim up; bsl00066 selected by swim-up). The data from individual animals arranged in ascending order of tail length prior to swim-up.

Table 1.   Comparisons of sperm tail lengths before and after selection by the ‘swim-up’ method.
SpeciesTreatmentNumber of animalsNumber of SpermatozoaRange (μm)Length (μm) ±SDBetween-animal coefficient of variationSignificance of swim-up effect
  1. SD of tail lengths calculated using the individual spermatozoa as the unit of measurement.

  2. Significance of selection treatments tested using paired t-tests across n individuals.

BoarUnselected990029.0–48.542.70 (±0.659)1.54= 0.00045
Swim-up990035.3–51.843.64 (±0.774)1.77
BullUnselected880042.0–69.757.61 (±1.78)3.09= 0.0067
Swim-up880041.3–70.058.23 (±1.72)2.95
HareUnselected15162015.4–70.242.39 (±0.832)1.96= 0.00024
Swim-up15150015.4–67.243.37 (±0.797)1.84
Figure 3.

 Summary box–whisker plots showing the effects of swim-up treatment on flagellar length on (a) hare, (b) boar and (c) bull spermatozoa. Small squares represent the median values; boxes indicate 25th and 87th percentiles, and whiskers represent the nonoutlier range. Filled circles represent outliers.

Five of the nine frequency histograms obtained from unselected spermatozoa from within single samples of unselected boar spermatozoa showed normal distributions (Shapiro–Wilks test; > 0.05) and three of the other four showed negative skews. The ‘within-animal’ CVs for unselected spermatozoa ranged from 2.72% to 5.15% (mean ± SD: 3.59 ± 0.79), which is smaller than the corresponding ‘between-animal’ CV (22.17%). After the swim-up technique, only one distribution remained negatively skewed; three were positively skewed and four showed statistical normality. A significant upwards shift in mean tail length occurred in 8/9 of the individual boars tested for the effect of swim-up (Mann–Whitney: < 0.05) (Fig. 2b) and similarly resulted in a highly significant outcome when the combined data were examined using the paired t-test (= 0.00045) (Table 1 and Fig. 3).

Four of the unselected populations of bull spermatozoa showed negative skews (< 0.05) and four were normally distributed; the swim-up technique did not affect these populations consistently. Two of the samples with negative skews converted to positive skews (< 0.05); three of the samples with normal distributions remained unchanged, whereas the fourth became negatively skewed. The intra-individual CVs for unselected sperm flagellar lengths from the bulls ranged from 2.42% to 5.03%, whereas the between-animals CV was higher (27.2%).

Bull spermatozoa showed the swim-up induced increase in sperm tail length (Fig. 3), although there was less consistency between the individuals. The Mann–Whitney tests showed that 4/8 bulls showed the effect at the < 0.05 level; one showed the effect with a probability (= 0.06) and the remaining three showed no statistical evidence for the effect. However, when the paired t-test was applied, there was a highly significant overall increase in sperm tail length (= 0.0067). The overall effect is illustrated in Fig. 3.

Experiment 2: effects of swim-up on boar sperm motility

No effects of the swim-up treatment was detected upon the individual CASA parameters when tested with either the paired t-tests or Wilcoxon matched pairs test (> 0.1; N = 9). Deriving ‘boar × treatment’ mean values for these tests was considered inappropriate for detecting such treatment effects (Holt et al., 2007) as the data distributions for VCL, VAP and VSL were significantly skewed in all 9 boars.

Treatment effects were therefore investigated using the PATN technique, which identified three subgroups of spermatozoa across the whole dataset. The characteristics of the three subgroups, shown in Table 2, indicated that PATN Group 1 represented a moderately active sperm population and Group 2 spermatozoa were highly active and progressive, as indicated by the exceptionally high VSL value. Group 3 represented spermatozoa showing low linearity but moderate curvilinear velocity, implying that they were nonprogressive. The proportion of Group 1 spermatozoa was significantly enriched by the swim-up treatment (Fig. 4) (mean ± SEM; pellet; 46.4 ± 7.7% vs swim-up; 66.7 ± 5.5%; = 0.031; DF = 8) whereas the proportion of Group 3 spermatozoa was slightly, but not significantly (= 0.064; DF = 8), diminished (mean ± SEM; pellet: 31.4 ± 3.1% vs swim-up: 25.6 ± 2.8%). The proportion of Group 2 (fast and progressive) spermatozoa was unaffected by the swim-up treatment (= 0.13).

Table 2.   Summary of boar sperm motility parameters for the groups identified by PATN analysis.
PATN groupNVCL (μm s−1)VAP (μm s−1)VSL (μm s−1)BCF (Hz)ALH (μm)LIN (%)
  1. Values are expressed as means (±standard deviation).

  2. ALH, amplitude of lateral head displacement; VAP, velocity along the average path; VCL, curvilinear velocity; VSL, straight line velocity; BCF, Beat Cross Frequency; LIN, Linearity of track.

1197258.8 (16.6)31.92 (14.3)17.86 (14.9)3.73 (3.8)5.47 (7.4)29.35 (21.2)
240698.9 (12.3)77.48 (12.3)70.06 (19.8)8.94 (3.1)6.55 (3.6)70.96 (18.5)
3103820.5 (7.4)15.83 (7.4)2.46 (2.5)0.01 (0.1)0.19 (1.8)16.06 (11.1)
Figure 4.

 Box–whisker plots showing the proportions (mean, SD and SEM) (%) of Group 1 boar spermatozoa (defined by PATN analysis) present in the unselected and swim-up samples. The difference between treatments was significant (paired t-test; P = 0.03; N = 9).

Experiment 3: comparison of boar spermatozoa in the utero-tubal junction and oviductal isthmus

Spermatozoa were present in the luminal flushings of all the sows used, regardless of the interval between AI and surgery and irrespective of the ovulatory status at the time of insemination. However, spermatozoa were only attached to the epithelial cells of the UTJ and isthmic regions of the two sows (2/12) which were still preovulatory when inseminated. No other regions had any substantial number of spermatozoa associated with the oviductal epithelium. SEM observations showed that spermatozoa in the UTJ region were generally found in large reservoirs within crypts and interfolds of the epithelia (Fig. 5a), whereas in the isthmic region, the spermatozoa were sparsely distributed and attached to the epithelium, primarily towards the UTJ end of the isthmus (Fig. 5b). The spermatozoa were attached independently or in small groups.

Figure 5.

 Scanning electron micrographs of boar spermatozoa bound to the epithelia of the utero-tubal junction (panel a) and oviductal isthmus (panel b) 4 h after artificial insemination. Bar in A = 10 μm and in B = 5 μm.

The small number of sows with spermatozoa attached to their oviductal epithelia precluded valid comparisons of mean flagellar lengths between the UTJ and the oviductal isthmus. Nevertheless, when we compared the flagellar length distributions at the UTJ and isthmus, it became apparent that there was no evidence for the size-selective prevention of sperm transport across this important entry point to the oviduct. In fact, the lower 5% percentile boundary of the distribution was lower (18.6 μm) for the isthmus than for the UTJ (29.4 μm), whereas there was little difference between the 95% percentile boundaries; 55.9 vs 53.7 μm for the UTJ and isthmus, respectively. Both sets of spermatozoa displayed normally distributed flagellar lengths that ranged continuously from approximately 20 to 73 μm (means ± SD; UTJ = 42.09 ± 9.31 (N = 61): isthmus 33.21 ± 10.81 (N = 53) (Fig. 6). There was no reduction of variance associated with sperm transport across the UTJ (Levene test; FDF=1 = 0.10: = 0.74).

Figure 6.

 Histograms showing length distributions of boar sperm flagellae measured from scanning electron micrographs taken in (a) the region of the utero-tubal junction and (b) oviductal isthmus. Both distributions are statistically normal.


The results of this study demonstrate that across the three species evaluated here, the swim-up procedure consistently resulted in a very small (approximately 1 μm) but highly significant upward shift in mean flagellar length. This result was achieved in nearly every individual examined. However, contrary to our expectation, it is remarkable that these shifts were not caused by the selective elimination of spermatozoa with short tails, which might be expected to swim slowly if flagellar length were the main determinant of velocity. This strongly suggests that our expectation was too simplistic and that other attributes of the spermatozoa are more important determinants of selection. Theoretical principles indicate that sperm swimming speed depends linearly on both flagellar length and beat frequency (Dresdner & Katz, 1981). Differences in beat frequency, a dynamic and energy-dependent attribute of spermatozoa, may therefore contribute more strongly than mean velocity to the effectiveness of sperm selection processes. This argument is supported by experimental evidence published by Katz (Katz et al., 1979). These authors measured beat frequency in spermatozoa from mice carrying T-complex mutations; these mutations cause defective sperm transport and transmission ratio distortion (Olds-Clarke & Sego, 1992; Johnson et al., 1995), The beat frequency values were distributed bimodally after 2 h of incubation, which led these authors to interpret their findings as indicating the presence of two sperm subpopulations in the same samples. They attributed the difference to suspected metabolic defects related to energy production in one of the subpopulations, and it is therefore likely that similar differences in energy production would be influential within sperm selection procedures like the swim-up technique. In our own data, we could not detect higher beat frequency after the swim-up procedure, but applying the CASA technique after the sperm selection process is less than ideal because of the strong likelihood that the sperm behaviour may change, because of capacitation or degenerative effects between applying the selection process and then making the measurements.

The sperm ‘mobility’ assay as a test for fertility advantage in chickens (Birkhead et al., 1999) is similar, but not identical, to the swim-up procedure reported here. The mobility assay involves measuring the ability of the motile sperm population to swim downwards into a medium containing the cell separation medium, Accudenz. The rate of colonization is measured by densitometry and the outcomes show that the results of competitive inseminations with paired semen samples mirror their superior ability to colonize the Accudenz layer. Such evidence emphasizes the importance of good sperm motility in deciding the outcomes of competitive inseminations. Sperm motility in this system was found to be determined largely by mitochondrial function (Froman & Kirby, 2005); in turn this was correlated with ability to generate adenosine triphosphate and the nature of mitochondrial DNA haplotypes.

Experiment 2 showed that the swim-up procedure produced significant enrichment for active and progressive spermatozoa, while slightly reducing the proportion of poorly progressive spermatozoa. One attractive interpretation of this result is that there may be differences in flagellar length between these two subpopulations, thus explaining how a slight shift in population mean flagellar length might be caused. However, biophysical analyses of the relationships between sperm morphology and swimming velocity have shown that not only flagellar characteristics, but also sperm head properties, particularly length/width ratio, can affect velocity (Malo et al., 2006; Gomendio et al., 2007; Gillies et al., 2009). Although the balance shifted between two of the sperm subpopulations, a third subpopulation (Group 2), comprising about 12% of the whole population, and showing exceptionally high velocity and linearity, exhibited little change in frequency before and after swim-up. This population is of considerable interest as it might represent the cohort of cells with the highest in vivo capability of traversing the UTJ. Current developments aimed at producing microfluidic devices for the separation of human spermatozoa are of relevance in this context, because populations of exceptionally high motility also show reduced frequency of sperm head abnormalities and enhanced zona-penetration rates during in vitro fertilization (Cho et al., 2003).

Although the in vivo experiment reported here was disappointing because we were unable to ascertain whether sperm transport from the UTJ to the oviductal isthmus produced any change in mean flagellar length, we were still able to conclude that sperm transport to, and across, the UTJ is not accompanied by the exclusion of short, or even exceptionally long, sperm tails. The similarities between our in vivo and in vitro results for the pig lend support to the view that the swim-up method was an appropriate experimental system to use. The entire range of flagellar lengths in all three species was preserved by the swim-up treatments, and the complete range could also be found at the point of entry to the oviduct itself. This is of interest as it provides empirical evidence showing clearly that, in the pig at least, there is no stringent in vivo selection bias for or against particular sperm tail lengths during the swim-up process, and by generalization of the boar data, during sperm transport to the oviductal sperm reservoir. However, if this marginal bias were repeated in vivo over many generations, it is likely to result in an upward size shift in sperm flagellar length. In a recent review (Ward, 1998) noted that sperm flagellar length is a heritable characteristic, which would be consistent with a slight tendency towards the selection of longer sperm tails in species where the female reproductive tract imposes barriers that spermatozoa must overcome through their own intrinsic motility.

Evidence from the present study of sperm transport in pigs confirmed that sperm reservoirs, i.e. colonies of spermatozoa bound to epithelial cell surfaces, became established at the UTJ and oviductal isthmus within 4h of preovulatory insemination (Hunter, 1981; Hunter et al., 1984; Suarez, 1998; Rodriguez-Martinez, 2001). The mammalian UTJ presents a significant obstacle to sperm entry into the oviduct, and only a privileged subpopulation of spermatozoa reaches this site. Entry is partly dependent upon intrinsic motility (Olds-Clarke, 1996) but, in mice at least, is also known to depend on the presence of specific sperm surface proteins such as fertilin-β (Cho et al., 1998). Before reaching the UTJ, spermatozoa must traverse the uterine horns, where they are thought to undergo some degree of selection. Recent evidence from the porcine system suggests that the selection is partly mediated by neutrophilic granulocytes, which invade the uterus shortly after insemination and then transiently bind and release intact and functional spermatozoa (Schuberth et al., 2008; Taylor et al., 2008). After release, the spermatozoa are allowed to progress towards the UTJ.

Sperm motility is known to be inhibited within the mammalian oviduct (Brussow et al., 2008), partly because the high potassium concentration (Overstreet & Cooper, 1975) suppresses motility and also because the sperm–oviduct interaction initiates specific signal transduction systems that modulate motility (Satake et al., 2008; Gonzalez-Fernandez et al., 2009). Genomic and proteomic studies of the female reproductive tract have shown that the arrival of spermatozoa in the female tract stimulates the novel transcription of many genes and the synthesis of multiple new proteins (Fazeli et al., 2004; Georgiou et al., 2005). The significance of these observations has yet to be explored but is likely to involve aspects of sperm selection. The passage of spermatozoa towards the site of fertilization, the oviductal ampulla, cannot therefore simply be viewed as a race which will be won by the fastest spermatozoa. The positive relationship between oviductal length and sperm mid-piece volume across 48 mammalian species (Anderson et al., 2005, 2006) identified by comparative analysis is therefore not easily explained by simple reference to high swimming speeds.

The epididymal samples all possessed more than expected short-tailed spermatozoa, with most distributions not conforming to statistical normality. The hare samples showed the highest within-animal CV but the lowest between-animal CV when compared to the other species. A suggested explanation for this observation relates to the source of the animals used. As the hares were all located around the Western Plains zoo at Dubbo in New South Wales, there is a strong likelihood that they were from a single population. On the other hand, although the boars and bulls were from agricultural breeding centres, they were not from single pig and cattle breeds. This explanation is consistent with data presented by Morrow & Gage (2001) who showed differences in flagellar length between cattle breeds. Other researchers, studying wild species, have interpreted the differences in CV in terms of sperm competition risk. Thus Kleven et al. (2008), who reported that both within- and between-animal CV values ranged between 0.5% and 3.5% across 22 species of passerine birds, found that the CV was negatively correlated with the incidence of sperm competition in a species; measurements by Laskemoen et al. (2007) in two other passerine birds were consistent with these relatively low CV values. Breed et al. (2007) reported considerably higher values (31.9–37.5% for the Asian bandicoot (Bandicota indica), a species in which sperm head shape variability is also evident and where there is little sperm competition. There is insufficient behavioural and genetic data to support or refute this supposition with respect to hares.

In conclusion, these experiments have demonstrated a marginal bias towards the selection of longer sperm tails during in vitro experiments, but no severe degree of bias against spermatozoa with exceptionally short or long flagellae. The in vivo experiment reported here also failed to produce any evidence for size-selective biases imposed by the UTJ during normal sperm transport. These results do not necessarily contradict comparative analyses that show correlations between sperm flagellar length, motility and the risk of sperm competition, as the marginal increases in mean flagellar length produced by the in vitro selection method used here are likely to induce significant flagellar elongation under conditions of strong environmental selection.


We thank the following for their help with the provision of samples for this study; Dr Amanda Pickard, Dr David Blyde and Catriona McCallum for obtaining hare spermatozoa in Western Plains Zoo, Dubbo, Australia; Dr Stuart Revell (Genus Freezing Unit, Ruthin, Wales) for providing bull semen samples and JSR-Genetics Ltd (Thorpe Willoughby, Yorkshire, UK) for providing boar semen samples. We are also grateful to Professor Juan-Maria Vazquez of the University of Murcia, Spain, for arranging the in vivo experiment.