7. FRAC analysis of fertile and subfertile bulls used in AI industry
Sperm samples from thirty Holstein bulls and one Jersey bull were obtained from the Select Sires, Inc. population of bulls that have been used in artificial insemination (AI). These bulls’ SCR scores ranged from a minimum of -18.2 to a maximum of +2.8 (Supplementary Table 3). Fertile bulls were classified as those that have SCR deviations >-3 and sub-fertile bulls are those that have SCR deviations <-3 (Amann and DeJarnette, 2012; DeJarnette et al., 2022; Harstine et al., 2018). All 31 bulls passed a rigorous semen and sperm analysis before they were used for artificial insemination in the field (DeJarnette et al., 2022). Therefore, it was surprising that six of them (five Holsteins and one Jersey) had an SCR that is significantly lower than the average and in the subfertile range. For our study, we divided the 31 bulls to two categories: 1) the fertile category included 25 bulls with SCR deviations above -3 and 2) the subfertile category included six bulls with SCR at or below -3. This is a ratio of about four fertile bulls for each subfertile bull, which is preferred in small control case studies (Breslow, 1982). To determine bull sperm centriole quality, we used a quantitative immunofluorescence-based assay named FRAC, which was developed to compare staining intensity ratios in the proximal centriole, distal centriole and the axoneme and, thereby, relative protein levels in human sperm (Turner et al., 2021)(Figure 1)(Supplementary Table 2). Since the FRAC ratio is a relative number, increase in a FRAC ratio can be because of an intensity increase at that location or because of an intensity reduction in one of or both two other locations. Therefore, FRAC ratio in the Axoneme can go up if there is intensity reduction in the PC, DC, or both, or alternatively if a centriolar protein mis-localizes to the axoneme.
8. FRAC experiments and raters are highly reproducible
The repeatability of FRAC tests is important for its reliability. Therefore, to gain insight into quantification reproducibility, we performed two types of evaluations.
First, we have compared three raters that analyzed the same four different pictures. Each picture included ~10 sperm for a total of 40 sperm per rater. Each rater generated nine averaged FRAC ratios per picture for a total of 36 values. We found the raters had an excellent intraclass correlation coefficient (ICC) of 0.9900 with 95% confidence interval between 0.994 and 0.998. Therefore, we concluded that FRAC have “excellent” inter-rater reliability (Supplementary Figure 1).
Second, we assessed the overall reproducibility of independent stainings. In this study, we analyzed one straw from each bull by dividing its sperm to multiple coverslips. Then, we performed at least six independent stainings of high-quality sperm on six slides prepared from each straw. The first set of three slides were stained with Tubulin, Acetylated tubulin, and POC1B. Additional sets of three slides were stained with Tubulin, Acetylated tubulin, and FAM161A. Therefore, for this analysis, we compared the tubulin and acetylated tubulin of the three independent stainings that included POC1B to each other as well as to the three independent stainings that included FAM161A, instead of POC1B. This process was repeated for the high-quality sperm of all 31 bulls (Supplementary Figure 2). We found excellent ICCs of more than 0.9400 for all bulls’ high-quality sperm. The 95% confidence intervals of 27 of the 31 ranged between 0.900 and 1, indicating an excellent inter-staining reliability. The ranges of the remaining four bulls were between 0.840 and 1, indicating a good to excellent inter-staining reliability.
Mean FRAC ratios of high fertility bulls’ high-quality sperm had a mostly Gaussian distribution.
FRAC testing assumes statistical Gaussian (i.e., “Normal”) distribution, which is needed to establish that 95% samples are expected to fall within two standard deviations of the mean (Chou et al., 1998). Therefore, we checked for normality using two tests: D’Agostino skewness test and Anscombe-Glynn kurtosis test (Ghasemi and Zahediasl, 2012) and found that most of the mean FRAC ratios in the fertile bulls’ high-quality sperm population had near Gaussian/normal distribution (Supplementary Table 4).
In contrast, the distributions of FRAC values of for fertile bulls’ individual sperm mostly show narrow distributions around one peak that rarely exhibit Gaussian distribution (Supplementary Figure 3). The skewness (the degree of asymmetry of a distribution around its mean) is smaller than +/- one for all the mean ratios of Acetylated tubulin and POC1B and most of mean ratios of tubulin (PC and Ax) and FAM161A (PC and DC). The skewness was larger in DC tubulin and axonemal FAM161A mean ratios (1.44 and 2.06). The skewness of axonemal FAM161A is likely because it is a centriole specific protein that is rarely found in the axoneme, and the distribution is capped at zero with values distributing only to its right (Supplementary Figure 3L). Surprisingly, the skewness of DC tubulin mean ratios was also larger (~1.44) and indicated a skew toward the right (Supplementary Figure 3E). The kurtosis (sharpness of the peak around the mean) is below +/- three for all the mean ratios of acetylated tubulin, tubulin and POC1B and most of mean the ratios of FAM161A (PC and DC). The kurtosis of axonemal FAM161A mean ratios was larger (5.14), probably because it is capped at zero (Supplementary Figure 3L).
High fertility bulls’ high-quality sperm mean FRAC ratios have a narrow distribution with only 2% of the parameters and 16% of bulls being outliers.
A good test should have a small reference range and a large effective dynamic range to identify deviation from normality. The effective dynamic range is defined as the range outside the smallest and largest values of the reference range. FRAC’s reference range is the average mean ratio plus and minus two standard deviations in the high-quality fertile sperm populations. We found that the reference range of fertile bulls’ high-quality sperm had different sizes for the various biomarkers in the three different locations, but it was always less 0.25 (Supplementary Figure 4). Since FRAC ratios fall between zero and one, the effective dynamic range of the mean ratios for the data we have collected is, in all cases, at least three times the reference range. This is a remarkably narrow distribution which facilitates identifying centriolar anomalies with high sensitivity.
The average reference range for POC1B was relatively small (0.11), and high for tubulin (0.19), acetylated tubulin (0.20), and FAM161A (0.20) (Supplementary Figure 4). This reduced variability appears to be because POC1B levels in the axoneme are consistently low. Interestingly, this is different than the reported average reference range in humans which reported higher average reference range for POC1B (0.15) and lower average reference range for tubulin (0.10) and acetylated tubulin (0.08) (Supplementary Figure 4). This may be due to lower general fertility in humans, a selection bias in either bovine or human samples, or that there is a species-to-species difference.
Normally, a new test is compared to a gold standard test, a test that has been thoroughly tested and has a reputation in the field as a reliable method (Cardoso et al., 2014). Because there is no “gold standard” to assess centriole quality, we use a reference range obtained from fertile bulls. An important feature of a good centriole-based test is that its reference range includes most of the high-quality centrioles for a given parameter. Indeed, we found that 21 of the 25 fertile bulls had all their 95% confidence intervals fall within the reference range. Only four of the 25 fertile bulls (16%) had mean ratio 95% confidence intervals outside of the high-quality sperm population’s reference range, making them outliers (Figure 2a and 2e). These include bulls 10, 20, and 22 (SCR = -1.5, 1, and 1.3, respectively) each had one outlier, while bull 29 (SCR = 2) had four outliers. Altogether, only 7 of the 300 analyzed parameters from high-quality sperm of fertile sires (2%) had mean ratio 95% confidence intervals outside of the high-quality sperm population 95% confidential interval (Figure 2a).
Low-quality sperm of fertile bulls have abnormal centriole protein distribution:
As sperm cells differentiate, they lose their cytoplasm and become denser; this distinction is used to separate the higher-quality sperm found in the pellet and the lower-quality sperm is found in the interface by differential centrifugation (Oshio et al., 1987; Sakkas, 2013). It is also during this differentiation that the sperm centrioles are remodeled (ref). Here, we analyzed the low-quality sperm population of the 25 fertile bulls (Figure 2B). Nine of the 25 (36%) analyzed fertile bulls had. This rate difference (9/25 versus 4/25) is not statistically significant (P=0.11). When examining individual parameters, the low-quality sperm from fertile bulls had 22 parameters FRAC ratio 95% confidence intervals outside of the high-quality sperm population reference range, this rate difference (7/300 versus 22/300) is statistically significant (P=0.004) (Figure 2E). The latter result suggests the fertile bulls' low-quality sperm population has lower quality centrioles relative to the fertile bulls’ high-quality sperm. Similar lower centriole quality results were reported with sperm of infertile men (Turner et al., 2021) but not in fertile men or men with unexplained infertility (Jaiswal et al., 2022).
9. High-quality sperm of subfertile bulls return positive FRAC tests
We have analyzed five subfertile Holstein bulls and one subfertile Jersey bull (Figure 2C). All the high-quality sperm populations of these subfertile bulls (6/6, 100%) had at least one mean ratio outside of the high-quality sperm distribution range of the fertile bulls’ high-quality sperm. This rate is much higher than high-quality sperm of fertile bulls (4/25, 16%), and is significantly different (P=0.00008) (Figure 2E). These six bulls’ high-quality sperm populations also had nine outlier parameters, out of 72 parameters analyzed (13%). This is more than five times the rate of outliers observed in fertile bulls’ high-quality sperm (7/300, 2%), and is significantly different (P= 0.0001). This result suggests that subfertile bulls’ high-quality sperm population has inferior centrioles compared to fertile bulls’ high-quality sperm.
If we set a cutoff (aka threshold) of one FRAC outlier 95% confidence interval to indicate subfertility, the test is 100% sensitive, and 84% specific, which is a good diagnostic performance. A more stringent cutoff to indicate subfertility may be having two FRAC outlier 95% confidence intervals. Two outlier parameters were present in half of the subfertile bulls’ high-quality sperm populations (3/6, 50%). This rate is over ten times higher than high-quality sperm of fertile bulls (1/25, 4%) and is significantly different (P=0.0025). Such a cutoff has a sensitivity of 50% and specificity of 96% for subfertility.
None (0/6, 0%) of the subfertile bulls’ high-quality sperm populations had more than two outliers. However, one fertile bull did, (1/25, 4%) resulting in a larger proportion than in the subfertile population. This difference is not significant (P=0.62). We suspect that this is due to having a smaller group of subfertile individuals, and that, if the sample size were larger, we would see individuals with more than two outliers in their high-quality sperm population.
To gain more insight into FRAC ratios of subfertile bulls, we examine the distributions of their single-sperm FRAC ratios (ratio distributions) (see graphs on Supplementary Table 5). As expected, most of the single bulls’ ratios are distributed around one peak (Supplementary Table 5). All subfertile bulls had an apparent shift in the distribution relative to the reference population (marked in yellow highlight), Bull 1 and 2 in the Tubulin PC parameter, Bulls 2-6 in acetylated tubulin in the PC, and Bull 4 and 5 in acetylated tubulin in the DC.
However, Bull 6’s distribution had a long single tail toward the right with some additional smaller peaks of Acetylated Tubulin in the PC. Bull 6 also had sperm with abnormally high FRAC values in FAM161A in the PC and abnormally low FRAC values in FAM161A in the DC. This extended distribution suggests that Bull 6 has a sub-population of anomalous sperm.
Together, these results indicated that a bull that has at least one outlier variable has a 60% chance of being subfertile (positive predictive value), and that a bull that has two or more outliers has a 75% chance of being subfertile. Also, these observations suggest that the high-quality sperm population of subfertile bulls also have lower quality sperm centrioles than the high-quality sperm population of fertile bulls regardless of the type of comparison (number of bulls with an outlier or number of outlier parameters), as expected from our hypothesis. Overall, the data suggests that abnormal centriole protein distribution is common in the unexplained subfertile bulls. Therefore, selecting a bull with a negative FRAC test, may improve predictions of successful breeding.
10. Subfertile bulls’ low-quality sperm population has inferior centriole protein distribution compared to fertile bulls’ high-quality sperm population
We analyzed the six subfertile bulls' low-quality sperm populations (Figure 2D). Individual-parameter mean FRAC ratios outside of the reference range are found in both the low-quality sperm populations of subfertile bulls (6/72, 8.3%) and the low-quality sperm of fertile bulls (22/300, 7.3%), (P=0.77) (Figure 2E). Half (3/6) of the subfertile bulls’ low-quality sperm populations had multiple out-of-range parameters (from 2 of the 12 analyzed parameters), while a small percent of fertile bulls’ high-quality sperm population had multiple out-of-range parameters (1/26, 4%). This result suggests that subfertile bulls’ low-quality sperm population has inferior centriole protein distribution compared to fertile bulls’ high-quality sperm population (P=0.002). Overall, the above data suggests that abnormal centriole protein distributions are common in subfertile bulls. Therefore, selecting a bull with high-quality centrioles may improves prediction of successful breeding.
11. Acetylated tubulin is the most common biomarker of lower quality centrioles
We used two categories of centriole markers. Tubulin, POC1B, and FAM161A label structural proteins (Tubulin labeled the centriole microtubule skeleton; POC1B and FAM161A label the centriole luminal scaffold that in the DC forms the rods). Acetylated tubulin is a post translational modification of tubulin. Therefore, these markers allow us to examine which of these distinct functional aspects are more sensitive for identifying subfertility. To identify marker sensitivity, we counted the number of outlier parameters found in each of the four sperm populations (Table 2). We found that acetylated tubulin had the highest number of outliers in total and specifically in high-quality sperm from subfertile bulls, suggesting it has the best potential to identify subfertility.
Table 2 - Acetylated tubulin is the most common marker of lower quality centrioles:
Marker
|
Location
|
High-quality
fertile
|
Low-quality
fertile
|
High-quality
subfertile
|
Low-quality subfertile
|
Location Total
|
Ace Tubulin
|
PC
|
1
|
1
|
5
|
0
|
7
|
DC
|
0
|
2
|
2
|
3
|
7
|
Ax
|
0
|
5
|
0
|
1
|
6
|
|
Total
|
1
|
8
|
7
|
4
|
20
|
Tubulin
|
PC
|
1
|
1
|
0
|
0
|
2
|
DC
|
1
|
1
|
2
|
0
|
4
|
Ax
|
0
|
0
|
0
|
0
|
0
|
|
Total
|
2
|
2
|
2
|
0
|
6
|
POC1B
|
PC
|
1
|
2
|
0
|
0
|
3
|
DC
|
1
|
2
|
0
|
0
|
3
|
Ax
|
0
|
3
|
0
|
0
|
3
|
|
Total
|
2
|
7
|
0
|
0
|
9
|
FAM161A
|
PC
|
0
|
4
|
0
|
0
|
4
|
DC
|
1
|
0
|
0
|
0
|
1
|
Ax
|
1
|
1
|
0
|
0
|
2
|
|
Total
|
2
|
5
|
0
|
0
|
7
|
The number of outlier 95% confidence interval FRAC ratios of the 31 bulls used in this study. Acetylated tubulin has the highest number of outliers in total (yellow highlight) and highest number of outliers subfertile bulls in high-quality sperm (green highlight).
12. Acetylated tubulin in the PC corelates with SCR and subfertility
Because we found that acetylated tubulin most effectively identifies lower-quality centrioles (Table 2), we wondered if the mean FRAC ratios of acetylated tubulin have a quantitative correlation with SCR. We found a highly statistically significant (R=-0.58; P=0.0007) linear regression correlation between SCR and acetylated tubulin levels in the PCs of high-quality sperm (Figure 3). This SCR correlation is much higher than any other reported variables from semen analysis, whose R values range between 0.10 and 0.16 (DeJarnette et al., 2022). In PC Acetylated tubulin, the SCR score determined about one third of the variance observed (R2=0.3316). Based on this correlation, a cutoff of 0.31 in PC acetylated tubulin mean ratio identified five out of the six subfertile bulls and 0 of the 25 fertile bulls (Figure 3B). This cutoff value has a sensitivity (the true positive rate) of 83% (5/6), and specificity of 100% (25/25) in the study population. A weaker correlation was found in DC Acetylated tubulin (P=0.02, R2=0.1719) and no correlation was found in Ax Acetylated tubulin (P=0.90).
Interestingly, the odds ratio (the strength of association between test-proposed fertility status and accepted fertility status) is much higher in the test based on the 0.31 cutoff generated from the linear regression analysis than tests based on bulls’ positive FRAC test results (Figure 3B).
Together, this data suggests that FRAC linear regression analysis can help identify bulls with lower sperm quality that are not recognized by current rigorous semen analyses.
13. Anti-acetylated tubulin antibody marking the DC-Ax junction.
Because acetylated tubulin immunostaining was found to be a strong biomarker for subfertility, we examined its labeling pattern using stimulated emission depletion (STED) super-resolution microscopy (lateral resolution of ~30 nm) (Figure 4A) and four times expansion microscopy in combination with structured illumination microscopy (SIM) (lateral resolution of ~40 nm) (Figure 4B) or STED (achievable lateral resolution of ~10 nm) (Figure 4C). All three approaches yielded similar results; acetylated tubulin signals were present in barrel-shaped PC, a fan-like DC, and the Ax. We presume that acetylated tubulin signals correspond to the microtubules.
Cryo-electron microscopy of the DC found that all the bovine sperm axonemes microtubules are continues with the DC microtubules (Khanal et al., 2021). Consistent with that, the acetylated tubulin immunostaining central lines in the DC appeared continuous with Ax staining in 96% of the analyzed sperm (Figure 4D). These lines are probably the microtubule central pair of the Ax that extend into the DC. Unexpectedly, however, in most cases, at the junction of the DC and Ax, the acetylated tubulin signal appeared discontinuous in both the left and right sides of the DC (73% of the analyzed sperm) (Figure 4D). This staining gap suggests that there is a region of reduced acetylation of the outer microtubules at the DC-Ax junction.