Sperm morphology differences associated with pig fertility

Artificial insemination is routine in commercial pig breeding, and as such, the use of high-quality semen samples is imperative. Here, we have developed a novel, semi-automated, software-based approach to assess pig sperm nucleus morphology in greater detail than was previously possible. This analysis identified subtle morphological differences between samples assessed by the industry as normal and those assessed as abnormal. 50 normal and 50 abnormal samples that were initially categorised using manual assessment to industry standards, were investigated using this new method, with at least 200 fixed stained sperm heads analysed in each case. Differences in sperm nuclear morphology were observed between normal and abnormal samples; specifically, normal samples were associated with higher mean nuclear area, a consequence of a greater head width and a lower variability between sperm heads. This novel, unbiased and fast analysis method demonstrates a significant difference in sperm head morphology between normal and abnormal pig sperm and has the potential to be further developed to be used as a tool for sperm morphology assessment both in the pig breeding industry and potentially in human assisted reproductive technologies.


Introduction 41
Male fertility is a consequence of both the number and quality of sperm [1]. In humans, many issues 42 with male infertility are addressed using one of a range of assisted reproductive technology 43 procedures, such as in vitro fertilisation and intracytoplasmic sperm injection. In agriculturally 44 significant species such as pigs, cattle, and sheep, where a key goal is to maximise the production of 45 meat at a low cost, male fertility is also a challenge [2]. To this end, improving reproductive traits is 46 paramount. In such species, the critical aim is often to identify sub-fertile animals quickly and 47 cheaply so that they can be removed from breeding schemes; boars from a nucleus herd with 48 fertility problems have the potential to reduce litter sizes throughout the breeding population [3].

50
Artificial insemination (AI) is the oldest [4] and most routinely used technique in commercial animal 51 breeding, especially in livestock species [4][5][6][7][8][9][10]. Over the past three decades, the use of AI has 52 greatly benefited the pig breeding industry, particularly in Europe where over 80% of sows are bred 53 through AI [4]. In North America, the technique is also widely implemented, especially in large 54 farming units [9]. The principal objective of AI in the pig breeding industry is to permit the 55 dissemination of genetics from high genetic merit boars to as many sows as possible. Without AI, 56 more boars would be needed and hence animals of lower genetic merit would be required in 57 breeding programmes. Moreover, the technique enables the opportunity to introduce superior 58 genetic traits into sows whilst reducing the incidence of disease transmission, an advantage that 59 does not exist with natural mating [11]. AI is achieved by depositing spermatozoa into the female 60 genital tract using artificial devices and processes. The standardised method of insemination is the

64
In humans, semen analysis is widely used to evaluate male fertility [12] and may also be used for the 65 determination of reproductive toxicity in therapeutic and environmental agents [13,14]. Various 66 4 physical characteristics of semen are assessed and whilst parameters such as volume, appearance, 67 pH, and viscosity are considered important [15], several studies have shown that good sperm 68 morphology is critical when determining semen quality and hence quantifying male fertility [16][17][18][19][20][21].

69
Generally, the cut-off values of what is considered "normal" vary, and are dependent on the fertility 70 clinic. However, the following benchmarks were published in the World Health Organisation's 5 th 71 edition of "normal semen analysis": morphology (≥4% normal forms), total motility (≥40%), vitality 72 (≥58% live), sperm concentration (≥15,000,000 per mL) and volume (≥1.5 mL) [12,22]. To date, a 73 number of studies have been performed to analyse semen composition [23,24] and to establish the 74 relationship between sperm quality and fertility in men [25][26][27][28][29][30]. One such study used several 75 comparative semen analyses of fertile and infertile men, to determine the most appropriate 76 measurements that could be used in the determination of fertility potential in men [31]. Here, it was 77 established that whilst threshold values for sperm motility, concentration and morphology could be 78 used in the classification of males into fertile, indeterminate fertility, or sub-fertile categories, these 79 measures cannot be used independently for the diagnosis of male infertility [31].

81
In livestock species, a key contribution to successful fertilisation following AI is also the use of high-82 quality semen during insemination. As such, routine assessment of semen quality is a standard 83 process in the animal breeding industry [

102
Based on these nuclear morphometrics, three subpopulations, namely, large, small-elongated and 103 small-round were identified. Whilst it has previously been shown that sperm shape differed between  Prior to preparation of samples for this study, semen samples were identified as either normal or 121 abnormal using computer assisted sperm analysis (CASA) followed by manual assessment.

122
Specifically, samples that had a normal morphology score of above 70% (obtained from CASA) and a 123 motility score of above 4 (motility was graded from 1 to 5, 1 being dead and 5 being excellent) were 124 graded as normal and those falling below these criteria were graded as abnormal.

Results 187
Analysis of sperm nuclear morphology from 50 normal and 50 abnormal animals yielded measures 188 from 11,534 and 11,326 nuclei, respectively. Correlation analysis of measured sperm head 189 characters indicated that, as expected, many of the measures were highly correlated (Table S1)

225
The three clusters identified in the training group were also recovered in the test group ( Figure S5), 226 with analysis of sperm head morphology showing the same differences between clusters (data not 227 shown). As in the training group, sperm heads from normal animals were overrepresented in cluster 228 1 and underrepresented in cluster 3 ( Figure S5). This supports the idea that the frequency of certain  (Table S2). Conversely, when the aim was to retain 238 samples that were identified as normal through cluster membership, 13 of the 20 normal samples 239 and 6 of the 20 abnormal samples would be retained (Table S2). Similarly, attempts to predict 240 fertility using other approaches -for example defining thresholds based on rates of variability within 241 samples -also resulted in the inclusion of abnormal animals or the exclusion of normal ones. To 242 formally assess the diagnostic capacity of cluster membership, a receiver operating characteristic 243 (ROC) analysis was also performed using proportions in Table S1 as the range of cut-off values for 244 the predictor variables. This identified that using the proportion of sperm heads that do not fall into  Sperm head morphology in pigs is usually assessed as being either normal or abnormal based on 276 CASA data and manual assessment of morphology undertaken by the industry. These analyses have 277 identified differences between the sperm heads of these groups, with sperm heads from normal 278 samples having a higher overall area -a consequence of greater width and height -and being less 279 variable in shape than those from abnormal samples (Figures 2 and S1-4).

281
Our analysis also identifies three clusters of morphology types (Figure 3). These clusters group sperm 282 heads that have a low variability and a high area (cluster 1), those that are long and narrow (cluster

294
Given that automated approaches to identify abnormal semen samples would be of value in pig 295 production, we sought to determine if the clusters we identified could be used as a predictive tool 296 for semen analysis. Whilst grouping based on the previously identified clusters was able to 297 successfully distinguish some samples as normal or abnormal, it was found that cluster membership 298 alone could not accurately predict fertility in all 40 samples in the test group. To further test the 299 diagnostic capacity of these clusters, ROC analysis ( Figure S6) was used to identify how suitable 13 cluster membership would be as a predictor variable in identifying whether a given semen sample 301 has a normal or abnormal profile. This analysis revealed that if the proportion of sperm heads that 302 do not fall into cluster 3 was used a predictor variable, there was a 72.3% chance of correctly 303 distinguishing a normal semen sample from an abnormal sample, with an AUC of between 0.7 and 304 0.8 generally being considered to be acceptable.

Conclusions 307
Here we show that morphometric analysis of pig sperm using NMA software distinguishes 308 morphologically distinct populations of nuclei. This analysis identifies differences in sperm nucleus 309 morphology between animals deemed to be commercially acceptable and unacceptable using 310 manual assessment to industry standards. These findings suggest that the use of NMA software 311 provides a high-throughput and more accurate method to identify pig sperm with a higher fertilising 312 potential. This approach may therefore have immediate utility for the pig breeding industry, and