The study of sperm head vacuoles using deep learning algorithm and its correlation with protamine mRNA ratio

Background As regards the routine semen analysis is not sufficient to assess male fertility status, is it necessary to use other morphological sperm examination that may be more relevant in regard to the promotion of assisted reproduction outcomes? This study was designed for examination of sperm vacuole characteristics, its association with other sperm parameters and protamine 1 to protamine 2 ratio, and predict assisted pregnancy outcomes. Methods 98 Semen samples from subfertile men were classified based on Vanderzwalmen's criteria as follows: grade I, no vacuoles; grade II, ≤ 2small vacuoles; grade III, ≥ 1 large vacuole; grade IV, large vacuole with other abnormalities. The location, frequency and size of vacuoles were assessed using high magnification, a deep learning algorithm, and scanning electron microscope methods. The chromatin integrity (toluidine blue staining), condensation status (aniline blue), viability and acrosome integrity (triple staining), and protamination status (CMA3 staining) was evaluated for vacuolated samples. Protamine 1 and gene was analyzed by quantitative PCR. The assisted reproduction were also followed for each cycle.


Introduction
Forasmuch as normal sperm parameters were found in almost 15% of infertile males, the routine semen analysis is not sufficient to assess male fertility status (Perdrix et al, 2011). Nowadays, several assays have been suggested to progress the male infertility diagnosis with more details (lewis, 2007), which can be included sperm morphology (using light and electron microscopy), nucleus assessment (chromatin integrity and condensation, protamination status, aneuploidy), and the function of sperm (e.g., motility, viability, oxidative stress, acrosome reaction, sperm-zona pellucida interaction) (Perdrix et al, 2011). In addition, the evaluation of the detailed morphology of motile sperm was performed in real-time at a high magnification (up to × 6600) which is called the motile sperm organelle morphology examination (MSOME). In fact, MSOME is seen sperm morphology with more details, which ones are not appeared when viewed at × 400 or × 200 magnifications (Bartoov et al., 2002).
In addition, the successful fertilization is affected by competent sperm, the most important of which is replacement of DNA-binding histones by protamines (Steger et al., 2011;Rogenhofer et al., 2013).
Recently, an association between improper protamine mRNA/protein ratio and male infertility has been found (de Mateo et al., 2009;Hammoud et al., 2009;Depa-Martynow et al., 2012;Rogenhofer et al., 2013). This ratio is known as a suitable biomarker for successful fertilization (Carrell and Liu, excluded from this study to remove their effects on sperm quality.
The semen samples were collected via masturbation after three to four days of sexual abstinence.
The semen samples were analyzed according to the World Health Organization (WHO) criteria (WHO, 2010). The semen parameters such as pH, volume, motility, morphology, concentration, viability were assessed. Using high magnification (×1000), a novel deep learning algorithm (Javadi and Mirroshandel, 2019) and scanning electron microscopy (SEM), the semen samples were categorized into four groups according to Vanderzwalmen's criteria as follows: grade I) no vacuoles; grade II) ≤ 2 small vacuoles (which occupy < 4% of the head's area); grade III) more than two small vacuoles or ≥1 large vacuole (which occupy between 13% to 50% of the head's surface area); and grade IV) large vacuole with other abnormalities (Figure 1) (Vanderzwalmen et al, 2008).

Assay using a novel deep learning algorithm
Using deep learning algorithm, sperm morphology, especially vacuole was analyzed. This algorithm was performed with a high accuracy (94.65 %) to detect sperm's vacuoles. In addition, this method worked very fast and categorized sperm images in real-time. Therefore, the classification of spermatozoa was done using this algorithm in line with the results of high magnification (×1000) and SEM images (Javadi and Mirroshandel, 2019).

Scanning electron microscopy (SEM)
For correct measurement of sperm vacuole, each semen samples were evaluated by scanning electron microscopy (SEM) to observe the smallest details. The fixation of semen samples washed via density gradient centrifugation method was performed using Karnovsky solution for 30 min at 4 °C.
After centrifugation at 4,000 × g for 30 min, the samples were washed and the postfixation was done with 1% osmium tetroxide for 30 min. Afterwards, the ascending degrees of ethanol (50%, 70%, 80%, 90%, 96%, and absolute alcohol) were used to dehydrate. The drying was performed at a critical point (Balzers CPD-010). The specimens coated with gold (Balzers MED-010) were examined in a Philips FEM 515 scanning electron microscope.

Toluidine blue stain
The abnormality in the sperm chromatin structure was distinguished using toluidine blue (TB) staining. In this way, the air-dried smears on silane-coated slides were fixed in 96% ethanol-acetone medium (1:1) at 4°C for 1 hour. To hydrolysis, slides were put in 0.1 N HCl at 4°C for 5 min, then were washed. The staining was done with 0.05% toluidine blue (TB, in 50% Mcllvaine's citrate phosphate buffer, pH 3.5, Merck) for 5 min at room temperature (RT). Approximately 200 sperms were evaluated in each slide using a light microscope. The observation of light blue or deep violet/purple heads is the sign of existence normal or abnormal chromatin structure, respectively.

Aniline blue stain
The adhesion between lysine residues of histones and aniline blue (AB) stain were detected the abnormal condensation of sperm chromatin. Briefly, the smears were fixed in 4% formalin (Junsei Chemical, Tokyo, Japan). After washing, the slides were stained in 5% AB (Sigma-Aldrich Co., St. Louis, MO, USA) in a solution of 4% acetic acid (pH 3.5) for 5 min at RT. Almost 200 spermatozoa in each slide were observed under a light microscope. The sperms with dark-blue or colorless head were considered as abnormal or normal chromatin condensation, respectively.

Acrosome reaction assessment
The acrosome status (reacted acrosome or intact acrosome) was evaluated using triple staining and according to Talbot and Chacon method (Talbot & Chacon, 1981). In brief, sperms were put in 2% trypan blue (1:1), incubated at 37 °C for 15 min, and centrifuged at 600 × g for 5-10 min). Then, the sperm pellet was washed to obtain a clear/ pale blue supernatant. The sperms were fixed using glutaraldehyde (3% glutaraldehyde in 0.1 M cacodylate buffer at pH 7.4) for 30 to 60 min, centrifuged at 6000 × g for 5 min, and stained with Bismark brown Y at 40 °C for 5 min. The staining with Rose Bengal was done at 24°C for 20-45 min. Then, the smears were washed (in water), dehydrated (in an ascending degree of alcohol), and cleared (in xylene). At the end, almost 300 sperms in each slide were examined under a light microscope. Four staining templates were seen as follows: i) Dead sperm and intact acrosome as dark-blue post-acrosomal regions and pink acrosomes, respectively, ii) Dead sperm and degenerated acrosome as dark-blue post-acrosomal regions and blue/white acrosomes, respectively, iii) Alive sperm and intact acrosome as light brown post-acrosomal regions and pink acrosomes, respectively, and iv) Alive sperm and degenerated acrosome as light brown post-acrosomal regions and blue/white acrosomes, respectively.

Determination of sperm deprotamination by chromomycin A3
The sperm protamination was distinguished by chromomycin A3 (CMA3) stain ( Figure 2) as a detector of guanosine-cytosine-rich sequence. The fixation of air-dried smears was done in methanol/glacial acetic acid (3:1) for 20 min at 4 °C. Then, the slides were put in 100 µl CMA3 solution (0.25 mg/ml CMA3 in McIlvaine's buffer, containing 10 µm MgCl 2 ) for 20 min. The sperms with dull yellow staining (CMA3 negative) and bright yellow fluorescence (CMA3 positive) were considered as normal and abnormal chromatin protamination, respectively.

RNA extraction and first strand cDNA synthesis
RNA extraction was performed using the RNeasy Mini kit (Roche Molecular Biochemicals, Mannheim, Germany) and stored at -80 ˚C. The first strand cDNA synthesis was done the cDNA kit (Thermo Scientific, EU), according to the manufacturer's protocol at 42 ˚C for 60 min, and stored at -20 ˚C.

Real-time quantitative reverse transcription polymerase chain reaction
Real-time qRT-PCR was done to quantify mRNA transcript levels of protamine-1 (prm1) and protamine-2 (prm2) genes. The primer pairs of prm1, prm2, and GAPDH (housekeeping) genes were designed by GenBank at NCBI (Table 1). The analyzing gene expression was conducted by real time thermal cycler (Applied Biosystems, Foster City, USA) and QuantiTect SYBR Green RT-PCR kit (Applied Biosystems) was also used for amplifying the reference and the target genes (5 ml cDNA per sample) in the same run. The conditions were 95 ˚C for 5 min (holding step), 95 ˚C for 15 s, 58 ˚C for 30 s, and 72 ˚C for 15 s (cycling stem), which was followed by a melt curve step at 95 ˚C for 15 sec, 60 ˚C for 1 min, and 95 ˚C for 15 sec. All PCR amplifications were carried out in triple and mean values were calculated. Determining of relative quantitation for target genes was performed using ΔΔCT method. A first selection of motile spermatozoa in the poly-vinyl pyrrolidone (PVP) drop was performed at ×400 magnification during ICSI cycle. Photo-documented was analyzed using a novel deep learning algorithm as real-time and used for the grading. The best spermatozoa selected was immobilized and injected as described previously for conventional ICSI (Vanderzwalmen et al., 1996) at ×200 and ×400 magnifications. The injected oocytes were incubated. The selected spermatozoa were also assessed at ×1000 magnification (under a Hoffman modulation contrast) in line with deep learning algorithm' results.
If possible, the selection of the best spermatozoa without vacuoles and other abnormalities was considered for injection into the oocytes. Depending on the rate of sperm morphology impairment, it may take approximately 2 to 15 min to selection the best spermatozoa. After this time, the secondbest spermatozoa with the least number of vacuoles and/or other abnormalities were selected for injection. The number of oocytes that have to be injected also affected on the research for normal spermatozoa.
Then, the analysis of fertilization was done by observing of two pronuclei in 16-18 hours after ICSI or IVF. In addition, the evaluation of embryo development rate and embryo transfer (ET) was performed at four to five days after fertilization.

Statistical analysis
Using the Chi-square test, the relationship between vacuolization grade and other semen parameters was analyzed. Fisher's exact test was used to assess the statistical correlation of vacuolization grade with clinical variables. The categorized protamine-1 to protamine-2 ratios were analyzed using the Bonferroni-adjusted Mann-Whitney U-test. Statistical analysis was done using SPSS version 20 (IBM, Armonk, NY, USA) and the values with P<0.05 were considered significant. To evaluate the effect of each feature/factor of vacuole (size, location and frequency) on the sperm parameters and ART outcomes, CorrelationAttributeEval and Ranker modules of WEKA software were used.

Grouping of vacuolated spermatozoa
The classification of spermatozoa was done using a novel deep learning algorithm (Javadi and Mirroshandel, 2019). The results of this algorithm were similar to the results of high magnification (× 1000). In addition, scanning electron microscope (SEM) images showed more details of vacuole (size, location, and frequency) in the semen samples of each patient (Fig. 3). Therefore, grouping of samples was done very precisely based on the novel deep learning algorithm, high magnification (× 1000), and the SEM images.

Vacuolization and sperm parameters
The data of Table 2 show significant difference in the rate of abnormal condensation of chromatin sperm (AB staining) in the grade IV (p < 0.05) in comparison to control group (grade I). There was no significant difference in the viability and abnormality of the DNA structure of vacuolated spermatozoa among different grades (p > 0.05).

The protamine mRNA ratio in the vacuolated spermatozoa from fertile and subfertile men
The assessment of gene expression (prm1 and prm2) shows that there is a significant difference in prm1 gene expression between III (median 0.4457; p < 0.05) and IV (p < 0.0001) groups in comparison to control group (group I), while there is no significant difference between II group (median 0.83184) compared with the control group (median 1.0201; p = 0.09) (Fig. 4). In addition, the analysis of prm2 gene expression shows significant differences between groups of II (0.6623; p < 0.01), III (median 0.60262; p < 0.0001), and IV (median 0.2772; p < 0.0001) compared to the control group (median 1.001; group I).
The protamine mRNA ration was evaluated between different vacuolization grades in the fertile and subfertile men. Vacuolated spermatozoa from subfertile men with grade IV (median 3.40006 ± 1.81; p < 0.01) displayed a significant difference in the protamine mRNA ratio in comparison to control group (median 1.02 ± 0.81; p < 0.01).

Vacuolated spermatozoa and IVF/ICSI outcomes
The results of the influences of different grading of sperm vacuolization and normal sperm on the clinical outcomes are shown in

Factors Ranking
Another important experiment is to measure the effect of each feature/factor of vacuole, including size, location and frequency on male fertility potential. In this way, CorrelationAttributeEval and Ranker modules of WEKA software were used. Table 4 shows the effects of different features/factors.
The effect of vacuole location (nuclear) weighed more than the effect of other parameters on pregnancy.

Discussion
A novel insight was provided in this study that how vacuolization affects sperm fertility potential and is a better predictor of IVF/ICSI outcomes following evaluation of sperm using high magnification, deep learning algorithm, and SEM images. The results of this study show that variations in vacuole with higher size, greater frequency, and nuclear location were seen in protamine-deficient sperms (CMA3 positive and aberrant prm1 and prm2 gene expression) than in non-deficient ones. In addition, the presence of non-nuclear vacuole leads to increased immature acrosome reaction and decreased fertilization rate under IVF cycles.
Although many studies have indicated diagnostic limitations during the routine semen analysis for the infertile couples, this analysis is still performed in many clinical practices (López et al., 2013). However, this conventional semen analysis does not recognize the subtle abnormalities in the male genome, DNA structure and condensation (López et al., 2013). The abnormalities in chromatin structure and condensation is known to be correlated with numerous indicators of assisted reproductive outcomes, including fertilization, embryo development rate and quality, pregnancy and spontaneous miscarriage (Larson-Cook et al., 2003;Seli et al., 2004;López et al., 2013).
Although it is determined that human sperms have a highly dynamic and key roles in embryo development, the utility of more detail analysis of sperm is still a matter of debate (Sergerie et al., 2005;Cassuto et al., 2012;López et al., 2013). In the present study, the predictive value of vacuolated sperm testing was distinguished between potentially pregnant and not potentially pregnant couples under IVF or ICSI cycles. As mentioned above, these poor outcomes may have related to abnormalities of chromatin condensation and sperm deprotamination (CMA3 positive with aberrant prm1 and prm2 gene expression). While, the cause of abnormal sperm chromatin condensation is still unclear. The results of this study suggest a direct correlation between sperm nuclear vacuolization and abnormalities in sperm chromatin packaging. It seems that the contribution of the immune seminal cells, mature sperms and immature germ cells lead to the production of reactive oxygen species that can cause vacuolated head in sperms (López et al., 2013). It has been also reported that poor chromatin condensation and aneuploidy could be observed in spermatozoa with large vacuoles (Perdrix et al., 2011).
In this study, the protamine mRNA ratio in the fertile men was obtained median 0.793 ± 0.221 from vacuolated spermatozoa with different grade. In addition, this ratio from subfertile men was 0.739 ± 0.212 and 3.400 ± 1.281 in the vacuolated spermatozoa with grades III and IV, respectively. While, the protamine mRNA ratio has been reported in the previous studies as follows: 0.83 ± 0.05 (Carrell and Liu, 2001; n = 50), 1.3 ± 0.1 (de Mateo et al., 2009; n = 12) and 0.98 ± 0.02 (Nanassy et al., 2011;n = 77), a range of 0.54 to 1.43 of the protamine ratio has been reported in normozoospermic men (Nanassy et al., 2011). Therefore, this study's outcomes indicate that vacuolization affects negatively the protamine ratio in the subfertile men. So that, a low protamine ratio was seen in the vacuolated spermatozoa with grade III (protamine-1 was underexpressed). Also, a high protamine ratio was observed in the vacuolated spermatozoa with grade IV (normal expression of protamine-1 and underexpression of protamine-2). Aoki et al. (2006) reported prm-1 under-expression and prm-2 over-expression in infertile patients with a low protamine ratio. On the other hand, in patients with a high protamine ratio, prm-2 was underexpressed and prm-1 has normal expression. Numerous studies also indicated a significantly aberrant protamine ratio in infertile men (Carrell and Liu, 2001;Mengual et al., 2003;Nasr-Esfahani et al., 2004;Aoki et al., 2005;De Mateo et al., 2009) and our result is in line with the reported ratio in the above mentioned studies.
In addition, it is widely accepted that there is a correlation between sperm quality and infertility and our results are in accordance with the previous studies (López et al., 2013). In this way, the embryos resulting from morphologically abnormal sperm cells lead to significantly lower pregnancy rates (de Vos et al., 2003). The correlation between spermatozoa with large nuclear vacuoles and ICSI outcomes has been reported (López et al., 2013). While the origin and consequences of vacuoles of sperm head are also a problem of controversy.
Therefore, the association among different sizes, locations, and frequencies of vacuole with chromatin status, IVF/ICSI outcomes, and weight of each feature (size, location, and frequency of vacuoles) on pregnancy rate are essential that this study considered them. Kacem et al. (2010) showed that a large sperm head vacuole could originate from spermatogenesis damaging, abnormal maturation or modifications during the acrosome reaction. Our results are consistent with the results of this study. So that, the immature acrosome reaction was greater in spermatozoa with grade III and IV, therefore, the fertilization rate was decreased in these groups. Of note, to the best of our knowledge, this is the first study in which high magnification with SEM images and deep learning algorithm (Javadi and Mirroshandel, 2019) were used to assess sperm morphology. In addition, sperm parameters, including chromatin structure, chromatin condensation, protamination status, acrosome reaction with gene expression involved in chromatin protamination were studied to trace IVF and ICSI outcomes.

Conclusions
The results of this study indicate that the semen samples from subfertile men are characterized by a higher ratio of vacuolization grades, although there are also in normozoospermia samples. This frequency of vacuolization correlated to abnormal chromatin condensation, greater sperm deprotamination, declined prm1 and prm2 gene expression, and a high protamine mRNA ratio. Also tracing the IVF/ICSI outcomes shows that the poor fertilization rate during (IVF cycles), embryo quality, and declined clinical pregnancy rate may have related to the abnormal maturation and sperm head vacuoles. Therefore, the evaluation of the vacuole status of semen samples as a definite parameter before starting treatment cycles appears to be a useful technique to introduce the best treatment cycle (IVF or ICSI) in couples undergoing ART.

Ethics approval and consent to participate
This study was approved by the Guilan University of Medical Sciences committee. In addition, the informed consent was obtained from all the selected participants in the present study (IR.GUMS.REC.1397.154).     Scanning electron microscope images of vacuolated spermatozoa. The presence of small and large vacuoles, its frequency and located in the nuclear or non-nuclear position is clear.

Figure 3
Scanning electron microscope images of vacuolated spermatozoa. The presence of small and large vacuoles, its frequency and located in the nuclear or non-nuclear position is clear.

Figure 4
The protamine 1 and protamine 2 gene expression of vacuolated spermatozoa from subfertile men. A significant difference can be seen in the prm1 gene expression (underexpression) of spermatozoa with grades of III (p < 0.05) and IV (p < 0.0001) compared to control group. There is significant difference in