Patients and samples
Azoospermic men were double questioned before and after the operation and samples were ruled out after the operation, if they were not willing to continue their participation in the study. An approximately 50mg of fresh testicular tissues, were submerged immediately under the RNAlater stabilizing reagent (Ambion Life Science, Austin, TE, USA, AM7024) according to the manufacturer instruction. Micro-TESE open surgery team was fully informed as the first piece of testicular tissue was used for RNA extraction and the next pieces for pathology and SR. Submerged samples were stored at 4 °C for 24 h and then processed for further RNA extraction. Out of 50 samples included in, 40 were diagnosed as non-obstructive and 10 as obstructive-control individuals according to the pathological results. Inclusion and exclusion criteria were as follows: samples with weak RNA integrity, the ones with variable Cqs even after multiple rounds of separate analysis, and without clear pathology were omitted from this study. Unfortunately, we omitted 9 samples as they were reported with unknown pathology.
Written informed consents were taken and full explanation was donated to azoospermic men before sampling. The experimentations and consent forms were approved by institutional review board of Isfahan University ethical committee. All procedures performed in the study involving human participants were in accordance with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.
Schlegel technique was adopted and an expert surgeon has done all the micro-TESE open surgeries under the microscope in order to lessen the obstruction of testicular vessels12. Meticulous sperm processing with initial mechanical dissection of seminiferous tubules was followed by extensive exercise to receive the maximum rate of retrieval13.
Hematoxylin and eosin (H&E) staining of paraffin embedded tissues was performed according to the standard protocol14. Two microscopic slides containing at least 100 different tubular sections for each specimen were examined by a specialist pathologist and the results were reported as follows: (i) N: normal spermatogenesis with all types of spermatogenic cell lineages in sections, (ii) SH: seminiferous tubule hyalinization, (iii) SCOS: sertoli cell-only syndrome or germ cell aplasia, (iv) eMA: early maturation arrest, (v) lMA: late maturation arrest, (vi) Hypo: hypospermatogenesis. Individuals with normal spermatogenesis were considered as obstructive azoospermia (OA) and these were used as control individuals according to the previous reports15. Other pathologies with abnormal spermatogenesis were classified as non-obstructive azoospermia (NOA).
GEO database was explored with the keyword “azoospermia” for microarray datasets. A rigid inclusion-exclusion criteria was applied as follow and a total of 9 datasets corresponding to homo sapiens species were found. Among these datasets those with any treatments and therapies were excluded. Samples with cryptorchidism phenotype and with detected mutations were also excluded. In this regard, GSE145467, GSE45885, GSE9194, GSE108886, GSE9210, GSE14310 were selected. All the candidated datasets were log2 scaled and quantile normalized if necessary. Hierarchical clustering of each dataset was illustrated using Euclidian distance. Principal component analysis (PCA) plot were drawn and outliers were detected and removed. GSE9194 and GSE9210 were excluded, respectively due to low quality and low feature intersection with other datasets. SVA16 and Limma 17packages were used to remove batch effects and subsequently, PCA and hierarchical clustering were used again to check the quality of the batch effect removal. Effect size of features was calculated using Limma package with Benjamini-Hochberg correction. We applied p values to determine the corresponding false discovery rates (FDR). Finally, the variation of testis-specific thioredoxin gene2 (TXNDC2) alongside protamination genes (TNP1, PRM1, PRM2) were recorded. Testis-specific thioredoxin gene 8 (TXNDC8) was not included in GSE14310 dataset, and meta-analysis was performed on the resting GSE45885 and GSE108886 datasets. R version 4.0.1 was used for meta-analysis.
RNA isolation and cDNA synthesis
RNA extraction was carried out as reported previously2. Nanodrop One (Thermo Scientific, USA) was used for quantification and then, 1 μg of total RNA was treated with DnaseI (Thermo Scientific, Lithuania; EN0522) according to the manufacturer instruction. TaKaRa PrimerScript II 1st stand cDNA synthesis kit (TaKaRa, Otsu, Japan; 6210B) was used to random prime the first strand of cDNA. The qualities of the extracted RNAs were checked by 2% conventional agarose gel electrophoresis stained with ethidium bromide (data not shown).
Reverse transcription quantitative real-time PCR (RT-qPCR)
Primers were adopted for RT-qPCR and their concentration was optimized according to our previous study. SYBR Premix Ex Taq II (TaKaRa; RR820L) was the quantifying dye in Corbett 6000 Rotor-Gene thermocycler (Corbett Life Science, Mortlake, Australia). Equal amounts of cDNA were amplified in triplicates and the average cycle of quantification (Cq) values were further analyzed.
Melting curve analysis
After the last run of amplification, melting curve analysis via green channel was performed according to the manual of Corbett 6000 Rotor-Gene machine. Gradual increase in temperature (1.0 °C/s) was applied from 65 to 95 degrees Celsius and the amount of fluorescence emission was recorded continuously. The deviation of fluorescence change over temperature on y axis was plotted against the temperature on x axis using Rotor-Gene embedded software 1.7.
Gene expression analysis
GAPDH and RPL37 were used simultaneously as reference genes for RT-qPCR data normalization based on our previous finding2. REST2009 was used for statistical analysis.
Raw mean Cqs were exported to SPSS v.21.0 (IBM Corp., Armonk, NY, USA) and normalization of the data was carried out if necessary. Normalized mean Cqs of the genes were compared between individuals with positive and negative SR using t-Test. A one-way between subjects ANOVA-coupled with Scheffe Post hoc comparison was conducted to visualize the differences of mRNA expression levels between different pathological status. Multiple linear regression approaches were applied to model the relationship between the expression levels of PRM1,PRM2, and TXNDC2. A receiver operating characteristic curve (ROC) predictive model was obtained to demonstrate the predictive ability of the three expressed genes for SR. The area under the curve (AUC) was determined to assess the diagnostic accuracy. In all statistics, p values smaller than 0.05 were considered as significant.