1 Introduction

Infertility is defined as the inability of a couple to achieve spontaneous pregnancy after 1 year of regular, unprotected sexual intercourse [1]. Almost 8–12% of couples in reproductive age is affected by this global health issue [2]. Male factors contribute to approximately 50% of all cases of infertility and male infertility alone accounts for approximately one-third of all infertility cases.

The diagnostic ability of available male fertility investigative tools is limited.

Semen analysis (SA), also known as “seminogram”, is the gold standard technique for determining men’s fertility [3]. Unfortunately, SA provides limited information and cannot discriminate fertile from infertile men on an individual basis [4]. Moreover, widely overlapping ranges of seminal parameters have left clinicians in search of better seminal biomarkers. Therefore, new technologies are needed since a high percentage (ranging from 6 to 27%) of infertile men show normal semen parameters.

Advanced research allowed to explore new potential biomarkers for specific alterations in fields of genomics, proteomics and metabolomics, featuring the so-called “omics” era. The researchers are employing these novel biomarkers from blood, urine and breath [5, 6].

Through knowledge of spermatozoa metabolomics, we can reach a comprehensive understanding of what is happening in the cells and find the key of molecular mechanisms that regulate their biology [7].

In particular, untargeted metabolomic profiling is a powerful tool for biomarker discovery by observing the changes in metabolite concentrations of various biofluids and by identifying altered metabolic pathways [8]. An important branch of metabolomics, unexplored for human semen, is volatilomics. Volatilome is the totality of Volatile Organic Compounds (VOCs) derived from cellular metabolism and, recently, it is considered useful to a better understanding of physiological and pathophysiological processes [9].

In this work, we propose a new technique based on the application of a MOX sensor, used as additional detector after gas chromatograph (GC) separation, to evaluate metabolome profile of human semen applied to infertility study. Mass spectrometer (MS), which is the master detector of the system, is here used exclusively to identify the compounds detected by the VOC sensor. The two detectors work in parallel by a two-way splitter, that splits the helium (He) flow, eluting from the chromatographic column, in 1:1 ratio towards the two detectors through two segments of defunctionalized chromatographic column.

For data interpretation, statistical analysis was carried out by PLS-DA, the most known tool to perform classification and regression in metabolomics. One of the main advantages of PLS-DA is that it has the ability to analyze highly collinear and noisy data [10]. This method is particularly appropriate for the analysis of large, highly-complex data sets as are the sensor resistance profiles versus time (labelled “sensorgrams”) of all the samples. In addition, PLS-DA applied to sensorgrams may be an useful tool for generating parsimonious models through feature selection and data reduction, as well as providing more predictive results.

2 Materials and Methods

The study was focused on 57 reproductive-age men. First, semen analysis (seminogram) was carried out on semen samples collected from subjects at the Biological Medical Center “Tecnomed” (Nardò, Lecce—Italy). Next, on the same semen samples, headspace VOC analysis was performed by the GC/[MS + sensor] system.

Initially, semen samples were classified based on sperm motility in two groups: asthenospermic, with a progressive motility (PR) < 32% and normozoospermic (PR > 32%) samples. Subsequently, asthenozoospermic samples have been splitted in severe and slight asthenozoospermic ones.

The seminal VOCs were extracted by Solid-Phase Microextraction (SPME) technique; SPME was carried out by a Carboxen®/Polydimethylsiloxane (CAR/PDMS) fiber (cod. 57318, Supelco) which was exposed to each human semen sample headspace overnight. The GC-MS analysis of the extracted seminal volatiles was performed using a GC (6890 N series, Agilent Technologies) coupled to a MS (5973 series, Agilent Technologies) equipped with a ZB-624 capillary column (Phenomenex) with the injector temperature set at 250 °C to allow thermally desorption of VOCs. The carrier gas was high purity helium with a flow rate of 1 ml/min.

GC/MS analysis was carried out in full-scan mode with a scan range 30–500 amu at 3.2 scans/s. Chromatograms were analysed by Enhanced Data Analysis software and the identification of the volatile compounds was achieved by comparing mass spectra with those of the data system library (NIST14, p > 80%).

For GC/sensor analysis the capillary column was inserted, by a splitter (Agilent G3180B Two-Way Splitter), into a tiny chamber hosting a VOC sensor (MiCS-5521, e2v technologies, UK) and it was positioned near sensor surface. The MOX sensor operating temperature was 400 °C. The MOX sensor traces (resistance vs. time, i.e. “sensorgrams”) were used for data analysis. The sensorgram data obtained by GC/MOX sensor experiments and physiological data were elaborated by multivariate data analysis techniques using web-based tools available with open access server MetaboAnalyst (version 4). Partial least-squares discriminant analysis (PLS-DA) were applied to the preprocessed sensorgrams. In particular, to ensure that each donor profile is on the same scale, resistance profiles were standardized using range-scaling between 0 and 1 [11].

3 Results

Seminograms allow us to classify the 57 samples in three motility groups. We studied 16 human sperm samples with high motility, 20 samples with slight asthenozoospermia and 21 severe asthenozoospermic samples.

In this paper, the use of GC/MS system only aims to support sensor analysis and to identify the compounds inducing resistance variation.

In Fig. 1, overlapping of a chromatogram and the corresponding sensorgram (in relation to run time) for each sample group is shown. General background contamination, such as column bleed (signed by asterisks), is not perceived by VOC sensor.

Fig. 1
figure 1

Overlay of chromatograms and sensorgrams of one sample for each sample groups (high motility, asthenozoospermia and severe asthenozoospermia). Asterisk sign column bleed

Full size image

Forty-nine total VOCs are detected by GC/MOX sensor system. Some of these (43%) are present only in one individual, while the remaining 57% in at least two semen samples.

Despite high variability, some compounds have a different expression in the three sample groups, whereas other compounds are present exclusively in a group. In particular, those VOCs which are closely related with severe asthenozoospermia are shown in Fig. 2.

Fig. 2
figure 2

Sensorgrams (variation of resistance vs. time) of a severe asthenozoospermic sample (PR = 0%) and matched VOCs

Full size image

2D- and 3D-PLS-DA results obtained from sensor data were compared with data obtained by seminograms, considering 2 groups (asthenozoospermic and high motility samples) and 3 groups classification models where severe and slight asthenozoospermic samples are included into two separate groups. 2D-PLS-DA based on semen analysis showed an accuracy of 85% and of 65%, in the discrimination in two and three groups, respectively. In three dimensions, accuracy increase up to 92% (two classes) and 76% (three classes) (Fig. 3, panel A).

Fig. 3
figure 3

2D- and 3D-PLS-DA results from seminogram (panel A) and from sensorgram (panel B) data. 2 groups (top) and 3 groups (bottom) discrimination were carried out

Full size image

Using exclusively sensor resistance profiles, accuracy is of 77% (in 2D) and 78% (in 3D) for discrimination in asthenozoospermic and normozoospermic samples, while when we considered three classes the accuracy decreases to 37% in two dimensions and to 40% in three dimensions (Fig. 3, panel B).

Finally, we have investigated MOX sensor responses to different classes of organic compounds comparing VOCs detected by GC/MS and GC/MOX systems. Our results (Fig. 4) show that VOC sensor has different affinity to organic compound classes, having a higher reactivity with aldehydes (72% of total VOCs), ketones and acetamides (50%).

Fig. 4
figure 4

Covering of VOCs detected by GC/MOS respect those identified by GC/MS system, in relation with organic compound classes

Full size image

4 Discussion

Overlay of chromatograms and sensorgrams (Fig. 1) allowed us to highlight that chromatographic contaminants are not a problem for sensor analysis. In fact, while for gas-chromatograms, peaks and ions corresponding to column bleed must be removed, resistance variations of sensor occur only at the elution of those VOCs to which the sensor is sensitive.

Through time-correspondence between peaks of two systems, we are able to determine the VOCs responsible for resistance variation. The percentages of volatile compounds unique of a single patient and occurring in more patients (57% vs. 43%, respectively) reflect the well-known biological variability of human semen. Human semen (such as other biofluids) is a concentrate of compounds derived from cellular metabolism, but also from diet, environment and lifestyle. Result is a “melting pot” of different molecules [5].

Nevertheless, some VOCs have a different expression in patients with asthenozoospermia or with high sperm motility. For example, 1H, 1,2,4-Triazole is present exclusively in severe asthenozoospermic samples (Fig. 3) and Butanal, 3-methyl- and Butanal, 2-methyl- are more concentrated in samples with lower motility.

After normalization [11], sample resistance profiles are used to perform PLS-DA and to compare discrimination results with those obtained by PLS-DA of physiological data from seminograms (Fig. 4). This latter present a value of accuracy higher than that PLS-DA based on sensor data both in the classification in two and in three sperm sample groups. This is obvious if we consider that three of parameters in our seminograms is based on motility, that represents the characteristic on the basis of which we have classified our samples.

Anyway, the separation in asthenozoospermic and normozoospermic samples with sensorgram data is satisfactory (77–78%). To discriminate asthenozoospermic samples in severe and slight ones with a high accuracy, it is needful to increase the number of samples by a more extended experimental campaign.

VOC sensor showed a different affinity for organic classes, proving a higher response to aldehydes, ketones and acetamides (Fig. 4).

5 Conclusions

Despite very high biological variability of human semen samples, GC/MOX sensor system allowed us to develop a reproducible and novel method to detect human semen headspace VOCs based on which classify semen samples in relation with sperm motility. PLS-DA based on VOCs analysis by sensorgrams provided good classification between asthenozoospermic from normozoospermic samples, compared to PLS-DA based on sperm motility measured by seminogram (SA). This constitutes a first promising result in using semen volatilome detected by a MOX sensor in evaluate semen quality, and hence man fertility, by classification in standard classes.

Setup of a microGC coupled to VOC sensor could represent a new technology that physician and biologists could use to support gold standard method and improve the quality of the medical report on fertility by further additional information parameters.

Of course, further studies are required to extend this study to a larger sample population and to select the pattern of VOCs that can statistically discriminate asthenozoospermic from normozoospermic samples integrating VOC biomarkers analysis to current standard semen analysis. Moreover, large-scale semen volatilome characterization could be a powerful unexplored tool to understand pathway alterations in idiopathic infertility in which semen parameters are physiological.