1. Introduction
Germ cell tumors (GCTs) are developmental cancers since they recapitulate the various steps of embryonic and germ cell development [
1,
2]. They can arise in both the testes and ovaries and also in extragonadal topographies, reflecting the migration of nascent germ cells along the midline of the body towards the gonads [
1]. Among GCTs, the type II GCTs of the testis (TGCTs) are by far the most numerous and clinically challenging. These postpubertal-type tumors are among the most common solid neoplasms in adolescent and young-adult Caucasian men [
3], and derive from primordial germ cells/gonocytes that are arrested in their maturation, giving rise to a precursor lesion called germ cell neoplasia in situ (GCNIS) [
4], a central concept of the most recent World Health Organization Classification for these tumors [
5]. From GCNIS emerges the most common subtype―the seminoma (SE)―and, after a reprogramming process takes place, the various non-seminoma (NS) histologies―including embryonal carcinoma (EC), yolk sac tumor (YST), choriocarcinoma (CH), and teratoma (TE) [
5,
6].
This developmental perspective for TGCT origin and progression has led to the discovery of several biomarkers of the disease, both tissue-based (pluripotency factors, such as OCT3/4, SOX2, and SOX17) and liquid biopsy-based (including the ‘classical’ serum markers alpha fetoprotein (AFP) and human chorionic gonadotropin subunit beta (β-HCG)). Better non-invasive liquid-biopsy biomarkers are needed to overcome the limitations of the classical markers, which are overall elevated in only about 60% of patients at the time of diagnosis [
7,
8]. In the last decade, embryonic microRNAs of the 371~373 cluster (especially miR-371a-3p) have proved their value, outperforming the classical markers in various clinical settings [
9].
Another biomarker related to germ cell development is CRIPTO, also known as teratocarcinoma derived growth factor 1, TDGF1. It is an obligate co-receptor of NODAL, a member of the TGF-β signaling family, which has determinant roles in the process of commitment to the male germ cell fate, in embryogenesis, and in stem cell pluripotency more broadly [
10,
11,
12,
13]. The NODAL/CRIPTO signaling pathway is believed to regulate the delicate balance between the proliferation of germ cells and fate commitment, meaning that disturbance of this signaling cascade may lead both to infertility (due to insufficient spermatogonial stem cells) or to prolonged pluripotency, resulting in the development of TGCTs [
14]. This is in line with the finding of ectopic NODAL/CRIPTO signaling activation in NS [
15]. CRIPTO was also found to be at least partially regulated by promoter methylation [
16,
17]. As has been found for other neoplasms [
18], we previously showed, in a small set of samples, that CRIPTO is detectable in serum of TGCT patients [
16]. This result suggested that CRIPTO could constitute a non-invasive diagnostic marker of the disease, particularly if a sensitive assay can be developed. Moreover, and given the proximity of this bodily fluid to the testis, we hypothesized that CRIPTO detection in semen could prove a sensitive biomarker of TGCTs and a good surrogate to indicate the germ cell status of male patients, as was recently suggested for miR-371a-3p [
19,
20].
In this work, we assessed the value of CRIPTO as a serum diagnostic and prognostic biomarker of TGCTs, expanding our previous series, and assessed the sensitivity of detection in seminal plasma samples of patients with both TGCTs and infertility. Finally, we correlated our findings with the levels of miR-371a-3p detected in these bodily fluids.
3. Discussion
The biology of GCTs closely relates to developmental germline events; their emergence results from defective programming in their cells of origin―the primordial germ cells (PGCs) [
1]. During this period of vulnerability during fetal life, disturbances in the germ cell niche that alter PGC differentiation can lead to arrest in maturation and development of the precursor lesion GCNIS. GCNIS cells are the ‘cancer stem cells’ of the testis and give rise to the various TGCT subtypes after puberty [
23]. Despite a report of infrequent CRIPTO expression in urological malignancies (including only four testicular tumors) [
24], more recently we and others have described the high expression of CRIPTO in TGCTs and related cell lines (with a decreased expression upon exposure to differentiation-inducing agents, such as retinoic acid), and hypothesized that aberrant maintenance or re-activation of NODAL/CRIPTO signaling during fetal life may be a possible mechanism for the genesis of these tumors [
14,
16,
25,
26]. The role of CRIPTO during fetal germline development is linked strongly to pluripotency of these cells in mice [
14], and we have also shown it to be expressed in the spermatogonial stem cells of the adult human testis [
16]. NODAL signaling, which is facilitated by CRIPTO, has also been implicated in the maintenance of normal somatic testis development in humans [
11]. In this work, we aimed to explore the expression of CRIPTO in liquid biopsy samples of patients with TGCTs and of infertile males, extending our previous work [
16] and providing novel insight on the role of this pathway in regulating this balance between tumor development and infertility. We additionally aimed at correlating our findings with the relative levels of miR-371a-3p, the most promising TGCT liquid biopsy biomarker identified to date [
27].
In this study, significantly higher CRIPTO concentrations were detected in serum samples from NS patients when compared to controls, and readings were higher than for the SE patient cohort, although this was not statistically significant (
Figure 1A). The highest mean concentration of CRIPTO was found in patients harboring pure EC (1.76 ng/mL). These results are in line with our previous analyses where, using RT-qPCR and immunohistochemistry, we found the highest CRIPTO expression in EC (and YST) compared to other subtypes and controls, which was the reason for enriching our cohort on NS samples, together with the need to represent the various subtypes within this heterogeneous group of tumors [
16]. In our previous ELISA assessment of a small set of TGCT patient sera, we found the highest mean CRIPTO concentration in SE, and a reading for only 1/15 controls (6% positivity) [
16]. This discrepancy is likely accounted for by the small cohort of 44 tumors and 15 controls assessed initially compared to this larger study, comprising 270 tumors and 48 controls. Further, our current assay had greater sensitivity as we achieved 63% positive CRIPTO readings (background levels) in our control serum cohort (compared to 6% in the small study).
We note in our current analyses at least one obvious outlier in the control blood sera group with a very high CRIPTO concentration (10.52 ng/mL). If this one sample is removed, the overall mean CRIPTO control concentration is decreased, and statistical significance is increased for many of our subgroup analyses (as presented in
Supplementary Table S1). Although ‘control’ individuals were blood donors presumed to be healthy at the time of sample collection, we hypothesize that this outlier could be due to the presence of an undiagnosed tumor (somatic or TGCT) or other condition; unfortunately, this possibility cannot be excluded nor verified due to the anonymized dataset. Further enlargement of the control group to replicate natural interindividual variation is desirable. Our control group CRIPTO concentration mean of 0.79 ± 0.29 ng/mL (or 0.59 ± 0.16 ng/mL with the highest outlier removed) is similar to the control mean of Pilgaard et al. [
28] of 0.60 ng/mL, using the same assay.
For a diagnostic assay to be useful in the clinical setting, it must be specific and sensitive. Our assay detected 56% of NS and 48% of SE patients and had a false positive rate (in controls, outlier included) of 21% using a control CRIPTO concentration cutoff set to 0.79 ng/mL. This sensitivity was further increased to 57% NS and 52% of SE and with a false positive rate (in controls, outlier included) of 29%, using the lower cutoff of 0.59 ng/mL. These results are approaching those of Pilgaard and colleagues who detected 70% of new glioblastoma multiforme cases using this assay [
28] and are an improvement on our previous smaller analysis where we detected 36% of GCTs [
16]. Within the NS tumor cohort, CRIPTO concentration out-performed classical serum markers AFP and β-HCG for the indication of disease stage (stage I vs. II and III;
Figure 2A,
Table 2), in line with the findings that the highest CRIPTO expression in glioblastoma patients (ELISA) [
28] and esophageal squamous cell carcinoma (immunohistochemistry) [
29] correlates strongly with poorer survival prognosis.
Detection of GCNIS in individuals prior to tumor development is the ultimate prognostic/diagnostic goal to reduce the need for chemotherapy and radiation treatments and improve overall GCT patient survival. In our small GCNIS-only patient group, we detected 40% (2/5) of patients harboring GCNIS using the higher CRIPTO cutoff of 0.79 ng/mL, and this was further increased to 60% (3/5) of patients using the lower cutoff of 0.59 ng/mL. The assessment of larger studies that include validated tumor-free controls would help in setting the optimal CRIPTO cutoff value to more accurately determine tumor presence and possibly stage of severity. A larger sample cohort for GCNIS-only patients is also required to determine whether CRIPTO is a sufficiently specific and sensitive prognostic marker for this pre-cancerous condition. Further assay optimization, using alternate CRIPTO antibodies and concentrations, as well as possibly other assay platforms (such as ELISA-qPCR or multiplexing with other targets), may aid in increasing specificity and sensitivity to clinically useful levels. Relevant to this purpose, our analyses indicated a need for reasonably prompt determination of CRIPTO concentration in serum samples since slightly higher readings were found in serum samples stored for long periods (
Supplementary Figure S3A).
It is logical to assume that seminal plasma would be of higher relevance than blood sera for detecting specific testicular malignancies, including the presence of pre-cancerous GCNIS alone, given the closer proximity to the tumor/GCNIS cells of origin. Although we assessed CRIPTO concentration in a small cohort (
n = 6) of control (normospermic) seminal plasma samples, we found the mean concentration to be five-times greater than that in the blood sera control group (
Figure 4,
Table 3). This finding of higher basal levels of CRIPTO in semen is likely due to (1) endogenous CRIPTO expression in the spermatogonial stem cells (SSCs) of the human testis [
16], and (2) the closer proximity to the cells of origin (SSCs), without the need for the biomarker to cross the blood–testis barrier. Such higher basal readings may be more clinically useful, due to higher sensitivity (lower false-negative rate; we detected CRIPTO in 92% of the total cohort of seminal plasma samples, compared to 68% of the sera total cohort). This discovery also suggested that CRIPTO should be investigated as a fertility biomarker (independent of malignancies, which was the primary aim of our study).
Within our seminal plasma dataset of azoospermic men (which was our experimental setting for studies in this biofluid), we came across four patients with TGCT/GCNIS and tested them for these markers as well. Although the low number of samples does not allow us to establish any conclusion, we describe that CRIPTO can be detected in the seminal plasma of these patients. In this preliminary impression, the three tested TGCT cases showed higher baseline CRIPTO concentrations (10.24 ng/mL) than controls (4.26 ng/mL), while the GCNIS-only male showed levels comparable to controls. It is clear that a much larger dataset of controls, TGCT, and GCNIS-only samples is needed in order to confirm whether or not CRIPTO would be a useful cancer biomarker in this fluid, as we have suggested for blood sera.
To investigate the relevance of CRIPTO concentration to fertility, we assessed this in a relatively large and well-defined seminal plasma cohort (
n = 79) of azoospermic males. Despite the absence of spermatozoa, CRIPTO was detected at levels similar to controls for non-obstructive azoospermia with JS ≥ 4, and at higher average levels for non-obstructive azoospermia with JS < 4 (not statically significant;
Figure 5A). Further, CRIPTO was detected in non-obstructive and also in clinically assumed obstructive (epidydimal) azoospermia, but not in the single structural obstruction (CBAVD) sample. This finding is likely due to the fact that our ELISA detects the smaller cleaved portion of the CRIPTO protein that is shed from the cell surface (presumably the SSCs), rather than relying on the local presence of the cells of origin. The trend for higher CRIPTO with the lower JS may reflect the skewed abundances of cell populations in the testis: reduction/loss of differentiated spermatozoa (low JS; negative for CRIPTO) biases the total population towards undifferentiated (CRIPTO positive; SSCs) cells. Seemingly counterintuitive to this argument, however, we found that CRIPTO was strongly and positively correlated with the VCM index (
Figure 6). Yet, in this instance, it is important to note that this correlation existed within the control group, which had an overall lower average CRIPTO concentrations than the non-obstructive azoospermia and TGCT sample groups. The reason for detection of CRIPTO in the epididymal obstructive azoospermia cases is still elusive, but we hypothesize it might be due to incomplete obstruction allowing passage of the cleaved CRIPTO protein despite no passage of sperm cells. Alternatively, another origin, like the seminal vesicles, must be kept in mind.
Finally, in this study, we also re-visited previous CRIPTO methylation analyses using a new technology platform. We have previously shown by direct sequencing that the methylation levels of two CpG regions of the CRIPTO promoter were low in human (T)GCT cell lines and did not correlate with CRIPTO gene expression [
16]. In this study, using an EPIC methylation array, we found similar methylation readouts among the same four (T)GCT cell lines (representing SE and NS;
Supplementary Figure S1A), although overall levels were higher than previously determined (mean beta values between 0.75–0.86). In a previous analysis of TGCT tissue biopsy samples, we found a correlation between promoter methylation and gene expression for certain subtypes (EC, TE, and CH), although we failed to do so for others (YST and SE) [
16]. In this current study, we confirmed differential methylation levels among histological subtypes (
Supplementary Figure S1B,C), as described by Costa et al. [
17] (which showed higher CRIPTO methylation in NS). However, there was no clear correlation with CRIPTO concentration in blood serum (for example, methylation levels are relatively high in EC (
Supplementary Figure S1B), the subtype that also shows the highest average CRIPTO concentration,
Figure 1B). Differences in the specific promoter regions investigated between the two studies likely account for some of the discrepancies we are reporting here, although we can conclude that methylation of the CRIPTO promoter may be partially, but not completely, regulating CRIPTO gene expression in GCTs.
Because the microRNA miR-371a-3p is the most informative TGCT serum marker known to date, we were very interested to compare relative miR-371a-3p expression with our CRIPTO concentration dataset in both sera and semen. Assessment of over 80% of the same sera samples for miR-371a-3p revealed a weak, although positive, and significant correlation between the two biomarkers (
Figure 3A), and this correlation was stronger in the total 84 seminal plasma samples assessed for both (
Figure 3B), perhaps due to closer proximity of this bodily fluid to the testis’ source and, therefore, higher (and more sensitive) overall readings (as discussed above).
As has been reported previously, we confirmed robust and reliable detection of the NS and SE malignancies using miR-371a-3p expression relative to controls [
30]. Similarly to CRIPTO, both TE and GCNIS-only samples were not significantly different from controls (as reported previously [
31]), however, using the baseline cutoff of 21.55, it detected 92.2% of NS, 91.7% of SE, 60.0% of GCNIS-only, and 85.7% of TE (
Figure 1D–F). In our particular sample cohort, and unexpectedly, we were unable to detect a significant difference between NS tumor stages I, II, and III, as had been reported previously [
32]. We hypothesize this can be due to the limited number of samples, specific composition of the cohort, and possibly due to some cases of the overstaging of stage II TGCTs based on imaging criteria. However, our dataset did reveal that long periods in storage (up to 16 years) did not significantly influence readings, highlighting another versatility of miR371a-3p as a serum biomarker (
Supplementary Figure S3B).
Assessment of miR-371a-3p in the same azoospermic seminal plasma cohort in which we assessed CRIPTO concentration similarly revealed higher basal levels than in blood sera (as had been reported previously [
20,
33]) and levels not significantly differing from normospermic males (
Table 4,
Figure 5B). We did detect miR-371a-3p in the obstructive azoospermia and non-obstructive azoospermia cohorts, as well as the TGCT cohort and single GCNIS-only sample, which is similar to a previous report [
20]. In the non-obstructive azoospermia samples, we also found the highest average miR-371a-3p reading in the group with the lower Johnsen’s score; we hypothesize this may be due to a cell-type dilution effect as we described above for CRIPTO. While we expected to detect CRIPTO in non-obstructive azoospermic seminal plasma due to its ability to be cleaved and be expelled from the testis independent of any spermatozoa, it is still unclear how miR-371a-3p might also be detectable in this fluid. Since it was detected in non-obstructive azoospermic men, we hypothesize that SSCs may also be the source of miR-371a-3p; again, its detection in obstructive azoospermia samples (although in lower levels) deserves further investigation in future studies, and may indicate some contribution from the seminal vesicles.
Relative levels of miR-371a-3p have been shown to be positively correlated with sperm concentration [
19,
20], so it was suggested that this microRNA could function as a surrogate of germ cell composition/niche in males. In this work, as for CRIPTO, we found a positive correlation between miR-371a-3p and the VCM index (although not statically significant;
Figure 6). Such a correlation may be due to spermatozoa harboring miR-371a-3p as has been suggested [
20] as opposed or in addition to, immature germ cells/SSCs also expressing the microRNA [
4]. Our previous work has suggested that the source of miR-371a-3p in semen samples is the germ cell compartment [
19], with detection in testicular parenchyma and semen, but not in other urogenital tract locations.
Limitations of our work include its retrospective nature, and especially the limited number of samples in certain categories. Our preliminary data describing levels of CRIPTO and miR-371a-3p in seminal plasma samples of TGCT/GCNIS patients indicate that these biomarkers can be detected in these patients, but much larger studies are needed to conclude on the usefulness of these biomarkers in this context. This includes a large sample set of adequate normospermic controls and patients with TGCT particularly. For sera samples, more age-matched controls and GCNIS-only patient samples should be assayed, to conclude definitely on the usefulness of these markers for its detection, together with optimal storage and processing of samples. These future studies, ongoing in our Institute, will allow us to better set the relevant cutoffs according to the tested body fluid, and also to refine the ELISA assay for CRIPTO.