MicroRNAs are small non-coding RNAs that are becoming more and more popular as biomarkers of disease, including cancer. They are interesting for clinical use in part due to their stability in bodily fluids, making them attractive non-invasive liquid biopsy biomarker candidates, but also due to ease of detection with relatively low-cost methodologies widely available, such as PCR-based ones (1). These microRNAs are also dynamic and versatile, reflecting the status of disease and often being useful for diagnostic, prognostic, and follow-up purposes. However, still, few of these promising microRNAs actually make it to full integration in the clinic, in part due to technological challenges related to detection and reporting (2).

Testicular germ cell tumors (TGCTs) are among the most common solid neoplasms arising in young-adult Caucasian men (3). For these tumors, a set of microRNAs that regulate embryonic development have proved their value as accurate liquid biopsy biomarkers of the disease (4–6). Among them, the hsa-miR-371a-3p has shown the best clinical results, outperforming the classical serum tumor markers nowadays available in the routine [alpha fetoprotein (AFP), human chorionic gonadotropin (HCG), and lactate dehydrogenase (LDH)], which show important limitations (7–11). Since the study of Voorhoeve et al. in 2006 (12), an overwhelming amount of evidence has been built in the last decade, demonstrating the accuracy of hsa-miR-371a-3p, determined by real-time quantitative PCR (RT-qPCR), for the diagnosis and follow-up of TGCT patients [except for pure teratoma (13)], with sensitivities and specificities of mostly >90% (14–17). This microRNA has a short half-life, correlates with tumor burden, can be reliably detected in several bodily fluids, and is able to predict recurrences and viable tumor in post-chemotherapy masses (18–27). The confirmation of these findings in large, prospective, multicentric studies increased even more the interest in this biomarker (28, 29), leading to organization of clinical trials (NCT03067181 and NCT04435756) and proposals for introduction of a quantitative test in the clinic (30, 31).

However, there is still room to attempt to improve such a test. An increased technical sensitivity is desirable for stage I patients with low tumor burden, sometimes missed by the hsa-miR-371a-3p test, not only for diagnosis but especially for early discrimination of patients at risk of relapsing, allowing the timely adjustment of the therapeutic strategy (surveillance versus adjuvant chemotherapy), sparing young patients from unnecessary cytotoxic treatments (32, 33). Moreover, most protocols for the determination of hsa-miR-371a-3p levels rely on preamplification steps, which may be argued to produce some variability related to increased cycling (besides being costly, time-consuming, and facilitating events of contamination and unspecific amplifications). Also, the commonly used protocols rely on normalization to housekeeping microRNAs (variable from study to study), giving a relative quantification (as opposed to the absolute number of copies of the marker), and the method of normalizing and reporting RT-qPCR data differs among studies, raising the need for a consensus procedure and pipeline (31, 34).

In recent years, droplet digital PCR (ddPCR) has emerged as a new methodology for accurately quantifying liquid biopsy biomarkers and for the detection of minimal residual disease (35, 36). DdPCR provides absolute quantification, obviating the need for normalization to housekeeping microRNAs and facilitating setting of cutoffs of positivity, and is less influenced by inhibitory substances, due to the partitioning process. However, its application for microRNA quantification is still largely unexplored [with few studies available (37–44)], and only one very recent study attempted to apply such protocol to hsa-miR-371a-3p determination in TGCTs (45).

In this work, we describe a pipeline for the quantification and reporting of hsa-miR-371a-3p in plasma samples of TGCT patients using ddPCR (DigiMir test) and provide a direct comparison with two other available RT-qPCR methods for the quantification of this biomarker. We perform several technical optimizations of the ddPCR pipeline, which obviated the preamplification step, and finally validate our findings in a larger cohort of plasma samples.

In this work, we describe a pipeline for quantifying the most remarkable liquid biopsy biomarker in the field of TGCTs based on ddPCR, being approximately 3 times cheaper than RT-qPCR-based techniques. In fact, and despite an increased interest on ddPCR for accurately determining biomarkers in samples with low burden and low input (like in circulating markers such as microRNAs and cell-free DNA in plasma), few studies have decided to invest on microRNA detection using this technology (42) [the same for circulating tumor cells (48)], perhaps because the more conventional RT-qPCR-based techniques have been performing well in this task, at low cost and with more disseminated know-how. DdPCR has emerged as a very robust way to determine mutation status in circulation (49), and also as a way to study DNA methylation patterns in circulating cell-free DNA (50). Studies on microRNAs are very recent, and report different protocols for achieving distinct needs. Tavano et al. reported a protocol to determine hsa-miR-1290 for the diagnosis of pancreatic cancer in plasma using ddPCR (37), with this method being superior to routine CA19-9 determination, while Zhao and collaborators investigated ddPCR to determine a panel of microRNAs diagnostic of gastric cancer (40). Campomenosi and co-workers compared RT-qPCR with ddPCR for the assessment of several microRNAs in serum of lung cancer patients and demonstrated a strong positive correlation between the two techniques, with equal or smaller variation in the ddPCR approach (38). This high concordance between the two methods reported by the authors led us to seek a similar methodological approach for our study. Indeed, we have performed back-to-back comparison of three independent pipelines for quantifying hsa-miR-371a-3p. Two of the protocols were RT-qPCR-based (the TaqMan Advanced method, where a global microRNA synthesis with preamplification is performed, and the TaqMan non-Advanced method, with targeted microRNA synthesis for the desired microRNAs and preamplification). While both pipelines produced similar results, the TaqMan Advanced methodology [less reported in literature specifically in the field of TGCTs (23)] produced more variability in the recovery of non-human spike-in ath-miR-159a and resulted in a lower sensitivity of detection of low burden stage I cases. We hypothesize that this could be explained by global preamplification (instead of targeted reaction) interfering with sensitivity of detection (51). For these reasons, the targeted approach for cDNA synthesis was considered more reliable and was chosen to translate into ddPCR quantification. Nevertheless, both RT-qPCR protocols detected hsa-miR-371a-3p in a manner consistent with disease burden, with the stage III patient rendering higher levels of hsa-miR-371a-3p.

In this work, we attempted to omit the preamplification stage by stepping-up to the ddPCR methodology. Preamplification may theoretically introduce bias by increased cycling and is cumbersome and frequently the most expensive section of the experimental protocols, but is highly adopted in most recent works in the field [as summarized in (31)], but was not performed in older works (9), because of the increasing sensitivity of detection of low amounts of circulating microRNA, such as in stage I small tumors. However, it can also elicit sporadic detection of trace amounts of hsa-miR-371a-3p in teratoma-only cases or even in healthy subjects, which are undesirable and could create discomfort when reporting the results—also evidenced in this work. In this study, we have optimized the DigiMir pipeline (after fulfilling ddPCR technical requirements and precision assessment, such as assuring the greatest separation of positive and negative droplets, temperature gradient, LOB and LOD calculation, and dilution series, as required and reported (39, 41, 44) and were able to quantify hsa-miR-371a-3p absolute copy number in a manner consistent with tumor burden (higher in stage III), with all negative controls and the healthy blood donor being below the threshold of positivity, hence being considered negative.

We designed experimental settings to illustrate variability in quantification of microRNAs with the three distinct pipelines, again with the TaqMan Advanced RT-qPCR protocol demonstrating to be poorer, with higher variability in results upon the various repetitions of the run under different conditions. Shifting to ddPCR, recovery of spike-in was homogeneous among samples [within the range reported in (39)] and variability of detection of hsa-miR-371a-3p was acceptable, with a strong positive correlation found between the amounts determined at varied moments [as reported also in (39)]. Most importantly, like Campomenosi (38), we also found the quantification results by RT-qPCR (normalized to housekeeping hsa-miR-30b-5p) and those of ddPCR (with the advantage of obviating normalization) to be strongly positively correlated.

Then, in an early attempt to provide clinical validation, we have applied the defined DigiMir pipeline to a set of 107 plasma samples, including those from TGCT patients, non-TGCT testicular masses, healthy male blood donors, and one patient with AFP secretion due to a hepatocarcinoma. Remarkably, hsa-miR-371a-3p identified all but two TGCT patients (which were also negative for AFP, HCG, and LDH), and correctly discriminated all 47 healthy blood donor male individuals as negative. The resultant specificity of 100% and a sensitivity of 94% are similar or even superior to the values reported in previous RT-qPCR-based studies (52) [summarized in (34)], outperforming the combined sensitivity of the three classical serum tumor markers (61.5%). Indeed, there were 11 TGCTs negative for all three classical serum tumor markers available in routine, which were, however, detected by our assay, representing an important clinical benefit in diagnosis. Furthermore, as additional negative controls from a clinical standpoint, the three non-TGCT testicular masses were negative for hsa-miR-371a-3p, a point in favor of the specificity of our assay. Non-TGCT masses are not infrequent and constitute an important differential diagnosis of TGCTs, which can however only be confirmed after orchiectomy is performed (53). Importantly, for small benign tumors, partial orchiectomy may be an option, allowing to spare fertility and function of the testis (54). A large study already reported that such masses are negative for hsa-miR-371a-3p by RT-qPCR (55), and we show the same by ddPCR approach. Additionally, our assay has proved to be useful in clarifying elevations of AFP due to causes other than TGCTs, as has been recently reported for RT-qPCR (56). The patient with an important elevation of AFP due to a hepatocarcinoma was negative (with 0 copies) for hsa-miR-371a-3p. Also, of interest, a seminoma patient showed an elevation of AFP pre-operatively, which he maintained post-orchiectomy and during follow-up. The elevation was interpreted clinically as constitutional since no further disease could be identified, but caused discomfort in clinical decisions, especially for the decision to put the patient on surveillance only. In this setting, our hsa-miR-371a-3p assay would have been useful to the clinic, since it was positive at pre-orchiectomy (due to the presence of the seminoma), but completely negative (0 copies) after orchiectomy despite the persistence of AFP, confirming that the elevation was constitutional or derived from other causes.

Finally, we also tested an additional setting of relevance, which is the follow-up of these patients (57). Follow-up routinely involves repeat measurements of classical serum tumor markers (which show important limitations in detecting relapses) and continuing imaging [both costly, with limited sensitivity for very small metastatic deposits and exposing young patients to radiation if computed tomography is performed (30, 58, 59)]. In this sense, hsa-miR-371a-3p measurements are attractive for adequate monitoring and for guiding treatment decisions (34). In our study, all follow-up samples from patients showing disease resolution and no signs of disease recurrence (negative imaging and classical serum tumor markers) were negative for hsa-miR-371a-3p, while the follow-up sample of a patient with stage III S1 disease was positive.

The data we present are overall corroborated by the very recent study of Myklebust et al. (45), who also describe a ddPCR pipeline for determining hsa-miR-371a-3p. However, there are important differences in methodology between the two studies: (1) whereas they used serum, we tested plasma samples, which may have an impact on microRNA determination, namely, on the levels of the housekeeping microRNA miR-30b-5p (18); (2) the RNA extraction kits and methodologies differed; and (3) Myklebust and co-workers did not include a spike-in (technical control) in their experimental setting, whereas, in our pipeline, spike-in recovery was demonstrated in all experiments as quality control (ath-miR-159a). Nevertheless, the main results of both studies completely concur in that ddPCR is a promising new methodology for reporting has-miR-371a-3p, with high sensitivity (89% for the authors and 94% in our study) and specificity (100% in both works) and associating with tumor burden. Additionally, both investigations report a good correlation and technical performance in comparison with RT-qPCR, having the advantage of obviating the preamplification step and allowing to report absolute copy numbers (without the need of a standard curve) in a clinical setting. Specifically, our study brings the novelty of including a seminoma with constitutional elevation of AFP. In our view, this further supports the relevance of our findings and suggests that the DigiMir pipeline will be also useful in the follow-up setting. Validation in larger cohorts of TGCT patients on surveillance, as performed in other studies (33), is warranted and would contribute to clarify the clinical utility of our assay.

To conclude, we provide a proof-of-concept investigation describing a methodology to quantify the most relevant liquid biopsy biomarker of TGCTs in plasma samples, disclosing adjustments to the protocol and technologies that can make the quantification less variable and more precise, achieving an accuracy of 97.5% in a validation cohort of 107 samples. A limitation of the study is the small patient cohort, not allowing to focus on specific settings such as stage I tumors with recurrence, or <1-cm masses. However, this constituted a first clinical validation of our pipeline, which we technically describe in detail. Further investigations and validation of the protocol are warranted in different clinical contexts. In particular, it will be interesting to investigate if ddPCR may improve the follow-up of stage I patients on active surveillance, or aid in the assessment of marker-negative small testicular masses and viable germ cell malignancy in the post-chemotherapy metastatic context, to better personalize treatment and monitoring of TGCT patients (60).

The original contributions presented in the study are included in the article/supplementary material. Further inquiries can be directed to the corresponding authors.

This study was approved by the Ethics Committee (CES-IPO-12-018) of Portuguese Oncology Institute of Porto, Portugal. The patients/participants provided their written informed consent to participate in this study.

JPS, JL, and VC performed molecular analyses and wrote the manuscript. JL collected the clinical data. TB-R assisted in specific tasks related to molecular analyses. JPS, JL, and VC analyzed the data. CC-M processed clinical samples. JM provided clinical information about the patients. RH and CJ supervised the work and revised the manuscript. All authors contributed to the article and approved the submitted version.

The authors would like to acknowledge the support of the Programa Operacional Competitividade e Internacionalização (POCI), in the component FEDER, and by national funds (OE) through FCT/MCTES, in the scope of the project EpiMarkGermCell (PTDC/MEC-URO/29043/2017). The authors would also like to acknowledge the support of MSD (“Prémio de Investigação em Saúde”), Banco Carregosa/Secção Regional do Norte da Ordem dos Médicos (SRNOM), and Fundação Rui Osório de Castro/Millennium bcp. JL is the recipient of a fellowship from FCT—Fundação para a Ciência e Tecnologia—(SFRH/BD/132751/2017). VC received the support of a fellowship from “la Caixa” Foundation (ID 100010434). The fellowship code is LCF/BQ/DR20/11790013.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.