Serum Levels of MicroRNA-371a-3p (M371 Test) as a New Biomarker of Testicular Germ Cell Tumors: Results of a Prospective Multicentric Study

1The serum tumor markers β-human chorionic gonadotropin, α-fetoprotein (AFP), and lactate dehydrogenase (LDH) became essential tools in the clinical management of testicular germ cell tumors (GCTs) in the late 1970s., Current guidelines recommend the use of marker measurements for clinical staging, treatment monitoring, and follow-up of patients with GCTs.- One major drawback of the markers, however, is their low overall sensitivity. Only 50% of all GCTs express one of the three markers, and seminomas lack AFP expression entirely., Moreover, LDH expression is also found in several other diseases.

2The use of serum levels of microRNAs (miRs) from the miR-371-3 and miR-302/367 clusters as novel GCT biomarkers was first suggested in 2011. Generally, miRs represent small noncoding RNAs that are involved in the epigenetic regulation of gene expression. Previous studies suggested a high sensitivity (> 80%) and specificity (> 90%) of miR-371-3 and miR-302/367 for GCTs, with miR-371a-3p proving to be the most sensitive and most specific.- The serum levels of this miR seem to be associated with both clinical stage (CS) and tumor bulk, with levels dropping to normal with a half-life of less than 24 hours after the cancer is cured., Of note, seminomas were found to express miR-371a-3p in more than 85% of patients.

3The current consensus is that measurements of miR-371a-3p, also called the M371 test, greatly outperform the classic markers, and thus, a clinical implementation of the test seems warranted. However, the available data are based solely on seven independent small- to moderate-sized studies with retrospective and prospective modes of patient accrual as well as divergent miR measurement techniques.,,- Moreover, the majority of patients examined thus far have been in the early CSs, and only a few had advanced stages. Accordingly, the body of evidence is both limited by and open to bias. Therefore, the aim of the current study was to prospectively evaluate the utility of the M371 test in a large and representative patient population enrolled from a large number of European institutions and to involve various histologies and CSs. In particular, we aimed to evaluate the diagnostic sensitivity and specificity of the test for the primary diagnosis of GCT and to assess its usefulness for monitoring GCT treatment.

4We performed the prospective study at 37 institutions in Germany, Austria, Switzerland, Hungary, and Italy between September 2015 and December 2016 (Appendix Table A1, online only). A total of 1,364 consecutive male patients ages 16 to 69 years were recruited. We excluded 490 patients from the study for various reasons (Fig 1). The final study population consisted of 616 patients with GCT and 258 controls (Table 1). Of the patients with GCT, 522 provided preoperative samples, and 118 with systemic disease underwent repeated sampling over the course of chemotherapy. Controls consisted of 133 males ages 18 to 60 years who presented with nonmalignant testicular disease (NMTD) and 125 healthy male blood donors of the same ages. The rationale for the sample size is detailed in the Appendix (online only). The following patient-related data were registered: date of blood aspiration, patient age, histology, primary tumor size, local pathologic stage (pT), CS according to the Lugano classification, serum levels of classic tumor markers, and treatment received. No information was available with regard to follow-up examinations of the included patients. All patients gave informed consent. The study received ethical approval by Ärztekammer Bremen (#301, 2015).

5For the measurement of serum miR-371a-3p levels, we used the method described previously. Briefly, RNA was isolated from cubital vein serum, and then reverse transcription was performed for both miR-371a-3p and the endogenous control miR-30b-5p to cDNA. Quantitative polymerase chain reaction was done after preamplification. Measurement results were documented as relative quantity (RQ) values. Laboratory details are provided in the Appendix.

6A two-sided Mann-Whitney U test was used to assess differences between two unrelated groups of samples. A Wilcoxon signed rank test was applied for the comparison of repeated measurements in individual patients. Receiver operating characteristic (ROC) analysis was performed with empirical data, and the optimal cutoff value (the highest Youden index) was determined. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and positive and negative likelihood ratios were calculated. For the calculation of predictive values, only patients from four large primary care urologic institutions were evaluated (see Appendix for rationale). Kernel density estimation was used as a model of the RQ distribution in an unlimited sample size. The 95% CIs for discriminative measurements derived from density estimation were calculated by bootstrapping, with 2,500 simulations. Differences among categorical data were calculated with an exact χ2 test. To test the association between tumor diameter and miR-371a-3p expression, linear regression was used. The association of tumor diameter and sensitivity was tested with logistic regression. RQ values were log-transformed for kernel density estimation and regression analysis, whereby values of 0 were assumed to be equivalent to 0.001. Bonferroni correction was applied to the comparison of miR-371a-3p sensitivity with the classic markers to adjust for multiple testing. All tests were two-sided, and significance was assumed at P <.05. Statistical analyses were performed using SPSS version 24 (IBM Corporation, Armonk, NY) or R version 3.2.3 (The R Foundation, Vienna, Austria) software.

7The median expression of miR-371a-3p was significantly higher in the entire GCT group and in all the CS subgroups compared with the controls. Patients with CS greater than I had a higher serum level than those with CS I (all P <.001; Fig 2A). Seminoma was found to have significantly lower miR-371a-3p values than nonseminoma. However, this difference was only detectable in CS I patients (Fig 2B). Teratoma had the lowest expression values of all subtypes (Fig 2C). Among controls, healthy blood donors and patients with NMTD did not have significantly different median RQ values (P =.4).

8An ROC analysis that was based on preoperative samples of the patients with GCT and controls revealed an area under the curve (AUC) of 0.97 and an optimal cutoff at an RQ of 5 (highest Youden index). On the basis of this cutoff, patients with GCT could be discriminated from controls with a diagnostic sensitivity of 91.8% and a specificity of 96.1%. After using kernel density estimation to model the distribution of RQ values, the AUC was 0.966, whereas the sensitivity was 90.1% and the specificity 94.0% (Fig 3A). PPV was 97.2%, and NPV was 82.7%. Table 2 lists a synopsis of all discriminative parameters for the entire group of GCTs and the subgroups of seminoma and nonseminoma.

9The M371 test discriminated patients with localized GCT (CS I) from those with systemic disease, with an AUC of 0.76 (Appendix Fig A1, online only) and a diagnostic sensitivity and specificity of 83.4% and 60.1%, respectively. Comparison of the sensitivities of the M371 test (empirical data) with the classic GCT markers (Fig 3B) revealed the sensitivity of the new test to be significantly higher than each of the classic markers and even the combination of all three. The superior sensitivity of the M371 test also was found in subgroup analyses of CSs and the two histologic subgroups (Figs 3C to F), with the greatest superiority documented in CS I (Fig 3C) and in seminoma (Fig 3E).

10We found a significant regression of tumor diameters with log-transformed miR-371a-3p serum levels (R2 = 0.653; P <.001) in CS I patients. Subgroup analyses also revealed this regression in seminoma, mixed nonseminoma, and embryonal carcinoma (R2 = 0.686, 0.745, and 0.619, respectively; each P <.001; Fig 4A), with much higher slopes of the regression curves in mixed nonseminoma and embryonal carcinoma than in seminoma (Fig 4A). Accordingly, the sensitivity of miR-371a-3p for detecting seminoma was significantly lower in the two lowest tumor size categories (≤ 9 mm and 10 to 19 mm) than in the larger categories (P <.001; Fig 4B). In nonseminoma, no divergent sensitivities were found among the various size categories (P =.8; Fig 4C). Localized tumors (pT1) had significantly lower median miR levels than advanced local stages (> pT1) among CS I patients (R2 = 0.664; P <.001; Appendix Fig A2, online only).

11Paired measurements of miR-371a-3p in 424 patients before and after orchiectomy revealed a significant drop in miR expression levels in both local and systemic disease (P <.001 for each category; Fig 5A). Of the CS I patients, 91.77% had decreased levels after surgery as opposed to 82.41% of patients with metastases (P =.008 for proportions of decreasing levels). Appendix Figure A3 (online only) shows additional details.

12Serial measurements of miR-371a-3p levels during chemotherapy in 118 patients with systemic disease revealed a significant decrease after the first cycle of therapy. In the 70 patients with CS IIa,b disease, subsequent cycles did not cause a further significant decrease in miR levels (Fig 5B). In CS III patients (n = 46), we observed another significant decrease in miR levels after the second cycle of chemotherapy, with no further significant changes with additional courses (Fig 5C). In two patients with mediastinal GCTs, miR levels dropped upon treatment in the same manner as in CS III patients (data not shown). Two patients had a fatal outcome and achieved strikingly elevated miR levels upon progression (Figs 5B and C as denoted with a †). Patients in the International Germ Cell Cancer Collaborative Group good prognosis category had significantly lower miR-371a-3p median values than those in the poor prognosis category (P =.04; Appendix Fig A4, online only).

13Patients with relapsed disease (n = 46) had a significantly higher median serum miR level than controls (P <.001; Fig 2D). Elevated levels were found in 38 patients, which corresponded to a sensitivity of 82.6%, a specificity of 96.1%, and an AUC of 0.921 for relapse detection. Serial measurements during treatment revealed significant decreases in miR levels in 28 of 29 patients (Fig 5D).

14This study provides a considerable body of evidence that supports the usefulness of miR-371a-3p serum levels as a new biomarker of GCTs. Five features of the M371 test are noteworthy: The test has a 90.1% sensitivity and a 94.0% specificity for establishing the primary diagnosis of GCT; it is relevant for the two main histologic subgroups of GCT; miR serum levels correlate with primary tumor size, local stage, and CSs; miR levels mirror treatment-related disease changes; and miR levels are elevated in recurrences.

15According to current guidelines, the classic tumor markers are not adequately effective for assessing primary diagnosis of GCT because of their low sensitivity and specificity.- Currently, clinical and ultrasound examination followed by surgical exploration represent the mainstays of diagnosing GCT. Scrotal magnetic resonance imaging is a second-line diagnostic tool; surveillance can be used in incidentally detected small masses., The high discriminative power of the test is evidenced by an AUC of 0.966 in the ROC analysis, the overall accuracy of 93.2%, and particularly the very high positive likelihood ratio of 23.675 (Table 2). Accordingly, in cases that remain indeterminate despite evaluation with all guideline-recommended examinations, the M371 test may add useful information for clinical decision making before surgery., One foreseeable weakness of the test could be in the detection of pure seminomas with sizes of less than 1 cm because only 59% of these express the marker (Fig 4B).

16Serum levels of miR-371a-3p significantly correlate with CS of GCTs (Fig 2A) and with the size of the primary tumor (Fig 4). The most probable explanation for this finding would be a close correlation between miR levels and the amount of tumor (ie, number of cells) present. In fact, this feature is one of the six prerequisites an ideal tumor marker is supposed to have., Accordingly, the M371 test can discriminate locally confined from metastasized disease with a sensitivity of 83.4% but with a specificity of only 60.1%. Computed tomography, the mainstay of clinical staging, has an overall accuracy of 82% with a sensitivity of 59% in detecting retroperitoneal metastases.,,, Therefore, even in conjunction with the analysis of classic markers, some cases remain unresolved. The M371 test might aid in assessing the correct CS.

17In contrast to the classic markers, both of the two main histologic subgroups expressed miR-371a-3p with frequencies of approximately 90%, with the exception of teratoma. In line with previous serum-,, and tissue-based studies,, we found CS I patients with pure teratoma to have very low to zero detectable miR-371a-3p expression. Of note, patients with metastasized teratoma displayed elevated miR expression (Fig 2C). Admixtures of other GCT components might have caused the elevation in metastatic teratoma. Nonseminomas have a significantly higher median serum level than seminomas in locally confined disease, and this differential expression has been reported previously but not unanimously., On the basis of the evidence that suggests that the miR-371-3 cluster is expressed primarily by undifferentiated stem cells,, it has been hypothesized that the close biologic association between embryonal carcinoma and undifferentiated stem cells results in higher expression of miR-371a-3p in nonseminoma than in the more differentiated seminoma and even less in the well-differentiated teratoma. Contrasting results were reported from a tissue-based examination wherein seminoma was found to have significantly higher expression of miR-371a-3p than nonseminoma. However, these findings must be assessed with caution because tissue miR levels do not correlate with serum levels.

18An important drawback of the markers AFP, β-human chorionic gonadotropin, and LDH is their overall low sensitivity and low specificity, particularly in the lower CSs., The median miR-371a-3p expression rates are likewise lower in early stages than in advanced disease; however, these differences are small and range between sensitivities of 86.7% and 98.4%. Hence, the M371 test exceptionally outperforms the classic markers at all CSs (Fig 3). Clinically, this feature is most relevant toward the management of seminomas, where more useful tools for the monitoring of patients under surveillance are needed.

19The hypothesis that serum miR-371a-3p levels correlate with tumor bulk is supported by our observation that serum levels decreased in response to therapy (Fig 5). After orchiectomy, 91.8% of all CS I patients displayed decreased serum levels, with the majority dropping to within the normal range. The reason why a small proportion of patients had inadequate decreases of miR levels after orchiectomy remains unresolved because there were no follow-up data available in this study. Occult metastatic disease is the most probable hypothesis to explain this finding.

20In systemic disease, orchiectomy results in a decrease in miR levels in many patients, although in most, the decrease does not reach the normal range. Marker production by metastatic tissue is the most likely explanation for this observation. Of note, a few CS II patients dropped to normal miR levels postoperatively (Fig 5A). Although the interpretation of this finding is hampered by the lack of follow-up information, a tempting hypothesis would be clinical overstaging in these patients, which would be in accord with surgical studies that documented staging error in approximately 20% of CS II patients. Serial miR measurements in patients with systemic disease revealed significant decreases upon chemotherapy. Of note, the majority of CS II patients dropped to normal levels after only one cycle, with insignificant further decreases after additional courses (Fig 5B). Because miR levels correlate with tumor bulk, this finding could possibly allow for the hypothesis that the cumulative chemotherapy dosage required for low-volume disease deserves reconsideration.

21In CS III patients, there is also a highly significant decrease in miR levels after the first cycle of chemotherapy, with the second cycle producing further significant decreases. At treatment completion, the majority of patients with metastases have normal miR levels. Treatment failure was documented clinically in a few patients, and was accompanied by rising miR levels. The two patients with a lethal outcome displayed increasing miR levels with disease progression and reached the highest values of all patients.

22Recently, two independent investigations likewise showed significant decreases of serum M371 levels in patients with metastases who received chemotherapy, which supports the value of this marker for monitoring treatment outcome., Moreover, the Princess Margaret Group demonstrated the value of the marker in identifying viable residual cancer in postchemotherapy residual masses and confirmed the lack of M371 expression in teratoma.

23The M371 test has an 82.6% sensitivity at an AUC of 0.921 to detect relapse. One reason why the test may detect recurrences at a somewhat lower sensitivity than primary GCT might be because miR-negative subtypes like teratoma and somatic-type malignancy occur with higher frequency upon recurrence.,, Nevertheless, the M371 test may have the potential to detect recurrences at an early stage in a substantial number of patients as has been recently documented.,,, MiR levels likewise decreased upon treatment of relapse as they did in metastasized primary disease.

24Limitations of the current study stem from the lack of clinical information on patient follow-up because several individual sequences of miR levels could not be sufficiently charted. In addition, histologic work-ups of orchiectomy specimens were performed by local pathologists without central pathology review, which involve minor uncertainties with regard to the differential assessment of nonseminoma. Among the serial measurements of patients with metastases, not all the patients had undergone measurements after each chemotherapy cycle. Accordingly, some missing measurement points may have reduced statistical power. All serum samples were deep frozen for several months until measurement; thus, some sample deterioration may have occurred in the meantime. Strengths of the study involve the multicentric, multinational design, which resulted in a minimal selection bias. In addition, laboratory measurement techniques were uniformly used for all samples analyzed.

25In conclusion, the current study strongly confirms previous data with regard to the usefulness of the M371 test as a new serum biomarker of GCT that is informative in both seminoma and nonseminoma. Because of its high sensitivity and specificity, this test involves the potential of simplifying clinical pathways of the management of GCT, although further validation in an independent cohort is needed.