**4. Discussion**

Based on our immunohistochemical analysis of human prostate tissues, we confirmed that AKR1C3 might be crucial in PCa occurrence and progression. In particular, this is the first study to report that AKR1C3 immunostaining increases along the treatment course, that is, AKR1C3 expression elevates from the hormone-naïve status to the CRPC stage in the same patient. Moreover, our study is unique in showing that the PSA PFS rate of patients with high AKR1C3 expression in cancer cells derived from RP specimens was lower than that with low AKR1C3 expression. The majority of our CRPC cases received LHRH analog collectively with bicalutamide as a primary hormonal therapy. Nevertheless, we could not evaluate the correlation of pharmacological treatment with AKR1C3 immunostaining, since CRPC specimens were mostly obtained just during the time of transition to CRPC, before administration of docetaxel or androgen receptor-axis-targeted agents (ARATs), including enzalutamide and abiraterone. To the best of our knowledge, there is no report of AKR1C3 expression after ARAT treatment; however, AKR1C3 activation, both in vitro and in vivo, using PCa cell lines, has been shown as a factor of resistance against ARATs [25,26].

Lin et al. was the first to report a high-titer isoform-specific monoclonal antibody for AKR1C3 and demonstrated AKR1C3 expression in stromal cells, though only faintly in epithelial cells in normal prostate; however, in PCa cells, elevated expression was observed by immunostaining [23]. The same group also reported AKR1C3 to be positive in immunostaining, in 9 out of 11 PCa cases, and showed variation from strong to negative immunostaining within the same tumors, as in our study [26]. They also found no correlation in staining patterns between AR and AKR1C3 expression, consistent with our study (data not shown) [27]. Tian et al. examined the primary PCa biopsy specimens and showed that AKR1C3 expression by immunostaining gradually increases with an elevated GS in PCa epithelium [16]. In our cases, there was no correlation between GS and AKR1C3 immunostaining, which is incompatible with Tian's results. This might be because our cases underwent RP and most of them had GSs of less than 8, while in Tian's report, half of the PCa cases had GSs of 8 or higher. Stanbrough et al. revealed that AKR1C3 expression, as per immunohistochemistry, showed negative-to-heterogeneously weak staining in most primary PCa, but intermediate-to-strong AKR1C3 staining in CRPC specimens, which is in agreemen<sup>t</sup> with our results [22]. AKR1C3 expression analysis of cancer cells derived from RP showed immunostaining to be correlated with PSA PFS after RP. The result is reproducible even if we adopt IS of AKR1C3 as a representative of its expression (data not shown). In our knowledge, this is the first report of correlation of AKR1C3 expression with PSA PFS after RP. Additionally, AKR1C3 expression was an independent factor of poor PSA PFS when we analyzed the prognostic factors by multivariate analysis including the initial PSA level and Gleason grade group, which were previously considered to be significant predictors of recurrence-free survival after RP [28]. In future, we should analyze AKR1C3 expression and survival after RP in a much larger cohort to understand its role in clinical practice.

AKR1C3 is a multifunctional steroid-metabolizing enzyme that catalyzes androgen, estrogen, progesterone, and PG metabolism [29]. It reduces DHT to form 3 α-diol, which is a neurosteroid that acts as a positive allosteric modulator of the gamma-aminobutyric acid type A receptor (GABA AR) [30]. The 3 α-diol stimulates AR-negative PCa cells through GABA AR. Further, it up-regulates the epidermal growth factor (EGF) family members in AR-negative PCa cells and stimulates EGF receptor and Src. These results together sugges<sup>t</sup> that AKR1C3 modulates intraprostatic neurosteroid that, in turn, activates AR-negative PCa progression. AKR1C3 is known to regulate its expression by ERG via direct binding to *AKR1C3* gene [31]. Furthermore, ERG and AKR1C3 expression in human metastatic PCa tissues was revealed to positively correlate with each other by immunohistochemistry. AKR1C3 is known to regulate the stability of ubiquitin ligase Siah2, and thus enhance the Siah2-dependent regulation of AR activity via non-catalytic function [21]. Wang et al. reported that AKR1C3 could drive epithelial–mesenchymal transition (EMT) by activating the ERK signaling pathways and up-regulating transcription factors such as ZEB1, TWIST1, and SLUG, thereby facilitating PCa metastasis [19]. Therefore, AKR1C3 might be crucial in PCa progression.

There are several limitations in our study. The samples analyzed were relatively small, in particular, in consecutive PCa with CRPC progression. We were unaware of the role of AKR1C3 overexpression, in particular, in progression to CRPC, since we did not analyze the tissue concentrations of T and other androgens. Moreover, we did not know how AKR1C3 overexpression correlated in response to new ARATs, namely enzalutamide and abiraterone. Only three cases received ARATs before CRPC tissue extraction. Case 5 received abiraterone 3 months before laminectomy, case 7 was administered enzalutamide 8 months before TUR-P, and case 11 acquired enzalutamide 10 days before TUR-P.

In conclusion, expression of multifunctional AKR1C3, which is known to be the most up-regulated steroidogenic enzyme in patients with CRPC, might increase with the occurrence and longitudinal progression of the tumor in certain cases [17,32]. Targeting AKR1C3 might overcome the complicated disease, CRPC, and also control resistance against ARATs. We plan to further reveal the functions of AKR1C3 in PCa progression and discover potent and specific drugs to inhibit AKR1C3 function.

**Supplementary Materials:** The supplementary materials are available online at http://www.mdpi.com/2077-0383/ 8/5/601/s1.

**Author Contributions:** Y.M. and T.I. conceived the study. Y.M. and T.I. drafted the manuscript. Y.M. performed statistical analysis. Y.T. and S.S. performed pathological specimen evaluation and scoring. All the authors participated in patient enrolment and data collection, helped to draft the manuscript, and read and approved the final manuscript.

**Funding:** This study was supported by Grants-in-Aid for Scientific Research (project numbers: 15K10587 and 18H02936) from the Japan Society for the Promotion of Science.

**Acknowledgments:** We thank all members of Department of Urology and Diagnostic Pathology Kyoto University Graduate School of Medicine, for their devoted support and constructive advice.

**Conflicts of Interest:** The authors declare no conflicts of interest.
