Despite rapid progress in our understanding of the cellular and molecular etiology of cancer and the development of countless new anticancer agents and therapeutic strategies, little has changed in the treatment of many cancers over the last few decades. For instance, the standard of care for acute myeloid leukemia (AML) which consists of combined cytarabine and anthracycline therapy has been fundamentally unchanged for the past 30 years (1). The long-standing presence of this strategy is owed to its effectiveness with a mean response rate up to 70% and a lack of superior strategies for most AML subtypes (2, 3). New targeted therapy strategies including monoclonal antibodies and small molecule inhibitors are constantly being developed; however to date, none of these targeted therapies have proven more effective than the standard of care with the exception of the use of all-trans retinoic acid (ATRA) in acute promyelocytic leukemia (APL) which has become nearly curable in the majority of cases (4).

Notwithstanding, drug resistance is a major therapeutic challenge in the treatment of AML. Failure of initial therapy can be observed in up to 40% of AML patients, and even when initial therapy is effective, up to 70% of patients eventually succumb to their disease due to aggressive relapse within 5 years (5–7).

The cause of poor long-term survival is primarily drug resistance, which is either intrinsic in patients that fail initial therapy or acquired after chemotherapy through selection or acquisition of mutations (8). Indeed, relapsed AML is often composed of cells that have distinct molecular and cytogenetic characteristics leading to deficiencies or perturbations in various pathways associated with therapeutic resistance including DNA damage response and drug targeting, cell cycling, cell death, and drug trafficking pathways due to increased or altered drug targets are commonly observed (Figure 1) (8–10). Consequently, outcomes of relapsed disease are abysmal, which highlights a desperate need for novel therapeutic approaches with potential to overcome or prevent therapeutic resistance.

The most common treatment for AML includes an anthracycline like daunorubicin and a nucleoside analog like cytarabine in the “7 + 3” regimen where daunorubicin is administered IV for the first 3 days concomitantly to the IV infusion of cytarabine for 7 days (48, 49). The 7 + 3 regimen is termed induction therapy (because of its intent is to induce remission) and has been in place since the 1960s (50). The aim of induction therapy is achieving complete remission (CR), defined clinically as myeloid blast counts in the bone marrow below 5% or minimum residual disease status (49). Efforts to enhance this regimen by escalating dose or adding a third drug has only resulted in increased toxicity with minimal improvement in patient survival. Upon achieving CR, treatment can be consolidated using high doses of cytarabine. Unfortunately, despite undergoing such aggressive chemotherapy regimen with all the associated toxicities and side effect, many patients still relapse within 5 years (48, 49). This is in part due to lack of targeting of leukemic-initiating cells, selection of rare pre-existing resistant AML clones, or the mutagenic effects of the treatments, all of which increase the probability of generating more aggressive AML.

Fundamentally, drug resistance occurs in cells which can evade or withstand treatment. While tumor heterogeneity may explain selection of a pre-existing clone with a favorable mutation, acquired drug resistance is generally defined as the ability of a cell to resist response to the drug to which it was initially responsive. Acquired resistance may be achieved through multiple dosing of the same drug or through as little as a single dose may be explained by the mechanism of drug action.

For instance, anthracyclines used to treat AML such as daunorubicin, doxorubicin, and idarubicin, intercalate DNA, and stall proper DNA replication events (51). Anthracyclines can also target topoisomerase II which normally binds to the scaffold/matrix-associated protein region (S/MAR) to resolve DNA supercoils (52, 53). By binding to topoisomerase II in its open DNA-bound conformation, a stall occurs which can lead to a double-strand break. These double-strand breaks may be fixed aberrantly through non-homologous end joining which can lead to gene mutation. One common mutation in AML, t4:11, occurs at an S/MAR (54–56). This mutation has also been shown to occur in significant proportions in secondary AML patients as well (56). Loss or translocation of the S/MAR may further modulate various miRNAs. As demonstrated by Chavali et al., protein binding to the S/MAR induces histone acetylation that leads to the increased expression of the miR-17-92 cluster and the miRNAs miR-221, miR-93, miR-17, and let-7b (57). As DNA damage is most likely to occur in these regions due to daunorubicin, it is likely that dysregulation of miRNA expression can be due to daunorubicin-induced damage directly.

Cytarabine, on the other hand, is a cytosine analog that terminates translation and replication events. It primarily inhibits cells in S phase (DNA replication) but can also inhibit the progression from G₁ phase into S phase (58, 59). It is known that cytarabine is first metabolized into the triphosphate bound product by deoxycytidine kinase (DCK) and other nucleoside analog enzymes whereby it can then incorporate into the DNA. It is shown that its incorporation can often lead to extensive DNA damage including chromatid breaks (60). Stalled replication forks can also lead to bypass mechanisms such as translesion synthesis (61). This method of DNA replication is more error prone and can lead to mutation events as well. Each of these mechanisms can be demonstrated to have direct or indirect consequences for miRNA function.

As described with both drugs, genotoxic effects can lead to breaks that are then repaired using homologous or non-homologous repair mechanisms leading to miRNA alterations and the upregulation of drug resistance mechanisms. Conversely, miRNA which regulate these associated pathways may also contribute to drug resistance when perturbed by increasing tolerance to DNA damage. For instance, ataxia telangiectasia mutated (ATM) is an important DNA damage sensing and DNA damage response protein that has been demonstrated to contribute to chemoresistance (62). In experiments conducted in leukemic HL60, NB4, and K562 cell lines, it was found that the overexpression of miR-181a leads to increased cell proliferation and increased cell cycling through ATM targeting and downregulation (63). Similarly, miR-128 was reported to affect the propensity for DNA damage in AML cells. In a study conducted in HL60 cells, it was observed that the transfection of miR-128 led to increased apoptosis, drug sensitivity, and the amount of DNA damage tolerated; however, the mechanism is yet to be elucidated (64). miR-128 is thought to be upregulated in various cancers, but its levels are reduced in AML cells carrying NPM1 mutations (Figure 2; Table 1) (65, 66).

Recently, Lai et al. identified a mechanism by which miR-128 is likely targeting Rad51 directly and leading to the increased DNA damage response in OCI-AML3 and MV4-11 AML cell lines. In these experiments, miR-128 led to the sensitization of these cell lines to sapacitabine, a novel oral nucleoside analog prodrug (46). In other cancers, Rad51 has been shown to be a direct target of other miRNAs such as miR-506, miR-103, and miR-107. Clinical significance in chemoresistant high-grade serous ovarian cancers was established for miR-506 while a miRNA mimic library screen revealed miR-103 and miR-107 as strong drug resistance contributors in the U2OS cell line, a model for osteosarcoma (Figure 2; Table 1) (45, 47). To date, proteins that are thought to be integral to the activity of anthracyclines and nucleoside analogs such as topoisomerase II and the DNA polymerases are not known to interact with miRNAs. However, topoisomerase II has been demonstrated to be downregulated in drug resistant subtypes of AML (67, 68). miRNA targeting may prove to be a mechanism of topoisomerase II downregulation, but more research is required to establish important links of miRNA-induced dysregulation of DNA repair machinery in drug resistant AML.

The cell cycle represents a series of events that require the input of various checkpoint proteins known as cyclins and cyclin-dependent kinases (CDK) to proceed into division (69). These proteins, in turn, receive input from DNA damage sensing proteins such as ATM/ATR and CHK1/2 (70, 71). The majority of rapidly dividing cancer cells can be found in one of two major phases: the interphase; which consists of G1, S phase (DNA replication) followed by G2; and the M phase, where cells undergo mitosis. Cell cycle manipulation can be a drug resistance mechanism as cell cycle arrest at different phases or quiescence can lead to chemotherapy evasion; however, increased proliferation can also contribute to resistance (72–75).

The process of cell division begins in G1 by the duplication of various proteins, chromatin remodeling, and the verification that the DNA is free of DNA damage. In a healthy cell, if substantial levels of DNA damage are found, ATM/ATR become activated leading to eventual CDK2 inhibition and arrest at the G1/S checkpoint through p21 signaling, where the mechanisms of action of many miRNAs have been elucidated (76). CDK2 has been found to be inhibited by miR-638, where it was demonstrated in HL-60, NB4, and THP-1 that an upregulation of miR-638 leads to a reduction in cell cycling and a differentiation block in APL (77). The differentiation block was found to occur at the G1/S checkpoint and differentiation inducers like ATRA were found to be more effective in cells with miR-638 downregulation (77).

CDK2 has been demonstrated to be a target of various miRNAs in cancer including miR-885-5p, miR-372, and miR-188 (Figure 3; Table 2). In contrast to miR-638 in AML, miR-885-5p was demonstrated to play a tumor suppressive role in neuroblastoma by inhibiting CDK2 and promoting senescence and apoptosis (78). miR-372 demonstrated targeting of both CDK2 and cyclin A1, which is highly expressed during S phase. Like miR-885-5p, miR-372 was demonstrated to play a tumor suppressive role as demonstrated in HeLa cells and tissue samples of cervical cancer (79). miR-188 was demonstrated to directly bind several genes which play a role in cycling such as cyclin D1, cyclin D3, cyclin A2, cyclin E1, CDK2, and CDK4 with varying degrees and it demonstrated modest knockdown of CDK2 relative to the other genes (80). In this study, it was found that the arrest occurs at the G1/S transition and that miR-188 plays a tumor suppressive role (80).

Other miRNAs such as the miR-16 family members famously known for downregulation of BCL2 (Figure 4) are also shown to simultaneously directly target several cycling genes such as cyclin D1, cyclin D3, cyclin E1, and CDK6 (Figure 3; Table 2). As demonstrated in the A549 cell line by Liu et al., this targeting and likely the targeting of downstream effectors leads to the arrest in G1 and at G1/S, a phenomenon observed by others (81, 85, 86). The targeting of Cyclin E has since been demonstrated as playing an important role in certain cancers such as cervical cancer and breast cancer (86–89). The miR-15 and miR-16 family may be response elements of E2F1 and as such, may be contributing to a feedback mechanism (90).

The transcription factor E2F family may also be a target of miRNAs. E2F7, a transcriptional response element gene implicated in cell cycling, is downregulated by miR-26a in AML (82). This inhibition in turn reduces c-myc transcriptional activation and sequential miR-17-92 reduced transcription, which has previously been implicated in promoting a differentiation block (82, 91, 92). When active, miR-17-92 members may be in part directly targeting p21 and promoting cycling, as demonstrated in MLL transformed leukemic cells by Wong et al. (83). The inhibition of E2F7 may lead to a reduction of miRNAs involved in proliferation such as miR-25, miR-26a, miR-27b, miR-92a, and miR-7 thus behaving as a regulatory mechanism (93).

In other instances, miRNAs can behave as direct inhibitors of their own transcriptional repressor thus behaving as autoregulatory elements. It has been demonstrated by Pulikkan et al. that this is the case for miR-223 and E2F1 regulation (84). E2F1, an important response element in G1/S, can repress transcription of miR-223 which in turn can repress E2F1 (84, 94, 95). The differentiation block observed in APL may be further exacerbated by miRNAs like miR-223 (Figure 3; Table 2). The complexity of interactions within miRNA–mRNA networks demonstrates the need for further analyses elucidating the major pathways of feedback and feedforward signaling.

In the majority of blast cells that experience sufficient levels of DNA damage upon chemotherapy, programmed cell death (PCD) will become activated. PCD may take the form of apoptosis or autophagy. Apoptosis is characterized by specific changes in morphology such as cell shrinkage and pyknosis (96). Autophagy, on the other hand, is characterized by cellular degradation and the re-introduction of catabolic products into anabolic processes (97, 98). Autophagy can play both a detrimental and a beneficial role in cancer cells and it can also contribute to the generation of leukemia (98–100). Apoptosis, on the other hand, while it is an essential component of normal cell turnover, only its downregulation will often be a major contributor for aberrant cancer growth and its further suppression can lead to drug resistance.

Drug activation and drug clearance can be altered in cells to reduce the effective dose of the drug. These proteins are highly varied, but can largely be characterized into two major classes: the phase I and the phase II class of enzymes. Phase I enzymes typically perform redox reactions or hydrolysis reactions. While they often precede phase II enzyme activity, this is not always required. Phase II enzymes typically increase the polarity of the molecule through the addition of a sub-group such as UDP-glucoronate, sulfate, methane, acetate, or glutathione (143).

Anthracyclines are active drugs that can carry out their genotoxic effects directly. Their metabolism into the semiquinone form, the hydroxyaglycone form, deoxyaglycone form, or the alcohol form will decrease its likelihood of intercalating DNA as it reduces the anthracycline’s lipophilicity. It is unclear whether the anthracyclines lose efficacy through metabolism. As demonstrated from cardiotoxicity assays in rat and rabbit, the metabolites may have differing effects depending on the organism in question and the rate of metabolism. In rats, the alcohol form may retain some activity, but the effects of the active drug are more pronounced (143, 144). In rabbits, the alcohol derivative is implicated in the cardiotoxic effects of the anthracyclines (143, 145). It is thought that the enzymes CBR1/3 and AKR1A1/C3 can act on the parent drug to form the alcohol form. The hydroxaglycone and the deoxyaglycone forms can be generated in part by certain cytochrome P450 (CYP) enzymes such as CYP3A4/5, CYP2D6, xanthine dehydrogenase (XDH), and NAD(P)H quinone dehydrogenase 1 (NQO1) (146–150). XDH, NQO1 along with nitric oxide synthase can help in generating the semiquinone form (151–153).

Cytarabine and other nucleoside analogs require phosphorylation through DNA/RNA synthesizing enzymes such as the nucleoside kinases to become candidates for incorporation into nascent DNA. Cytarabine requires activation by several enzymes including deoxycytidine monophosphate kinase, nucleoside diphosphate kinase, and the rate limiting DCK (Figure 5) (154, 155). It is then metabolized by various enzymes including CYP3A4, 5′ nucleotidase, cytidine deaminase, and deoxycytidylate deaminase (154, 155).

Currently, there are few publications that highlight the role of miRNAs in anthracycline and cytosine analog metabolizing enzymes in AML. However, certain miRNAs such as miR-27b and miR-298 have demonstrated direct binding of CYP3A4 in a pancreatic cell line and miR-577, miR-1, miR-532-3p, and miR-627 were found to target CYP3A4 in HEK 293T cells (Figure 5; Table 4) (156, 157). In gemcitabine resistant colon and lung cancer cells, Hodzic et al. established a correlation between miRNA-330 and DCK expression levels (158). Further studies interrogating the role of computationally predicted miRNAs and miRNAs discovered in other cancer subtypes may help establish a role for miRNAs in metabolism in drug resistant AML.

The trafficking of the anticancer drugs can dramatically modulate treatment response as a reduction in influx or an increase in efflux will reduce the effective intracellular concentration of drug. Due to the lipophilicity of the anthracyclines, they can freely diffuse into the cell, but they can also bind to the SLC22A16 solute pump to enter cells (153, 159–161). While there are some reports that suggest the role of SLC22A16 in bleomycin resistance, the role of this transporter in anthracycline resistance is yet to be explored (162, 163). As such, while there are predicted miRNA-binding sites on this protein, none are yet confirmed.

Cytarabine and other cytosine analogs, on the other hand, necessitate the function of nucleoside transporters to enter the cell. The nucleoside transporters are composed of six major protein families: human equilibrative nucleoside transporters (hENTs) and human concentrative nucleoside transporters (hCNTs), organic anion transporters, organic cation transporters, peptide transporters, and the multidrug resistance protein family (MRP), with the hCNTs and hENTs playing the most major role of cytarabine import (164–166). In childhood leukemia, the hENT protein family has demonstrated to correlate with cytarabine resistance, but miRNA-mediated mechanisms are yet to be confirmed (167, 168).

In contrast, many efflux pumps can confer resistance to diverse and seemingly unrelated drugs and the characterization of several of these transporters has been extensive in AML. These ATP-binding cassette (ABC) proteins can be upregulated in the drug resistant forms of cancers and as such, the downregulation of miRNAs that target efflux pumps can contribute to resistance. Within this class, ABCB1 (P-glycoprotein, MDR1), ABCC1 (MRP1), ABCC2 (MRP2), and ABCG2 (BCRP) have been the most extensively examined out of 48 proteins within this functionally similar class (Figure 6) (169, 170). Indeed, previous treatments of drug resistant AML centered on the targeting of P-glycoprotein. It has been clearly demonstrated that the surface expression of P-glycoprotein is inversely proportional to the concentration of intracellular daunorubicin in blast cells and in tissue culture samples; however, blocking of P-glycoprotein did not yield positive results in clinical settings (171).

P-glycoprotein can be targeted by several miRNAs including miR-27a, miR-331-5p, miR-145, miR-298, miR-508-5p, miR-9, and miR-451 (Figure 6; Table 5). In leukemia, only miR-27a and miR-331-5p have been demonstrated to bind to P-glycoprotein in the K562 and HL-60 leukemia cell lines (172). In ovarian and cervix cell lines, it was demonstrated that the downregulation of both miR-27a and miR-451 can lead to downregulation of P-glycoprotein; however, in the case of miR-27a, this contradictory effect on P-glycoprotein is likely in part due to targeting of HPK2 upstream (173). This was further phenotypically demonstrated by the reduced uptake of intracellular dyes and by the response to cisplatin and methotrexate (174). In more recent experiments conducted in hepatocellular carcinoma cells, in addition to direct binding to P-glycoprotein and HPK2 binding, it was demonstrated that the inhibitory effect of miR-27a on P-glycoprotein may also be partially attributed to upstream modulation of the β-catenin pathway through direct binding of SFRP1 and potentially through FZD7 as well (172, 175). It is possible and likely that P-glycoprotein is involved in processes that are unrelated to drug trafficking as well such as apoptosis which may explain the contradictory expression in different cancers and the varying predisposition of its mutagenicity in certain cancers; however, its actions remain unclear (176).

Direct binding of miR-451 to P-glycoprotein transcripts was demonstrated in MCF-7 cells, where it was demonstrated that it could contribute to doxorubicin resistance; however, this has not yet been demonstrated to be clinically significant in cancer patients (179). In colon cancer cell-derived cell lines and HEK293 cells, it was demonstrated that miR-145 can play a role in the repression of P-glycoprotein and increase the efflux of rhodamine 123 (177). miR-298 was demonstrated to directly bind to the transcript in resistant breast cancer cell lines (178). This suggests that it may play a role in patients, but follow-up studies are needed. miR-508-5p was demonstrated to directly bind to P-glycoprotein in gastric cancers and its upregulation was found clinically as well (175). It has also been suggested by Munoz et al. that miR-9 may also target P-glycoprotein and confer resistance to temozolomide in glioblastoma multiforme cells (180). These miRNAs may also prove to be relevant in AML, but no studies have been attempted to date.

While the MRP1 gene has not demonstrated miRNA binding in AML, it was demonstrated in other cancers that the MRP1 gene can also be targeted by miRNAs such as miR-1291, miR-873, miR-221/222, and miR-326 (Figure 6; Table 5). In an analysis conducted by Pan et al., doxorubicin treatment of pancreatic cancer cells demonstrated that miR-1291 will become upregulated and target MRP1 directly (181). MRP1 downregulation contributes to multidrug resistance as well in other cancers such as ovarian cancer (182). It was recently demonstrated through in vivo and in vitro studies that miR-873 can be biologically significant in paclitaxel and cisplatin resistance in ovarian cancer cell lines where it can directly bind to MRP1 (182). Consequently, miR-873 is often downregulated in MRP1-dependent ovarian cancers. In melphalan-refractory multiple myeloma cells, Gulla et al. demonstrated that miR-221/222 may be binding and reducing MRP1 thus contributing to drug resistance (183). Finally, miR-326 was inversely correlated with MRP1 in multidrug resistant MCF7 cell lines (184). Less is known about MRP2 targeting by miRNAs, but in the liver cell line HepG2, miR-379 was demonstrated to be highly upregulated and to target MRP2 directly as a response to Rifampicin resistance (185).

BCRP, in contrast, has been shown to be a target of miR-520h, miR-328, and miR-519c and to potentially play a role in the hematopoietic system (Figure 6; Table 5). In CD34⁺CD38⁻ hematopoietic stem cells, it was demonstrated that miR-520h is enriched compared to CD34⁺ cells alone and that it can directly target BCRP in this fraction (188). An examination of miR-520h in leukemic cells and AML may demonstrate a similar trend of upregulation and a contribution of miR-520h to drug resistance, but more experiments are required. In mitoxantrone-resistant MCF-7 cells, Pan et al. showed that the expression of miR-328 is inversely correlated with BCRP and that it is directly suppressing BCRP, leading to resistance (186). To et al. demonstrated that miR-519c may play a role in downregulating BCRP in S1 colon cancer cell lines; however, they demonstrated that binding of miR-519c was limited to a longer form of the transcript only found in their parental cell line compared to their mitoxantrone-resistant counterpart (187, 189). This study highlights the importance of splice variants and how they may gain or lose miRNA-binding sites and thereby contribute to resistance.

Drug resistance is only a single aspect of clinical setbacks; however, it is a major contributor to therapy failure. Although treatment has improved substantially in some cancers in the past few decades, many other cancer types continue to demonstrate substantial patient populations that relapse after an initially successful treatment. While we focused on the regulation of drug resistance-associated miRNAs common between different cancers and drug classes, there are likely various miRNA that are specific to different drug treatments and cancers. However, the miRNA dysregulations discussed may have therapeutic value beyond AML. Furthermore, although we describe several drug resistance proteins, our analysis only focused on miRNA specifically implied in drug resistance where they were demonstrated to have direct activity and as such, the list is not exhaustive (190).

There are also many other molecular changes that occur in the development of drug resistance such as copy number variations, aberrant methylation, and aberrant post-transcriptional and post-translational processing (191, 192). The modulation of miRNAs offers a new perspective on drug resistance as miRNA replacement therapy and miRNA inhibition therapy raises the potential of developing new and effective drug therapies. Subtle miRNA changes can lead to significant changes in protein-coding gene expression and can consequently lead to changes in tumor progression and patient outcome. Experimental success in vitro and in vivo models may point to the likely coming of more miRNA-based clinical trials.

Previously, Mrx34 emerged as a promising therapy for the treatment of unresectable primary liver cancer. Due to multiple immune-related adverse events, this therapy was terminated in phase I although there was evidence of benefit in a subset of patients (193). Its promise came from being a p53-response element that was thought to mediate p53’s antitumor effects and consequently affecting downstream signaling in proliferation arrest and induction of apoptosis by targeting c-MYC, CDK6, and c-MET (194). However, recent research now demonstrates that it may not always behave as a tumor-suppressor either and furthermore, p53 may also be a direct target of miR-34a (138, 195). In liver cancers with β-catenin mutations, it is demonstrated that LNA-34a, a miR-34a inhibitor, displays antitumor effects. This is suggested to occur through blocking HNF-4α targeting which in turn decreases cyclin D1 and inhibits proliferation (196, 197).

A miR-16 mimic has also been recently introduced in patients in an open-label phase I clinical trial for mesothelioma and non-small cell lung cancer (NSLC). miR-16 was shown to be dysregulated in many different cancers (87, 89, 90, 103–105, 124, 198). A directed analysis in mesothelioma showed that miR-16 is reduced in patient samples and that a knock-in of a miR-16 mimic is tumor suppressive (198). This observation was repeated in xenografted mice with high success (198). Currently, there are no miRNA-based therapies for drug resistant AML or AML-related diseases.

Currently, there are two miRNA-based therapies intended to treat different cancers that are on-going or with pending results. MesomiR-1, a miR-16 mimic, was in a multi-center Phase I trial intended to treat mesothelioma and NSLC. This trial has been completed as of January 2017 and the results are currently pending. MRG-106 is a miRNA inhibitor that targets miR-155 that is currently being examined in cutaneous T-cell lymphoma and mycosis fungoides. Like mesomiR-1, it is also currently in phase I. It is thought to block the action of miR-155 from targeting tumor suppressors such as C/EBPβ and altering the TGF-β response (199). This study is currently still recruiting patients. These studies may offer promise of miRNA treatment as therapy and pave the way for future studies similar in nature.

Today, the main hurdle for miRNA-based therapies remains to be the method of delivery. Many types of viruses are thought to be potentially useful for treatment and many stabilizing modifications such as phosphorothioate, methyl- and fluoro-substitutions on RNA species may help to overcome this hurdle (200, 201). Given the diverse set of roles that miRNAs play in regular cellular function, it is evident that clear elucidation of specific miRNA mechanisms may be required before their integration into modern cancer therapy (202). In contrast, due to the dependence and overexpression of a few coding mRNA in tumorigenic cells, it is possible that miRNAs may have a higher therapeutic index. miRNAs may prove to be an important addition to treatment in the years to come to treat drug resistant cancers in the future.

MG contributed to the research, figure design, and writing of manuscript. LS contributed to the research, editing, and overall design of manuscript.

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.