*Review* **Emerging Link between Tsc1 and FNIP Co-Chaperones of Hsp90 and Cancer**

**Sarah J. Backe 1,2,†, Rebecca A. Sager 1,2,†, Katherine A. Meluni 1,2, Mark R. Woodford 1,2,3 , Dimitra Bourboulia 1,2,3 and Mehdi Mollapour 1,2,3,\***


**Abstract:** Heat shock protein-90 (Hsp90) is an ATP-dependent molecular chaperone that is tightly regulated by a group of proteins termed co-chaperones. This chaperone system is essential for the stabilization and activation of many key signaling proteins. Recent identification of the co-chaperones FNIP1, FNIP2, and Tsc1 has broadened the spectrum of Hsp90 regulators. These new co-chaperones mediate the stability of critical tumor suppressors FLCN and Tsc2 as well as the various classes of Hsp90 kinase and non-kinase clients. Many early observations of the roles of FNIP1, FNIP2, and Tsc1 suggested functions independent of FLCN and Tsc2 but have not been fully delineated. Given the broad cellular impact of Hsp90-dependent signaling, it is possible to explain the cellular activities of these new co-chaperones by their influence on Hsp90 function. Here, we review the literature on FNIP1, FNIP2, and Tsc1 as co-chaperones and discuss the potential downstream impact of this regulation on normal cellular function and in human diseases.

**Keywords:** tuberous sclerosis complex (TSC); Tsc1 (hamartin); Tsc2 (tuberin); heat shock protein 90 (Hsp90); FNIP1; FNIP2; co-chaperones; cancer; renal cell carcinoma; kidney cancer

### **1. Introduction**

Heat shock protein-90 (Hsp90) is a molecular chaperone essential for maintaining signaling competence in eukaryotic cells. Hsp90 is comprised of an N-terminal ATP binding domain, a middle domain for binding "client" proteins, and a site of constitutive dimerization at the carboxy-terminus [1–3]. Hsp90 function is coupled to its ability to bind and hydrolyze ATP and undergo a series of conformational changes known as the "chaperone cycle" [4,5]. This cycle facilitates the maturation and activation of more than 300-client proteins, including kinases, and non-kinases such as steroid hormone receptors, transcription factors, and tumor suppressors [6] (https://www.picard.ch/downloads/ Hsp90interactors.pdf, accessed on 12 February 2022). A number of these Hsp90 client proteins participate in oncogenesis, and this chaperone machine is often co-opted by cancers to maintain deregulated signaling pathways and buffer the effect of pathogenic mutations [7–11]. The breadth of signaling pathways mediated by its clients makes Hsp90 an attractive therapeutic target and dozens of Hsp90-directed small molecules have been developed. In fact, there are 14-ATP-competitive Hsp90 inhibitors in ongoing clinical trials in various cancers (www.clinicaltrials.gov, accessed on 1 May 2022) [12].

The binding and dissociation of Hsp90-modulating proteins, termed co-chaperones, tailors Hsp90 to particular clients and provides directionality to the chaperone cycle [13–16]. To date, more than 25 Hsp90 co-chaperones with varying characteristics and classifications

**Citation:** Backe, S.J.; Sager, R.A.; Meluni, K.A.; Woodford, M.R.; Bourboulia, D.; Mollapour, M. Emerging Link between Tsc1 and FNIP Co-Chaperones of Hsp90 and Cancer. *Biomolecules* **2022**, *12*, 928. https://doi.org/10.3390/ biom12070928

Academic Editor: Yongzhang Luo

Received: 1 June 2022 Accepted: 28 June 2022 Published: 1 July 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

have been identified. Prior to the recent characterization of the three large co-chaperones Tsc1, FNIP1, and FNIP2 (hereon referred to as FNIP1/2), known Hsp90 regulatory proteins existed within the range of 20–100 kDa [15]. These three large co-chaperones are each approximately 130 kDa [17–21] and were originally identified as stabilizers of specific tumor suppressor proteins implicated in the mTOR pathway [17–19]. The co-chaperone function of FNIP1/2 and Tsc1 gives us an opportunity to reevaluate the previous published work from a new perspective. Here we review the functions and roles of FNIP1/2, and Tsc1 that have been reported, describe their functions as new co-chaperones of Hsp90, and retrospectively evaluate how new functions can help contextualize previous observations. We also review their roles in cancer and cellular response to Hsp90 inhibitors as well as their emerging role in chaperoning of tumor suppressors.

#### **2. FNIP1 and FNIP2**

#### *2.1. FNIP1/2 Structure and Function*

Folliculin interacting proteins 1 and 2 (FNIP1/2) are named after their first identification in complex with the tumor suppressor folliculin (FLCN) [17,18]. Loss of FLCN function is implicated in Birt-Hogg-Dubé (BHD) syndrome, a hereditary condition characterized by benign fibrofolliculomas, pulmonary cysts, spontaneous pneumothorax, and renal tumors, which are most often of hybrid oncocytic or chromophobe histology [22]. FLCN interacts with FNIP1/2 via its C-terminus, which stabilizes the FLCN protein. This mechanism is supported by the instability of C-terminally truncated FLCN protein products resulting from FLCN mutations identified in BHD [17,18,22,23] and indeed, many of these mutated FLCN proteins fail to associate with FNIPs and are targeted for proteasomal degradation [24]. Recently, portions of the FLCN:FNIP2 structure have been resolved by cryo-EM [25,26]. The structures support previous evidence that FLCN contains a GTPase activating protein (GAP) domain and interacts with FNIP2 through its C-terminal differentially expressed in normal and neoplastic cells (DENN) domain. Additionally, the N-terminal Longin domains of FLCN and FNIP2 proteins also interact, emphasizing the complex nature of the interaction between FLCN and the FNIPs [25,26]. Despite this well-supported finding, the precise mechanism by which FLCN stability is achieved had remained elusive. Our group demonstrated that FLCN is a client of Hsp90 and depends on the co-chaperone activity of FNIP1 and FNIP2 for loading to Hsp90 and thus stability [20].

FNIP1 shares 74% similarity and 49% identity with FNIP2 [18], and the majority of research on FNIPs is exclusive to FNIP1. Initially, FNIP1 was found to be phosphorylated by AMP-activated protein kinase (AMPK) as well as facilitate AMPK-mediated phosphorylation of FLCN [17]. AMPK is a negative regulator of the mTOR nutrient-sensing pathway, and FNIP1 was found to translocate from the cytoplasm to lysosomes under starvation conditions [27], therefore a role for FNIP1 in mTOR signaling was suggested, though direct, mechanistic evidence remains elusive (Figure 1).

#### *2.2. FNIP1 Function in Skeletal Muscle and Adipocytes*

One pathway in which the FNIP1-AMPK interaction has been interrogated is skeletal muscle fiber type specification. Broadly, type I muscle fibers are highly aerobic, express elevated levels of myoglobin, and have high mitochondrial function, while type II muscle fibers are comparatively lower in both and favor anaerobic glycolysis [28]. AMPK is known to regulate mitochondrial biogenesis via peroxisome proliferator-activated receptorγ coactivator-1 α and β (PGC1α/PGC1β) and is activated under low energy conditions to suppress mTOR-dependent ATP utilization [29]. *FNIP1*−/<sup>−</sup> mice contain an abundance of type I muscle fibers, similar to mice with gain-of-function mutations in AMPK [30,31]. This suggests that at steady state FNIP1 suppresses AMPK and thus regulates mitochondrial biogenesis. Liu et al. furthered this line of inquiry by demonstrating that miR-499, an intron of the gene encoding the major slow-twitch type I myosin heavy chain (Myh7b), directly targets and inhibits translation of FNIP1 but not FNIP2 [32]. Similar results have been shown for miR-208b [33]. Interestingly, FNIP1-mediated AMPK inhibition can be

reversed by the flavonoid dihydromyricetin, which causes a decrease in FNIP1 expression and reactivates AMPK-mediated mitochondrial biogenesis [34]. These data provide a mechanism that explains the FNIP1-dependent regulation of AMPK in skeletal muscle. *Biomolecules* **2022**, *12*, x FOR PEER REVIEW 3 of 22

**Figure 1. FNIPs and Tsc1 in the mTOR pathway.** The mTOR pathway is a cellular signaling hub that integrates signals from several pathways and controls protein synthesis. A simplified schematic representation is shown to highlight the role of the FLCN/FNIPs and TSC complexes as mTOR regulators through GAP activity of RagA/C and Rheb, respectively. **Figure 1. FNIPs and Tsc1 in the mTOR pathway.** The mTOR pathway is a cellular signaling hub that integrates signals from several pathways and controls protein synthesis. A simplified schematic representation is shown to highlight the role of the FLCN/FNIPs and TSC complexes as mTOR regulators through GAP activity of RagA/C and Rheb, respectively.

*2.2. FNIP1 Function in Skeletal Muscle and Adipocytes*  One pathway in which the FNIP1-AMPK interaction has been interrogated is skeletal muscle fiber type specification. Broadly, type I muscle fibers are highly aerobic, express elevated levels of myoglobin, and have high mitochondrial function, while type II muscle fibers are comparatively lower in both and favor anaerobic glycolysis [28]. AMPK is known to regulate mitochondrial biogenesis via peroxisome proliferator-activated receptor-γ coactivator-1 α and β (PGC1α/PGC1β) and is activated under low energy conditions FNIP1 regulation of AMPK may be cell-type dependent however, as recent work has demonstrated that FNIP1 regulates cellular respiration in adipocytes in an AMPK/mTOR independent manner [35]. Specifically, FNIP1 was shown to regulate intracellular Ca2+ levels through stabilization of sarcoendoplasmic reticulum calcium transport ATPase (SERCA) and increasing SERCA Ca2+ pump activity. This study also suggested a pivotal role for FNIP1 in regulating metabolism and glucose homeostasis in adipocytes, independent of AMPK/mTOR [35].

#### to suppress mTOR-dependent ATP utilization [29]. *FNIP1*−/− mice contain an abundance of type I muscle fibers, similar to mice with gain-of-function mutations in AMPK [30,31]. *2.3. FNIP1 in Oxidative Stress*

This suggests that at steady state FNIP1 suppresses AMPK and thus regulates mitochondrial biogenesis. Liu et al. furthered this line of inquiry by demonstrating that miR-499, an intron of the gene encoding the major slow-twitch type I myosin heavy chain (Myh7b), directly targets and inhibits translation of FNIP1 but not FNIP2 [32]. Similar results have An interesting perspective on FNIP1 regulation of AMPK activity can be gained through an understanding of the factors governing FNIP1 protein dynamics. Recent work has shown that reductive stress promoted the degradation of FNIP1, but not FNIP2 [36]. The mechanism was traced to the chelation of Zn2+ by two reduced Cys residues in

been shown for miR-208b [33]. Interestingly, FNIP1-mediated AMPK inhibition can be reversed by the flavonoid dihydromyricetin, which causes a decrease in FNIP1 expression and reactivates AMPK-mediated mitochondrial biogenesis [34]. These data provide a FNIP1, which recruits CUL2FEM1B [36,37], the scaffold and recognition subunits of an E3-ubiquitin ligase complex [38]. Degradation of FNIP1 in this context promotes AMPK-PGC1α-mediated mitochondrial biogenesis to counteract reductive stress [36,37]. Interestingly, loss of FEM1B led to decreased lactate production [36], perhaps as a byproduct of FNIP1-dependent stabilization of FLCN and its recently described tumor suppressive effect on lactate dehydrogenase A [39].

#### *2.4. FNIP1 Function in B-Cell Development*

Another striking example of FNIP1 function is in lymphoid differentiation and maturation. Park et al. identified a pre-B cell "checkpoint" where loss of FNIP1 prevents mature B-cell development [40]. These cells were found to be sensitive to nutrient-deprivationinduced apoptosis seemingly due to failure of AMPK to suppress mTOR in the absence of FNIP1 [40]. Interestingly, FNIP1-deficient B-cell progenitors also exhibit elevated TFE3 transcription as well as increased lysosome function and autophagic flux [41]. Similarly, loss of FNIP1 prevents maturation of invariant natural killer T cells and increases their sensitivity to apoptosis [42]. *FNIP1* knockout was again determined to cause downstream mTOR activation, though in this case the effect is definitively indirect, as rapamycin treatment was not able to rescue the phenotype [42]. Concurrent research also found a marked pre-B cell blockade and confirmed that the effect stems from caspase activation and intrinsic apoptosis [43]. This effect was also observed in patients, as *FNIP1* mutation caused a clinically significant reduction in B cell numbers and hypogammaglobulinemia [44,45]. In addition to B-cell deficiency, FNIP1 loss leads to cardiomyopathy, which phenocopies AMPK gain-of-function mutations, consistent with a failure of FNIP1 to regulate AMPKmediated signaling [46]. Taken together, these data support a role for FNIP1 as an indirect regulator of mTOR through its suppression of AMPK activity, and likely also via its positive regulation of FLCN [47].

#### *2.5. Role of FNIP1/2 in Transcription*

Recent work has also demonstrated the impact of the FLCN-FNIP1/2 system on transcriptional reprogramming. It is well established that loss of FLCN induces nuclear localization of the transcription factors TFE3/TFEB and promotes a gene expression program favorable for tumor growth [41,48–51]. Similarly, it was recently shown that simultaneous deletion of *FNIP1/2* in a human renal proximal tubular epithelial cell (RPTEC) line induced a TFE3-mediated gene signature [52]. This is in agreement with previous data showing that *FLCN*-null and *FNIP1/2*-null mice developed phenotypically indistinguishable enlarged polycystic kidneys [53,54]. Additionally, loss of either *FLCN* or *FNIP1/2* induced a STAT2 dependent interferon response transcriptional program, though the impact of interferon signaling in *FLCN*-deficient tumors is unclear [52].

Despite the progress reviewed here, it remains difficult to disentangle the cellular roles of FNIP1/2 in the regulation of AMPK and TFE3 from that of FLCN tumor suppressive function. Given this, it is possible that FNIP1/2-mediated regulation of Hsp90 activity provides a unifying explanation for FNIP-mediated cellular effects.

#### **3. Tsc1**

#### *3.1. Structure and Function of the Tsc1/2 Complex*

Tuberous Sclerosis Complex (TSC) is an autosomal dominant genetic syndrome caused by mutations in either the *TSC1* or *TSC2* tumor suppressors. In addition to neural associations that include epilepsy, subependymal giant cell astrocytomas (SEGA), intellectual disability, and autism, TSC is also characterized by cutaneous, pulmonary, and renal manifestations, similar to BHD [23,55,56]. These include facial fibrofolliculomas, pulmonary lymphangiomyomatosis, and renal angiomyolipomas (AML). The *TSC2* gene was cloned first in 1993 followed by the non-homologous *TSC1* gene in 1997 [19,57]. The Tsc1 and Tsc2 proteins, also known as hamartin and tuberin, respectively, were then shown to directly interact and form a complex [58]. The 130 kDa Tsc1 and 200 kDa Tsc2 proteins share no

homology with each other [59]. Recently, partial structures of this complex were resolved by cryo-EM and revealed an elongated structure with a 2:2 stoichiometry. Further, Tsc1 was consistently found to have a coiled-coil domain which mediated Tsc1 dimerization and interaction with Tsc2 in vitro [60–62]. This is in contrast to a previous study using a yeast two-hybrid system which identified Tsc1 residues 302–430 as the critical region for Tsc2 interaction [63]. Tsc2 interaction with Tsc1 was primarily mediated through the N-terminal Tsc2-HEAT repeat domain, which is consistent with previous findings [60,62,63]. Furthermore, Tsc1 was required for Tsc2 maximal GAP activity likely through proper positioning of the Tsc2 catalytic asparagine-thumb [62].

The Tsc1/2 complex was demonstrated to inhibit mTOR signaling through the GAP activity of Tsc2 towards Rheb [64–66] (Figure 1). The effect of Tsc2 was greatly potentiated by the presence of Tsc1. In this TSC complex, Tsc1 has been shown to be important for the stabilization of Tsc2, preventing its interaction with the HERC1 ubiquitin ligase and its ubiquitination [67,68]. Early identification of Tsc1 and Tsc2 in complex and the role of this complex in the mTOR pathway focused a large portion of the TSC literature on this function and does not address separable functions of Tsc1 and Tsc2.

#### *3.2. Separable Functions of Tsc1 and Tsc2*

There are a number of differences in Tsc1 and Tsc2 function that have been identified, as well as mTOR-independent functions. Early reports suggested that although Tsc1 and Tsc2 often co-localize, the subcellular locations as well as tissue and organ expression patterns of Tsc1 and Tsc2 are not identical [69]. Germline mutations in *TSC1* cause a similar but not identical phenotype to *TSC2* mutations in animal models, suggesting commonalities to the pathways involved but some differences as well [70]. Renal tumors developed in heterozygous *TSC1* mice at a slower rate than in *TSC2*+/<sup>−</sup> mice. In addition to renal cystadenomas, *TSC1*+/<sup>−</sup> mice also develop liver hemangiomas, which are more common and more severe in female mice, demonstrating sex-dependent lethality [71]. Concordantly, an analysis of patients in the TOSCA database (TuberOus SClerosis registry to increase disease Awareness) revealed that female patients were significantly more likely to develop renal AML and experience hemorrhage [72]. Sex-dependent and estrogen linked effects exclusive to Tsc1 can also be seen in mammary development. Conditional Tsc1 loss in mammary epithelium impaired mammary development through suppression of Akt, ER, and cell cycle regulators and did not lead to tissue hyperplasia [73]. Moderate overexpression of Tsc1 also enhances overall health and cardiovascular health in an animal model and improves survival only in female mice [74]. Tsc1 and Tsc2 have also been shown to have separable functions in both cell signaling and cell cycle control [75,76]. Milolaza et al. describe the effect of Tsc1 and Tsc2 on the G1 to S phase transition of the cell cycle. Tsc1 and Tsc2 control cell proliferation independent of each other, and only Tsc2 function is affected by p27 expression [75]. Further evidence for separate functions of Tsc1 and Tsc2 comes from microarray and proteomic approaches, which reveal that each TSC gene triggers substantially different cellular responses [77–82].

#### *3.3. mTOR Independent Functions of Tsc1*

While the effects of Tsc1 loss are often ascribed to increased mTOR signaling and are at least partially responsive to rapamycin, there are also mTOR independent functions of Tsc1 that have been reported. Tsc1 haploinsufficiency without mTOR activation was shown to lead to renal cyst formation in *TSC1*+/<sup>−</sup> mice [83]. It has also recently been demonstrated that p21 activated kinase 2 (PAK2) is an effector of the Tsc1/Tsc2 complex. Loss of either Tsc1 or Tsc2 promotes hyperactivity of PAK2 downstream of Rheb, but independent of mTOR, as demonstrated by insensitivity to rapamycin treatment [84]. Tsc1 and Tsc2 also differentially modulate the cytoskeleton. *TSC1*−/<sup>−</sup> and *TSC2*−/<sup>−</sup> MEFs demonstrate rapamycin insensitive increase in number and length of cilia [85] whereas only Tsc2 loss promotes an mTOR-dependent pro-migratory phenotype [86]. On the other hand, Tsc1 loss was shown to dysregulate tight junction development in an mTOR independent

manner [87]. Collectively, these studies suggest a role for Tsc1 in cell integrity independent of mTOR.

Furthermore, it has long been observed that clinical features of TSC across multiple organ systems are more severe in patients with mutations in *TSC2* than in patients with *TSC1* mutations [88–90]. There is a higher incidence of intellectual disability in patients with *TSC2* mutations, and it has been suggested that severity of disability may correlate with predicted effects of mutations on Tsc1 and Tsc2 protein [91–93]. Epilepsy generally exhibits an earlier onset and is also more severe as a result of *TSC2* mutations [94,95]. Similarly, the mean age at diagnosis for patients with renal AML was lower in patients with *TSC2* mutations. Additionally, patients with *TSC2* mutations had a higher occurrence of renal AML, multiple renal cysts and polycystic kidney disease compared to patients with *TSC1* mutations [72]. In a mouse model, conditional knockout (CKO) of *TSC2* in GFAP-positive cells also produces a more severe epilepsy phenotype than *TSC1* CKO [96]. Additionally, it has been proposed that perhaps TSC resulting from *TSC1* mutation is not less common than *TSC2* disease but that it is less frequently diagnosed due to the milder clinical features [97].

Collectively, this evidence suggests a role for Tsc1 outside the TSC complex and mTOR signaling. Due to the described role of Tsc1 in stabilizing Tsc2 and protecting it from ubiquitination we questioned whether this protective role involved molecular chaperones and whether Tsc1 was involved in chaperoning Tsc2. In fact, Tsc2 is a client of Hsp90, and Tsc1 is a co-chaperone [21].

#### **4. Regulation of Hsp90 Chaperone Function by Co-Chaperones**

The action of co-chaperones towards Hsp90 generally meets one or more of the following criteria: (1) scaffolding of client proteins to Hsp90 (e.g., Hop, p50Cdc37); (2) modulation of Hsp90 ATPase activity (e.g., Aha1); (3) stabilization of specific chaperone complexes (e.g., p23) and are not themselves dependent on Hsp90 for stability [98]. We have shown that the newly identified large co-chaperones FNIP1/2 and Tsc1 are able to satisfy at least two of these observed co-chaperone functions (Figure 2). Indeed, we have a unique opportunity to advance our understanding of the function and effect of these proteins as we reconcile their known functions with their roles as Hsp90 co-chaperones.

Hsp90-dependent maturation and activation of client proteins relies on a continuum of regulated conformational changes of Hsp90 coupled to ATP hydrolysis. As currently understood, there are several "stages" to a generalized chaperone cycle. Initially, immature clients bind to the early chaperone heat shock protein 70 (Hsp70) and the Hsp70-Hsp90 organizing protein (Hop) forms a bridge to the "open" conformation of Hsp90, allowing the transfer of a client protein to Hsp90 [99]. ATP subsequently binds to the amino-terminal nucleotide binding pocket, and concurrent binding of the Activator of Hsp90 ATPase (Aha1) displaces Hop and induces transient N-domain dimerization, forming the "closed 1" state. Aha1 binds to the N-domain as well as the middle domain of Hsp90 and greatly increases the weak intrinsic ATPase activity of Hsp90 [100]. Interaction of the co-chaperone p23 with the N-domain of Hsp90 displaces Aha1 and stabilizes the "closed and twisted" conformation (closed 2). This allows completion of ATP hydrolysis, followed by release of a mature client protein and the return of Hsp90 to the open conformation [101–103].

The complement of co-chaperones that regulate Hsp90 during a single chaperone cycle is largely dependent on the individual requirements of the client protein [104]. For example, kinase clients are loaded to Hsp90 by the co-chaperone Cdc37, a decelerator of Hsp90 ATPase activity, and protein phosphatase 5 (PP5)-mediated dephosphorylation of Cdc37 is required for their release [105,106]. Alternatively, overexpression of Aha1 greatly decreases the folding of CFTR by accelerating the rate of Hsp90 ATP hydrolysis [107–109]. Similarly, steroid hormone receptors prefer a slower chaperone cycle and require the cochaperone p23, which is known to decelerate the action of Hsp90 [5,110–112]. In fact, GR itself is capable of modulating the conformation of Hsp90 such that Hsp90 ATPase activity

decreases [113], demonstrating the degree of specificity that can be achieved by modulation of Hsp90 complexes [15].

**Figure 2. The Hsp90 chaperone cycle.** Open Hsp90 is dimerized only through contacts in the CTD. ATP binding and an ordered series of conformational changes allow Hsp90 to adopt a closed conformation, which is N-terminally dimerized. ATP hydrolysis leads Hsp90 to return to the open conformation and begins another chaperone cycle. Throughout the chaperone cycle co-chaperones bind to Hsp90 and regulate its function. PTM of Hsp90 and PTM of co-chaperones provide further regulation of the chaperone cycle.

#### **5. FNIP1/2 and Tsc1: New Co-Chaperones of Hsp90**

Recent reports from Mollapour's group demonstrated that the tumor suppressors FLCN and Tsc2 are Hsp90 clients [20,21]. As FNIP1/2 and Tsc1, respectively, have established roles as guardians of these tumor suppressors [17,18,67,68], it follows that there may be a role for molecular chaperones in mediating FLCN and Tsc2 stability. Indeed FNIP1/2 and Tsc1 both interact with Hsp90 and Hsp70, as well as with overlapping complements of Hsp90 co-chaperones including PP5, Cdc37, Hop, and p23 and behave as Hsp90 co-chaperones [20,21] (Figure 3). These reports also demonstrate a role for these new co-chaperones in regulating both kinase and non-kinase clients, as well as provide clues to their chronology in the overall chaperone cycle.

1 FNIP1 and Tsc1 share a number of striking similarities in their actions as co-chaperones. Both FNIP1 and Tsc1 exhibit complex binding to Hsp90; contacts are made using multiple domains of these co-chaperones as well as multiple domains of Hsp90. The most well characterized interactions thus far however are that of FNIP1 and Tsc1 binding the Hsp90 middle domain via their carboxy-termini. It is through this interaction that they decelerate Hsp90 ATPase activity and compete with the accelerating co-chaperone Aha1 for Hsp90 occupancy. In addition to increasing the dwell time of ATP (and thus client proteins)

on Hsp90, interaction with these large co-chaperones also increases Hsp90 binding to its ATP-competitive inhibitors [20,21,114].

**Figure 3. FNIPs and Tsc1 co-chaperone interaction network.** Hsp90 co-chaperones are represented by colored circles. Interactions between co-chaperones are denoted by colored lines. FNIP1 interactions are colored red; FNIP2, yellow; Tsc1, blue; other, gray.

While the overall pattern of how FNIP1 and Tsc1 interact with Hsp90 is similar there are key differences between them. The C-terminal fragment of Tsc1 (Tsc1-D) binds to Hsp90 with higher affinity than does the C-terminal fragment of FNIP1 (FNIP1-D). Similarly, Tsc1-D is a potent inhibitor of Hsp90 ATPase activity and very effectively competes with Aha1 for Hsp90 binding as evidenced by in vitro competition experiments. FNIP1 and Tsc1 can also be distinguished by the complement of co-chaperones with which they interact therefore, providing clues to their distinct roles in the chaperone cycle. While neither is found in complex with Aha1, Tsc1 is able to interact with PP5 and Cdc37, whereas FNIP1/2 can additionally be found in complexes containing p23 and Hop (Figure 3). This may demonstrate some promiscuity of FNIPs, but likely reflects the necessity for FNIPs to work in concert with other co-chaperones, while Tsc1 may be capable of modulating Hsp90 independently. This potentially explains the observation that Tsc1 is a much more potent decelerator of Hsp90 ATP hydrolysis than FNIP1/2 [20,21,114]. Interestingly, Tsc1 also inhibits the ATPase activity of another molecular chaperone, Hsp70, in vitro [115]. Whether FNIPs share this function remains unknown.

Despite their shared role in facilitating chaperoning of both kinase and non-kinase clients, FNIP1/2 and Tsc1 over-expression and deletion have different effects. Non-kinase clients are destabilized upon knockdown/knockout of FNIP1/2 or Tsc1 and stabilized with overexpression of the co-chaperones. Interestingly, FNIP1/2 knockdown or overexpression affects the kinase clients in a comparable manner as the non-kinase clients, however overexpression or absence of Tsc1 both negatively affect kinase client stability [20,21]. This may be due to the participation of FNIP1/2 with a variety of chaperone complexes, whereas the semi-exclusive nature of Tsc1 co-chaperone activity disrupts the delicate balance of Hsp90 co-chaperone dynamics.

#### 1 *5.1. FNIP1/2 and Tsc1 in the Chaperone Cycle*

This large body of work on co-chaperone dynamics allows us to propose a model of FNIPs and Tsc1 co-chaperones in the Hsp90 chaperone cycle. Our previous work demonstrates that FNIPs and Tsc1 interact with Hsp70 in addition to Hsp90, and FNIP1 and Tsc1 are essential for scaffolding FLCN and Tsc2, respectively, to Hsp90 (Figure 4A,B) [20,21,23]. Subsequent ATP binding triggers conformational changes leading to the N-terminally dimerized 'closed' conformation of Hsp90 (Figure 4C) [116–119]. We have previously shown that Tsc1 and FNIP1 are not found in complex with Aha1 and that phosphorylation of Aha1-Y223 displaces Tsc1 from Hsp90 (Figure 4D) [20,21,109]. p23 is a late-acting co-chaperone that locks Hsp90 in the closed conformation to allow proper client maturation (Figure 4E) [103,120–127]. FNIP1/2, but not Tsc1, are found in complex with p23 (Figure 4F) [20,21]. We propose that p23:FNIP1:FNIP2 holds the matured client in its active conformation until there is a signal for client release, resetting Hsp90 for another cycle (Figure 4G,H). chaperone that locks Hsp90 in the closed conformation to allow proper client maturation (Figure 4E) [103,120–127]. FNIP1/2, but not Tsc1, are found in complex with p23 (Figure 4F) [20,21]. We propose that p23:FNIP1:FNIP2 holds the matured client in its active conformation until there is a signal for client release, resetting Hsp90 for another cycle (Figure 4G,H).

overexpression affects the kinase clients in a comparable manner as the non-kinase clients, however overexpression or absence of Tsc1 both negatively affect kinase client stability [20,21]. This may be due to the participation of FNIP1/2 with a variety of chaperone complexes, whereas the semi-exclusive nature of Tsc1 co-chaperone activity disrupts the del-

This large body of work on co-chaperone dynamics allows us to propose a model of FNIPs and Tsc1 co-chaperones in the Hsp90 chaperone cycle. Our previous work demonstrates that FNIPs and Tsc1 interact with Hsp70 in addition to Hsp90, and FNIP1 and Tsc1 are essential for scaffolding FLCN and Tsc2, respectively, to Hsp90 (Figure 4A,B) [20,21,23]. Subsequent ATP binding triggers conformational changes leading to the N-terminally dimerized 'closed' conformation of Hsp90 (Figure 4C) [116–119]. We have previously shown that Tsc1 and FNIP1 are not found in complex with Aha1 and that phosphorylation of Aha1-Y223 displaces Tsc1 from Hsp90 (Figure 4D) [20,21,109]. p23 is a late-acting co-

*Biomolecules* **2022**, *12*, x FOR PEER REVIEW 9 of 22

icate balance of Hsp90 co-chaperone dynamics.

*5.1. FNIP1/2 and Tsc1 in the Chaperone Cycle* 

**Figure 4***.* **FNIPs and Tsc1 in the Hsp90 chaperone cycle.** (**A**) FNIPs and Tsc1 co-chaperones scaffold a client from Hsp70 to Hsp90. (**B**) Hsp70 dissociates from the complex. (**C**) ATP binding triggers Hsp90 conformational rearrangements resulting in the 'closed' conformation. (**D**) Aha1 phosphorylated at Y223 displaces FNIPs/Tsc1 co-chaperones from the Hsp90 complex and promotes ATP hydrolysis to ADP + Pi. (**E**) p23 binds and stabilizes the closed conformation of Hsp90. (**F**) FNIP cochaperones bind to the Hsp90:client:p23 complex to promote client maturation. (**G**) The complex dissociates releasing the mature client. (**H**) Hsp90 is reset to begin another cycle. **Figure 4. FNIPs and Tsc1 in the Hsp90 chaperone cycle.** (**A**) FNIPs and Tsc1 co-chaperones scaffold a client from Hsp70 to Hsp90. (**B**) Hsp70 dissociates from the complex. (**C**) ATP binding triggers Hsp90 conformational rearrangements resulting in the 'closed' conformation. (**D**) Aha1 phosphorylated at Y223 displaces FNIPs/Tsc1 co-chaperones from the Hsp90 complex and promotes ATP hydrolysis to ADP + Pi. (**E**) p23 binds and stabilizes the closed conformation of Hsp90. (**F**) FNIP co-chaperones bind to the Hsp90:client:p23 complex to promote client maturation. (**G**) The complex dissociates releasing the mature client. (**H**) Hsp90 is reset to begin another cycle.

#### *5.2. FNIPs, Tsc1 and the Chaperone Code*

Hsp90 and its co-chaperones' functions are heavily regulated by post-translational modifications (PTM), collectively known as the 'chaperone code' [128,129]. An additional layer of Hsp90 regulation via FNIP1 is provided through FNIP1 post-translational modification. Recent work by our group identified a series of serine residues (S938, S939, S941, S946, and S948) in the Hsp90-binding region of the FNIP1 carboxy-terminus that are phosphorylated in a relay manner by casein-kinase 2 (CK2) [114]. This sequential phosphorylation promotes FNIP1 interaction with Hsp90 while dephosphorylation of these residues by the Hsp90 co-chaperone PP5 disrupts the Hsp90-FNIP1 complex. Furthermore, stepwise phosphorylation of FNIP1 provides gradual inhibition of Hsp90 ATPase activity and therefore increased activity of a subset of both kinase and non-kinase clients [114].

These new co-chaperones also affect Hsp90 binding to its ATP-competitive inhibitors. Generally, there is an inverse relationship between the rate of ATP hydrolysis and the ability of Hsp90 to bind ATP-competitive inhibitors [20,21,109]. Overexpression of FNIP1/2 or

Tsc1 decreases Hsp90 ATPase activity, thus increasing Hsp90 binding to its inhibitors. As expected, Hsp90 inhibitor binding is decreased upon knockdown of FNIP1/2 or loss of Tsc1 [20,21,114,130]. Interestingly, approximately 15% of bladder cancers have loss-offunction mutations in Tsc1. Tsc1 loss causes hypo-acetylation of Hsp90 on K407 and K419 leading to decreased binding of Hsp90 to its inhibitors, demonstrating another mechanism of Tsc1-mediated regulation of Hsp90 [130]. The precise mechanism of how Tsc1 loss compromises Hsp90 acetylation remains unknown, however it is important to note that Hsp90 acetylation can be restored by histone-deacetylase (HDAC) inhibition, sensitizing *TSC1*-null cells to Hsp90 inhibitors [130].

Targeting Hsp90 in cancer cell lines induces apoptosis, and Hsp90 inhibitors have been found to preferentially accumulate in cancer cells versus normal cells [131–135]. FNIP1/2 were found overexpressed in cancer cell lines originating from several different tissues, and knockdown decreased sensitivity of these cancers to Hsp90 inhibition [20]. This increased expression of FNIP1/2 provides one potential mechanism to explain the tumor selectivity of Hsp90 inhibitors. Similarly, bladder cancer cells lacking functional Tsc1 fail to accumulate Hsp90 inhibitors and are less sensitive to Hsp90 inhibition than those with wild-type Tsc1 [130].

Collectively, these studies provide support for a new functional role for the tumor suppressor Tsc1 and FNIP1/2 as co-chaperones of Hsp90. As Hsp90 co-chaperones the protective function of Tsc1 and the FNIPs goes beyond mediating stability of Tsc2 and FLCN, respectively, and provides insight into a larger role for these proteins in the cellular context.

#### **6. A New Perspective: FNIPs, Tsc1, and mTOR**

Early connection of FNIP1/2 and Tsc1 to the mTOR nutrient-sensing pathway has narrowed the focus of research conducted on these proteins. Recent research has demonstrated that FNIP1/2 and Tsc1 act as co-chaperones of Hsp90. This allows us to reevaluate the previous published work with a new perspective.

#### *6.1. FNIP1/2 Co-Chaperone Activity Contributes to mTOR Regulation*

As reviewed in this text, FNIP1 negatively regulates AMPK activity. Since the α and γ subunits of AMPK are known clients of Hsp90 [136], the effect of FNIP1 on AMPK could be mediated through the Hsp90 chaperone (Figure 5). In support of this idea, microarray data show that B220+CD43<sup>+</sup> pre-B cells from *FNIP1*−/<sup>−</sup> mice demonstrate a dramatic increase in expression of AMPK-responsive genes [40]. These data would suggest a role for FNIP1 in activation of mTOR, however we posit that this mechanism may actually be more complicated. First, mTOR is also an Hsp90 client protein [137] and will be subject to the influence of Hsp90 co-chaperones as with any other client protein. Second, it is likely that loss of FLCN is actually responsible for mTOR activation, as is suggested by Baba et al., whose data show that the induction of mTOR is mild in *FNIP1*−/<sup>−</sup> mice as compared to *FLCN*−/<sup>−</sup> and that *FNIP1* deletion fails to phenocopy BHD syndrome [17,43,54]. Concordantly, FNIP co-chaperone activity toward FLCN can explain the observation that non-degradable FNIP2 enhances FLCN expression and thus suppresses tumorigenesis in a BHD mouse xenograft model [138]. Together, these data highlight that the co-chaperone activity of FNIP1/2 is essential for FLCN-mediated mTOR suppression, but also underscore our inability to reconcile this observation with the current understanding of FNIP1/2 function.

#### *6.2. Co-Chaperone Activity of Tsc1 in Regulation of mTOR*

The newly identified role for Tsc1 as an Hsp90 co-chaperone may help clarify some of the phenotypic differences as a result of Tsc1 versus Tsc2 mutation. Due to its function as a co-chaperone, Tsc1 loss would trigger effects on many cellular pathways, not just mTOR signaling. This could explain the finding of renal cyst formation in *TSC1*+/<sup>−</sup> mice, as well as provide insight into the pro-migratory phenotype seen only in *TSC2*−/<sup>−</sup> MEFs [83,86]. Furthermore, Tsc1 loss, but not Tsc2 loss, causes hypo-acetylation of Hsp90 further demon-

strating a role for Tsc1 independent of both Tsc2 and mTOR [130]. The loss of Tsc1 has a dramatic negative effect on Hsp90 kinase and non-kinase clients, including Tsc2. It is reasonable therefore that loss of Tsc1 co-chaperone activity manifests itself independently of the mTOR pathway. ble FNIP2 enhances FLCN expression and thus suppresses tumorigenesis in a BHD mouse xenograft model [138]. Together, these data highlight that the co-chaperone activity of FNIP1/2 is essential for FLCN-mediated mTOR suppression, but also underscore our inability to reconcile this observation with the current understanding of FNIP1/2 function.

and that *FNIP1* deletion fails to phenocopy BHD syndrome [17,43,54]. Concordantly, FNIP co-chaperone activity toward FLCN can explain the observation that non-degrada-

*Biomolecules* **2022**, *12*, x FOR PEER REVIEW 11 of 22

**Figure 5***.* **Hsp90 clients in the mTOR pathway.** A schematic representation of the mTOR pathway highlighting the components that are Hsp90 clients (red). **Figure 5. Hsp90 clients in the mTOR pathway.** A schematic representation of the mTOR pathway highlighting the components that are Hsp90 clients (red).

*6.2. Co-Chaperone Activity of Tsc1 in Regulation of mTOR*  The newly identified role for Tsc1 as an Hsp90 co-chaperone may help clarify some of the phenotypic differences as a result of Tsc1 versus Tsc2 mutation. Due to its function as a co-chaperone, Tsc1 loss would trigger effects on many cellular pathways, not just mTOR signaling. This could explain the finding of renal cyst formation in *TSC1*+/− mice, as well as provide insight into the pro-migratory phenotype seen only in *TSC2*−/− MEFs [83,86]. Furthermore, Tsc1 loss, but not Tsc2 loss, causes hypo-acetylation of Hsp90 further As discussed above, Tsc1 loss has long been known to lead to a less severe phenotype than Tsc2 loss both in patients as well as animal models [88–90,94–97]. Canonically, Tsc2 loss leads to upregulation of mTOR signaling due to release of the inhibitory signal from Tsc2-Rheb [139]. Activation and phosphorylation of mTOR and its downstream targets as well as other pathway components such as Akt has been shown to be dependent on Hsp90; in fact, many mTOR pathway components are clients of Hsp90 [140–143] (Figure 5). Activation of the mTOR pathway is therefore dependent on proper function of the Hsp90 chaperone system. Perhaps the milder phenotype seen with Tsc1 loss is a result of the loss of co-chaperone activity toward Hsp90. Upon Tsc1 loss, Tsc2 is destabilized leading to

increased mTOR activity; however, the other proteins in the mTOR pathway that are Hsp90 clients would also be destabilized, potentially mitigating the downstream effect.

Rapalogs, such as sirolimus and everolimus, are rapamycin derivatives that are commonly used to treat patients with BHD and TSC. Armed with the information of the role of Hsp90 in BHD and TSC, perhaps preclinical examination of Hsp90 inhibitors in combination with rapalogs is warranted. In fact, Hsp90 inhibitors have been shown to synergize with PI3K/Akt/mTOR inhibitors in preclinical studies for the treatment of various cancers [144–146]. In line with this, Di Nardo et al. identified the heat-shock machinery as an exploitable target in Tsc2-deficient neurons [147]. Accordingly, one study has evaluated mTOR and Hsp90 inhibitors in combination in *TSC1* or *TSC2* deficient cancer models. Unfortunately, the results were inconsistent between cell line and mouse xenograft models, as synergism between Hsp90 inhibitor (GB) and mTOR inhibitors (Torin2, rapamycin) in cell lines did not translate to increased efficacy over monotherapy in xenograft models [148]. Taken together, these studies demonstrate the potential therapeutic benefit of co-targeting Hsp90 and mTOR in BHD and TSC patients. However, further investigation is needed.

#### **7. Specialized Function of FNIP1/2 and Tsc1: Chaperoning Tumor Suppressors**

An important and perhaps specialized role for these new Hsp90 co-chaperones is in the chaperoning of tumor suppressors. FLCN and Tsc2 are additions to the growing list of tumor suppressors that interact with Hsp90. The transcription factor p53 was the first reported tumor suppressor client of Hsp90 [149–152]. Since 1996, 17 functional Hsp90 interactions with both kinase and non-kinase tumor suppressors have been discovered (Table 1) [20,21,149,153–163]. Furthermore, several tumor suppressors including, VHL, BDC2, LKB1, p53, FLCN and LATS1/2 were found to interact with Hsp90 co-chaperones including, Hop, p23, Hsp110, Cdc37, PP5, and CHIP [20,23,150,154,159,160,164–167]. As FNIP1/2 and Tsc1 scaffold the tumor suppressors FLCN and Tsc2 to Hsp90, it follows that these co-chaperones may participate in the chaperoning of additional Hsp90-dependent tumor suppressor clients. In line with this notion, Tsc1 was identified as a genetic interactor with VHL in HeLa cells highlighting the need for further exploration in this area [168].


**Table 1.** Relationships of known tumor suppressor-Hsp90 interactions.

Additionally, there is new evidence supporting a compensatory mechanism between FNIP1/2 and Tsc1 co-chaperone activity. FLCN traditionally requires interaction via its C-terminus to FNIP1/2 to mediate its stability. Tsc1, however, is capable of interacting with a truncated FLCN mutant and supporting a low level of expression in the absence of FNIP1/2 binding [23]. Unexpectedly, the truncated FLCN-L460QsX25 was still able to interact with Hsp90 even though it did not bind to its loading co-chaperone FNIP1. Overexpression of Tsc1, but not FNIP1 was capable of stabilizing expression of the mutant FLCN. Notably, Tsc1 interaction with Tsc2 was compromised in this model resulting in loss of Tsc2 tumor suppressive function. Loss of such a compensatory mechanism may also explain why deletion of *FNIP1* synergized with *TSC1* deletion to activate mTOR and subsequently resulted in accelerated renal cyst formation in mice [189]. These findings necessitate investigation into how these large co-chaperones mediate chaperoning of tumor suppressors and the impact of tumor suppressor mutations on this relationship.

#### **8. Conclusions**

Hsp90 is an important component of the cellular homeostatic machinery and is regulated by post-translational modification and interaction with co-chaperones. There are more than 25 known co-chaperones that serve several functions including modulating Hsp90 conformations, loading client proteins to Hsp90, and modifying the rate of ATP hydrolysis. New data has identified newly characterized roles for three proteins, FNIP1, FNIP2 and Tsc1, as large co-chaperones of Hsp90. Though these proteins have established roles in the regulation of tumor suppressor proteins FLCN and Tsc2, it seems that many of their other ascribed functions are potentially explained at least in part by their effect on Hsp90. A more thorough understanding of the action and interplay of these new, large co-chaperones may unveil clues that will aid in developing the next generation of cancer therapeutics.

**Author Contributions:** Conceptualization, M.M.; writing—original draft preparation, S.J.B., R.A.S. and M.R.W.; writing—review and editing, S.J.B., R.A.S., M.R.W., D.B. and M.M.; visualization, S.J.B. and K.A.M.; supervision, M.M.; funding acquisition, M.M. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was partly supported with funds from National Institute of General Medical Sciences of the National Institutes of Health under Award Number R35GM139584 (M.M.). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. This work was also supported by Carol M. Baldwin Breast Cancer Fund (M.M.), SUNY Upstate Medical University, Upstate Foundation (M.M.).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Acknowledgments:** We are grateful to W. Marston Linehan & Len Neckers (Urologic Oncology Branch, NCI, USA), David J Kwiatkowski (Harvard Medical School, USA) and Chris Prodromou (University of Sussex, UK) for their valuable discussion.

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

#### **References**


## *Review* **TRAP1 Chaperones the Metabolic Switch in Cancer**

**Laura A. Wengert 1,2, Sarah J. Backe 1,2, Dimitra Bourboulia 1,2,3, Mehdi Mollapour 1,2,3 and Mark R. Woodford 1,2,3,\***


**Abstract:** Mitochondrial function is dependent on molecular chaperones, primarily due to their necessity in the formation of respiratory complexes and clearance of misfolded proteins. Heat shock proteins (Hsps) are a subset of molecular chaperones that function in all subcellular compartments, both constitutively and in response to stress. The Hsp90 chaperone TNF-receptor-associated protein-1 (TRAP1) is primarily localized to the mitochondria and controls both cellular metabolic reprogramming and mitochondrial apoptosis. TRAP1 upregulation facilitates the growth and progression of many cancers by promoting glycolytic metabolism and antagonizing the mitochondrial permeability transition that precedes multiple cell death pathways. TRAP1 attenuation induces apoptosis in cellular models of cancer, identifying TRAP1 as a potential therapeutic target in cancer. Similar to cytosolic Hsp90 proteins, TRAP1 is also subject to post-translational modifications (PTM) that regulate its function and mediate its impact on downstream effectors, or 'clients'. However, few effectors have been identified to date. Here, we will discuss the consequence of TRAP1 deregulation in cancer and the impact of post-translational modification on the known functions of TRAP1.

**Keywords:** TRAP1; Hsp90; chaperone; post-translational modification; cancer; mitochondria; metabolism; Warburg effect

#### **1. Introduction**

Molecular chaperones of the heat shock protein-90 (Hsp90) family are involved in signal integration and the cellular stress response. These chaperones mediate cell signaling through the stabilization and activation of their substrate proteins, known as clients (https: //www.picard.ch/downloads/Hsp90interactors.pdf, accessed 28 February 2022) [1]. The Hsp90 chaperone function is coupled to the ability to hydrolyze ATP, and chaperone activity can be precisely regulated by a heterogeneous group of proteins known as co-chaperones [2], as well as a diverse array of post-translational modifications (PTM) [3].

TNF-receptor-associated protein-1 (TRAP1) is the mitochondrial-dedicated Hsp90 family member and is localized to the mitochondrial matrix, inner mitochondrial membrane, and the intermembrane space [4–6]. TRAP1 was first identified through its interaction with the intracellular domain of the Type I TNF receptor [7], and early characterization of TRAP1 demonstrated ATP-binding ability and sensitivity to ATP-competitive Hsp90 inhibitors [8]. Despite this, TRAP1 was unable to form complexes with known cytosolic Hsp90 co-chaperones, nor could it promote the maturation of Hsp90 client proteins, suggesting a distinct mechanism of action for TRAP1 [8].

From this time, work has concentrated on the impact of TRAP1 on cellular processes, however identification of TRAP1 effectors and regulatory mechanisms of TRAP1 expression and activity are critical to understanding its biological function. TRAP1 has an established

**Citation:** Wengert, L.A.; Backe, S.J.; Bourboulia, D.; Mollapour, M.; Woodford, M.R. TRAP1 Chaperones the Metabolic Switch in Cancer. *Biomolecules* **2022**, *11*, 274. https:// doi.org/10.3390/biom12060786

Academic Editor: Raquel Soares

Received: 18 May 2022 Accepted: 2 June 2022 Published: 4 June 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

role as a master regulator of metabolic flux, and a large body of evidence has demonstrated that TRAP1 expression serves to suppress oxidative phosphorylation [9–11]. Further, TRAP1 also contributes to cell survival through complex formation with cyclophilin D (CypD), which regulates the opening of the permeability transition pore (PTP) [12]. These two known roles suggest a critical function for TRAP1 in maintaining cellular homeostasis [13]. Despite the critical importance of TRAP1 to these processes, the molecular mechanisms of TRAP1 function remain largely unresolved. Here, we will discuss recent advances in understanding the mechanisms of TRAP1 regulation, the impact of this regulation on TRAP1 function and downstream cellular processes, and the role of TRAP1 in cancer.

#### **2. Structural Basis of TRAP1 Activity**

Hsp90 family chaperones are characterized by their dimeric structure. Each of the two protomers are composed of an amino-terminal ATP-binding domain, followed by a middle domain, the primary interface for client interaction, and a C-terminal domain that allows constitutive dimerization of the protomers [14]. Hsp90 chaperone activity is coupled to its ability to hydrolyze ATP [15,16]. The 'chaperone cycle' begins with ATP binding to the 'open' conformation of Hsp90, followed by transient dimerization of the N-terminal domains of each protomer and ATP hydrolysis, and subsequent release of mature client proteins and regeneration of the 'open' Hsp90 dimer [17]. TRAP1 is broadly structurally similar to cytosolic Hsp90, with some notable exceptions, including a cleavable N-terminal mitochondrial localization signal and an N-terminal extension or 'strap' that stabilizes the 'closed' conformation of TRAP1 [18,19]. Asymmetrical post-translational modification and co-chaperone binding are important determinants of Hsp90 molecular chaperone function [18,20–24]. Interestingly, TRAP1 dimers are inherently asymmetric, and uniquely composed of one 'straight' and one 'buckled' protomer, with the buckled protomer demonstrating increased rates of ATP hydrolysis [25] (Figure 1). Recently, structural and cell-based studies have described a tetrameric form of TRAP1 induced in response to dysregulation of oxidative metabolism, although the impact of this TRAP1 state on its activity is as yet unknown [26]. Interestingly, whether TRAP1 ATPase activity is essential for the entire scope of its biological role also remains an open question [26]. *Biomolecules* **2022**, *12*, x FOR PEER REVIEW 3 of 21

**Figure 1.** Structures of human TRAP1 (PDB: 6xg6) and human Hsp90β (PDB: 5fwp) bound to nucleotide with the conserved N-, middle-, and C-domains denoted. One protomer of each is colored **Figure 1.** Structures of human TRAP1 (PDB: 6xg6) and human Hsp90β (PDB: 5fwp) bound to nucleotide

Controversially, TRAP1 has alternately been characterized as an oncogene and tumor suppressor, and it has been suggested that TRAP1 is essential for malignant transformation of cells but dispensable at later stages of tumor development [6,27]. Despite this controversy, much of the literature supports the idea that TRAP1 regulates metabolic transformation during tumorigenesis, TRAP1 is overexpressed in many cancers, and TRAP1 attenuation is detrimental to tumor cell survival [28–33]. It may be more appropriate to suggest that, similar to cytosolic Hsp90, many cancers may be 'addicted' to TRAP1 [34–36]. In fact, multiple pathways in which TRAP1 activity can drive tumorigen-

blue and the second is colored green. The regulatory N-terminal extension (strap) of each TRAP1 protomer can be observed overlapping the opposite protomer. The region of TRAP1 near the M-C boundary that 'buckles' during conformational rearrangement is incompletely resolved in the struc-

esis have been described () and will be reviewed in the following section.

**3. Impact of TRAP1 on Cancer Metabolism** 

with the conserved N-, middle-, and C-domains denoted. One protomer of each is colored blue and the second is colored green. The regulatory N-terminal extension (strap) of each TRAP1 protomer can be observed overlapping the opposite protomer. The region of TRAP1 near the M-C boundary that 'buckles' during conformational rearrangement is incompletely resolved in the structure. Additionally, the resolved residues of the charged linker domain (CL) of cytosolic Hsp90, which is absent in TRAP1, are labeled in the lower right quadrant.

#### **3. Impact of TRAP1 on Cancer Metabolism**

Controversially, TRAP1 has alternately been characterized as an oncogene and tumor suppressor, and it has been suggested that TRAP1 is essential for malignant transformation of cells but dispensable at later stages of tumor development [6,27]. Despite this controversy, much of the literature supports the idea that TRAP1 regulates metabolic transformation during tumorigenesis, TRAP1 is overexpressed in many cancers, and TRAP1 attenuation is detrimental to tumor cell survival [28–33]. It may be more appropriate to suggest that, similar to cytosolic Hsp90, many cancers may be 'addicted' to TRAP1 [34–36]. In fact, multiple pathways in which TRAP1 activity can drive tumorigenesis have been described (Figure 2) and will be reviewed in the following section. *Biomolecules* **2022**, *12*, x FOR PEER REVIEW 4 of 21

**Figure 2.** Role of human TRAP1 in mitochondria of normal cells and cancer cells. Normal expression levels (light blue) lead to TRAP1 regulation of ROS and calcium levels, integrity of cristae, function of ETC, and oversight of the PTP. As TRAP1 expression increases (dark blue), mitochondria lose calcium sensitivity, downregulate ROS, and prevent PTP opening, leading to metabolic reprogramming and evasion of apoptosis in cancer. **Figure 2.** Role of human TRAP1 in mitochondria of normal cells and cancer cells. Normal expression levels (light blue) lead to TRAP1 regulation of ROS and calcium levels, integrity of cristae, function of ETC, and oversight of the PTP. As TRAP1 expression increases (dark blue), mitochondria lose calcium sensitivity, downregulate ROS, and prevent PTP opening, leading to metabolic reprogramming and evasion of apoptosis in cancer.

#### *3.1. Metabolic Regulation 3.1. Metabolic Regulation*

remain obscured.

The cellular energy currency adenosine triphosphate (ATP) is generated as a consequence of the complete oxidation of glucose to CO2 and H2O, and each molecule of glucose can maximally result in 36–38 ATP molecules [37]. Normal cells produce ATP primarily through cellular respiration, which describes a process in which glucose metabolism by glycolysis is coupled to the tricarboxylic acid cycle (TCA). Concurrent mitochondrial electron transport generates the electrochemical gradient that provides the force by which ATP is disseminated throughout the cell [38]. ATP generation is highly dysregulated in cancers, and many cancer subtypes supplement their ATP supply by upregulating cytosolic glycolysis, simultaneously generating additional ATP driven by the terminal fermentation of pyruvate to lactate [39]. This hyperactive glycolytic phenotype is known as the The cellular energy currency adenosine triphosphate (ATP) is generated as a consequence of the complete oxidation of glucose to CO<sup>2</sup> and H2O, and each molecule of glucose can maximally result in 36–38 ATP molecules [37]. Normal cells produce ATP primarily through cellular respiration, which describes a process in which glucose metabolism by glycolysis is coupled to the tricarboxylic acid cycle (TCA). Concurrent mitochondrial electron transport generates the electrochemical gradient that provides the force by which ATP is disseminated throughout the cell [38]. ATP generation is highly dysregulated in cancers, and many cancer subtypes supplement their ATP supply by upregulating cytosolic glycolysis, simultaneously generating additional ATP driven by the terminal fermenta-

Warburg effect, and serves to support the accelerated growth of cancers through the increased synthesis of intermediates for anaplerotic metabolism and hypertrophy [40,41]. The phenotypic manifestations of metabolic dysregulation are variable and dependent on

Few specific biological roles and binding partners have been described for TRAP1, despite the broad understanding of its impact on metabolic flux. Two of the few described bona fide clients of TRAP1 however are subunits of electron transport chain (ETC) complexes, Complex II components succinate dehydrogenase subunit A/B (SDHA/B) [42–45], and Complex IV cytochrome *c* oxidase subunit 2 (COXII) [6,46,47]. Complex II/SDH is an iron–sulfur cluster-containing protein complex that functions to transfer electrons from succinate to coenzyme Q10-ubiquinone (Complex III) [48]. In agreement with the understanding of Hsp90 function, TRAP1 maintains SDH in a partially unfolded state [49], and TRAP1 inhibition releases active SDH, leading to an increase in its activity [27,44,50–52].

tion of pyruvate to lactate [39]. This hyperactive glycolytic phenotype is known as the Warburg effect, and serves to support the accelerated growth of cancers through the increased synthesis of intermediates for anaplerotic metabolism and hypertrophy [40,41]. The phenotypic manifestations of metabolic dysregulation are variable and dependent on cell type and genotype, and many of the details and nuances of this differential regulation remain obscured.

Few specific biological roles and binding partners have been described for TRAP1, despite the broad understanding of its impact on metabolic flux. Two of the few described bona fide clients of TRAP1 however are subunits of electron transport chain (ETC) complexes, Complex II components succinate dehydrogenase subunit A/B (SDHA/B) [42–45], and Complex IV cytochrome *c* oxidase subunit 2 (COXII) [6,46,47]. Complex II/SDH is an iron–sulfur cluster-containing protein complex that functions to transfer electrons from succinate to coenzyme Q10-ubiquinone (Complex III) [48]. In agreement with the understanding of Hsp90 function, TRAP1 maintains SDH in a partially unfolded state [49], and TRAP1 inhibition releases active SDH, leading to an increase in its activity [27,44,50–52]. Further, SDH activity [44,53,54] and the oxygen consumption rate [6,55] are inversely correlated with TRAP1 expression, implicating TRAP1 in promoting the Warburg effect [56]. Notably, SDH also oxidizes succinate to fumarate and thus integrates the TCA cycle and the ETC, indicative of the broad influence of TRAP1 on mitochondrial metabolism [56–58]. *Biomolecules* **2022**, *12*, x FOR PEER REVIEW 5 of 21 Further, SDH activity [44,53,54] and the oxygen consumption rate [6,55] are inversely correlated with TRAP1 expression, implicating TRAP1 in promoting the Warburg effect [56]. Notably, SDH also oxidizes succinate to fumarate and thus integrates the TCA cycle and the ETC, indicative of the broad influence of TRAP1 on mitochondrial metabolism [56–

> Complex IV of the ETC converts molecular oxygen to water, and in doing so enacts the final step in generating the electrochemical gradient that supports ATP production by Complex V (ATP synthase) [59]. COXII is a downstream effector of TRAP1 function in the regulation of apoptosis, and TRAP1 regulates COXII expression [47] and activity [6]. As downregulation or inhibition of TRAP1 has been shown to destabilize COXII [46,50] and deletion of TRAP1 was associated with decreased COXIV subunit levels [60], it is possible that TRAP1 chaperoning of COXII/IV is mechanistically similar to SDHA/B. TRAP1 has also been shown to interact with the Complex V subunit ATPB, although little is known about this interaction [27]. 58]. Complex IV of the ETC converts molecular oxygen to water, and in doing so enacts the final step in generating the electrochemical gradient that supports ATP production by Complex V (ATP synthase) [59]. COXII is a downstream effector of TRAP1 function in the regulation of apoptosis, and TRAP1 regulates COXII expression [47] and activity [6]. As downregulation or inhibition of TRAP1 has been shown to destabilize COXII [46,50] and deletion of TRAP1 was associated with decreased COXIV subunit levels [60], it is possible that TRAP1 chaperoning of COXII/IV is mechanistically similar to SDHA/B. TRAP1 has also been shown to interact with the Complex V subunit ATPB, although little is known about this interaction [27].

> Mitochondrial respiration drives the production of reactive oxygen species (ROS) and is responsible for most cellular ROS (Figure 3) [61]. In considering the role of TRAP1 in chaperoning SDH and COXII, TRAP1-mediated regulation of mitochondrial respiration suppresses ROS production [62], thereby contributing to the regulation of redox homeostasis, metabolic flux, and mitochondrial apoptosis. Mitochondrial respiration drives the production of reactive oxygen species (ROS) and is responsible for most cellular ROS () [61]. In considering the role of TRAP1 in chaperoning SDH and COXII, TRAP1-mediated regulation of mitochondrial respiration suppresses ROS production [62], thereby contributing to the regulation of redox homeostasis, metabolic flux, and mitochondrial apoptosis.

**Figure 3.** Simplified mitochondrial respiration schematic. Electron transport chain (ETC) complexes (I–V) are represented by orange ovals, and reactive oxygen species (ROS) generated as a byproduct of Complex I and III activity is represented by yellow starbursts. Succinate dehydrogenase (SDH)/Complex II connects the ETC to the tricarboxylic acid (TCA) cycle. TRAP1 interactors involved in this process have been highlighted in green. **Figure 3.** Simplified mitochondrial respiration schematic. Electron transport chain (ETC) complexes (I–V) are represented by orange ovals, and reactive oxygen species (ROS) generated as a byproduct of Complex I and III activity is represented by yellow starbursts. Succinate dehydrogenase (SDH)/Complex II connects the ETC to the tricarboxylic acid (TCA) cycle. TRAP1 interactors involved in this process have been highlighted in green.

Cancer-associated increases in TRAP1 expression suggest a role for TRAP1 in oncogenesis [30,63,64]. Indeed, TRAP1 deletion delayed tumor formation in a mouse model of

cinate [58]. Increased succinate inhibits the activity of prolyl hydroxylases, which are responsible for the hydroxylation of the transcription factor hypoxia inducible factor (HIF1α), a prerequisite for recognition by the VHL-dependent E3-ubiquitin ligase machinery [66]. Succinate-dependent HIF1α stabilization and activation promotes a well-established glycolytic transcriptional program [67], demonstrating yet another function of

TRAP1 in the regulation of cancer-associated metabolic dysregulation.

*3.2. Contribution to Tumorigenesis* 

#### *3.2. Contribution to Tumorigenesis*

Cancer-associated increases in TRAP1 expression suggest a role for TRAP1 in oncogenesis [30,63,64]. Indeed, TRAP1 deletion delayed tumor formation in a mouse model of breast cancer, providing direct evidence of the role of TRAP1 in tumor initiation [65]. Further, TRAP1-mediated SDH inhibition leads to accumulation of the oncometabolite succinate [58]. Increased succinate inhibits the activity of prolyl hydroxylases, which are responsible for the hydroxylation of the transcription factor hypoxia inducible factor (HIF1α), a prerequisite for recognition by the VHL-dependent E3-ubiquitin ligase machinery [66]. Succinate-dependent HIF1α stabilization and activation promotes a well-established glycolytic transcriptional program [67], demonstrating yet another function of TRAP1 in the regulation of cancer-associated metabolic dysregulation.

TRAP1 expression was found to be elevated in aggressive pre-neoplastic lesions in a rat model of hepatocarcinogenesis [68]. The master antioxidant transcription factor NRF2 was also activated in this model, and given the established role of TRAP1 in regulating intracellular ROS, TRAP1 likely participates in NRF2-driven ROS mitigation during tumor development [68]. NRF2 inhibition led to decreased TRAP1 levels independent of TRAP1 transcription [68], suggesting that post-translational regulation is essential for sustained TRAP1 expression in pre-cancerous and cancerous cells. Interestingly, pentose phosphate pathway (PPP) flux was found to be increased in this model, and was determined to be a consequence of elevated citrate synthase activity in aggressive pre-neoplastic lesions [68]. Citrate accumulation inhibits downstream metabolic enzymes phosphofructokinase and SDH and activates the anaplerotic PPP [69]. This increase in citrate synthase activity was alleviated following TRAP1 knockdown or inhibition, suggesting that citrate synthase may also be a TRAP1 client [68].

Cell cycle dysregulation is a well-established driver of tumorigenesis [70]. TRAP1 impacts the cell cycle through regulation of protein quality control in cooperation with the proteasome regulator TBP7 [71,72]. Loss of the TRAP1/TBP7 machinery leads to increased ubiquitination and degradation of the G2-M checkpoint proteins CDK1 and MAD2 and dysregulation of mitotic entry [72]. However, whether TBP7 is a client or perhaps even the first co-chaperone of TRAP1 remains to be seen.

Taken together, these data describe multiple mechanisms through which TRAP1 dysregulation can impact cellular metabolic flux and, potentially, tumorigenesis.

#### *3.3. Evasion of Apoptosis*

Mitochondrial involvement in cell death is mediated by the release of cytochrome *c* [73,74]. Sustained opening of the permeability transition pore (PTP) within the inner mitochondrial membrane (IMM) initiates a series of events that lead to cytochrome *c* release and apoptosis or necrosis. Upon PTP opening, particles under 1500 Da, such as ions (Ca2+ , K<sup>+</sup> , and H<sup>+</sup> ), water, and other solutes, flood the IMM, causing swelling and unfolding of the cristae and eventual outer mitochondrial membrane (OMM) rupture. Subsequent efflux of cytochrome *c* through the compromised OMM into the cytosol induces the caspase cascade [75,76]. This sustained PTP opening is known as the mitochondrial permeability transition (PT) [77], and it can be triggered by several mechanisms, including elevated ROS, Ca2+, or inorganic phosphate levels, as well as decreased pH or ATP depletion [78]. Interplay between these elements also plays a role in its regulation, as elevated ROS has been shown to decrease the amount of Ca2+ required to trigger the PTP [76].

TRAP1 attenuation induces opening of the PTP and release of cytochrome *c* [47], and expression of TRAP1 likely discourages the initiation of apoptosis through two distinct, but potentially overlapping mechanisms: (1) regulation of triggers that signal into the PTP, and (2) direct disruption of the physical mechanism of PTP opening. TRAP1 knockdown has been shown to lead to increased ROS accumulation under oxidative stress [79] and TRAP1 overexpression insulates cells against iron chelation-mediated ROS production [80]. These effects are likely a consequence of both direct and indirect roles of TRAP1 in minimizing ROS generation. TRAP1 is a direct regulator of oxidative phosphorylation through its chaperoning of Complexes II and IV of the ETC [6,44,46] and has an indirect role in quenching existing ROS, as TRAP1 expression is associated with increased levels of the reduced form of the antioxidant glutathione (GSH) [81]. TRAP1-dependent regulation of ROS generation also results in decreased oxidation of the phospholipid cardiolipin. This phospholipid is responsible for the binding of cytochrome *c* to the inner folds of cristae, and its oxidation results in an increase of free cytochrome *c* in the inner membrane space that can potentially escape into the cytosol [78].

Furthermore, TRAP1 has been shown to chaperone the calcium-binding protein Sorcin [82]. TRAP1 is also thought to be responsible for Sorcin translocation into the mitochondria, given that Sorcin lacks its own mitochondrial localization sequence [8,82]. Overexpression of Sorcin in neonatal cardiac myocytes has been shown to increase mitochondrial Ca2+ levels, while simultaneously decreasing cytochrome *c* release, indicating an increase in mitochondrial Ca2+ tolerance [83]. Therefore, the chaperoning of Sorcin by TRAP1 is important for desensitizing the PTP to Ca2+ levels. Understanding this regulation is particularly important for TRAP1, as Ca2+ can replace Mg2+ as a co-factor and induce an increased rate of TRAP1 ATP hydrolysis [84]. TRAP1 has also been shown to decrease ubiquitination of the mitochondrial contact site and cristae organizing system subunit 60 (MIC60) under conditions of extracellular acidosis [85]. MIC60 is a critical component of the protein complex MICOS, which is regarded as the master organizer of the IMM through the formation of contact sites with the outer membrane and maintenance of cristae junctions [86,87]. Thus, TRAP1 regulation of MIC60 contributes to its anti-apoptotic function through the preservation of mitochondrial integrity.

Proposals for the structure of the PTP have gone through various iterations, however the prevailing model is that the PTP is formed by coordinated activities of the adenine nucleotide translocator (ANT) and the F-ATP synthase [88–90]. Furthermore, cyclophilin D (CypD) is key to PTP regulation [12,91]. Though its role in this process is controversial, CypD peptidyl-prolyl isomerase activity is required, as is its binding to the mitochondrial peripheral stalk subunit of the F-ATP synthase [63,90,92]. In addition to attenuating the triggers that lead to PTP opening, TRAP1 has been shown to antagonize the opening of the PTP itself. There is a general consensus that TRAP1 accomplishes this by forming a complex with CypD, interfering with the ability of CypD to interact with the PTP [12,63,93] potentially at the peripheral stalk of F-ATP synthase [90].

Further, the mitochondrial chaperones Hsp60 and Hsp90 have been implicated in this process, as their association with CypD also prevents PTP opening; however, the architecture of this complex has yet to be characterized [12,63,93–96].

#### **4. Post-Translational Regulation of TRAP1**

Post-translational modification is critically important to mitochondrial function [97] and has previously been shown to regulate TRAP1, though relatively little is known about individual PTM sites (Table 1, Figure 4) [5,6,98,99]. A comprehensive study of cytosolic Hsp90 has demonstrated the importance of post-translational regulation to Hsp90 chaperone activity (reviewed in [3,100]), and in the absence of certain co-chaperone regulatory proteins, specific PTM events have been shown to functionally recapitulate their activity [101]. This phenomenon may be critically important for TRAP1 biology, as TRAP1 is thought to act without the assistance of co-chaperones [8,10].


**Table 1.** Reported PTMs of TRAP1. Paralog identifies conserved residues in Hsp90α. GSNOR—Snitrosoglutathione reductase, ERK—extracellular signal-regulated kinase.

S511 (red) are highlighted, while S568 is absent.

S511 (red) are highlighted, while S568 is absent.

**Figure 4.** Ribbon structure of human TRAP1 (PDB: 6xg6) with known PTM sites. C501 (yellow) and

### *4.1. Phosphorylation*

*4.1. Phosphorylation*  PINK1 is a mitochondrially targeted serine/threonine kinase whose mutation and inactivation is linked to Parkinson's disease [102]. PINK1 activity has previously been shown to be cytoprotective [103], and when exposed to H2O2, cells transfected with siRNA targeting PINK1 showed significant increases in cytochrome *c* release and apoptosis [5]. TRAP1 was shown to be phosphorylated by PINK1 and mediate PINK1 anti-apoptotic activity, as evidenced by the observation that TRAP1 knockdown sensitized cells to PINK1 attenuation [5,104,105]. Interestingly, TRAP1 inhibition leads to activation of PINK1 is a mitochondrially targeted serine/threonine kinase whose mutation and inactivation is linked to Parkinson's disease [102]. PINK1 activity has previously been shown to be cytoprotective [103], and when exposed to H2O2, cells transfected with siRNA targeting PINK1 showed significant increases in cytochrome *c* release and apoptosis [5]. TRAP1 was shown to be phosphorylated by PINK1 and mediate PINK1 anti-apoptotic activity, as evidenced by the observation that TRAP1 knockdown sensitized cells to PINK1 attenuation [5,104,105]. Interestingly, TRAP1 inhibition leads to activation of PINK1, suggesting a reciprocal regulatory relationship [106].

**Figure 4***.* Ribbon structure of human TRAP1 (PDB: 6xg6) with known PTM sites. C501 (yellow) and

PINK1, suggesting a reciprocal regulatory relationship [106]. TRAP1 has also been shown to interact with the mitochondrial serine protease HTRA2 in Parkinson's disease [55]. Canonically, HTRA2 participates in mitochondrial and cellular quality control through inhibition of IAPs (inhibitor of apoptosis proteins) and induction of cell death, while loss of HTRA2 is associated with aberrant mitochondrial function and Parkinson's disease (PD). Overexpression of HTRA2 led to decreased levels of TRAP1, suggesting that HTRA2 may play a role in regulating TRAP1 stability [55]. However, the effect of HTRA2 was independent of its protease activity and the interaction between HTRA2 and TRAP1 was abrogated through treatment with mitochondrial respiratory inhibitors [55]. TRAP1 overexpression is also capable of rescuing mitochondrial dysfunction-associated PINK1 and HTRA2 loss. Interestingly, HTRA2 is also a substrate of PINK1, demonstrating that further work is needed to understand the mechanistic reg-TRAP1 has also been shown to interact with the mitochondrial serine protease HTRA2 in Parkinson's disease [55]. Canonically, HTRA2 participates in mitochondrial and cellular quality control through inhibition of IAPs (inhibitor of apoptosis proteins) and induction of cell death, while loss of HTRA2 is associated with aberrant mitochondrial function and Parkinson's disease (PD). Overexpression of HTRA2 led to decreased levels of TRAP1, suggesting that HTRA2 may play a role in regulating TRAP1 stability [55]. However, the effect of HTRA2 was independent of its protease activity and the interaction between HTRA2 and TRAP1 was abrogated through treatment with mitochondrial respiratory inhibitors [55]. TRAP1 overexpression is also capable of rescuing mitochondrial dysfunction-associated PINK1 and HTRA2 loss. Interestingly, HTRA2 is also a substrate of PINK1, demonstrating that further work is needed to understand the mechanistic regulation of TRAP1 by HTRA2 and the role of PINK1 in this system.

ulation of TRAP1 by HTRA2 and the role of PINK1 in this system. Neurofibromatosis is caused by mutation and inactivation of the Ras regulatory protein neurofibromin and is characterized by elevated Erk1/2 activity [10]. Active Erk1/2 is Neurofibromatosis is caused by mutation and inactivation of the Ras regulatory protein neurofibromin and is characterized by elevated Erk1/2 activity [10]. Active Erk1/2 is associated with TRAP1-SDH in the mitochondria of these cells, and Erk1/2-mediated phos-

associated with TRAP1-SDH in the mitochondria of these cells, and Erk1/2-mediated phosphorylation of TRAP1-S511/S568 strengthens their association, suggestive of a chap-

phorylation at these residues prevents tumor growth, in a succinate-dependent manner [10]. Mitochondrial Erk1/2 was previously shown to antagonize PTP opening [107], perhaps indicating a role for TRAP1 phosphorylation in PTP regulation as well. Taken together, these data suggest that TRAP1 inhibition or combined TRAP1-Erk1/2 targeting

Interaction with mitochondrially localized c-Src remains the only described TRAP1– tyrosine kinase relationship [6]. Previous work has shown that mitochondrial c-Src is involved in the phosphorylation-mediated activation of ETC Complexes I, II, and IV

may be a viable therapeutic strategy in neurofibromatosis and other cancers.

phorylation of TRAP1-S511/S568 strengthens their association, suggestive of a chaperone– client relationship. Association of TRAP1 and SDH decreases SDH activity, leading to accumulation of the oncometabolite succinate [10]. TRAP1 attenuation or loss of phosphorylation at these residues prevents tumor growth, in a succinate-dependent manner [10]. Mitochondrial Erk1/2 was previously shown to antagonize PTP opening [107], perhaps indicating a role for TRAP1 phosphorylation in PTP regulation as well. Taken together, these data suggest that TRAP1 inhibition or combined TRAP1-Erk1/2 targeting may be a viable therapeutic strategy in neurofibromatosis and other cancers.

Interaction with mitochondrially localized c-Src remains the only described TRAP1– tyrosine kinase relationship [6]. Previous work has shown that mitochondrial c-Src is involved in the phosphorylation-mediated activation of ETC Complexes I, II, and IV [108,109]. TRAP1 binds to and maintains c-Src in an inactive state, providing a potential mechanism for TRAP1 suppression of oxidative metabolism and ROS mitigation [6]. Though TRAP1 tyrosine phosphorylation is induced by c-Src expression and abrogated by c-Src inhibition, direct phosphorylation of TRAP1 by c-Src remains to be demonstrated. Taken together, TRAP1 and c-Src play opposing roles in the regulation of mitochondrial metabolism, though the reciprocal impact of c-Src on TRAP1 remains unresolved.

#### *4.2. Acetylation–Deacetylation*

Acetylation modulates protein–protein interactions via neutralization of Lys residues and can be reversed by the activity of deacetylases. TRAP1 directly stabilizes one such deacetylase, sirtuin-3 (SIRT3), and augments SIRT3 activity in vitro and in glioma cells [27]. Interestingly, SIRT3 overexpression was also able to rescue the effects of TRAP1 inhibition by the TRAP1 inhibitor gamitrinib [27]. One potential explanation for this observation is that SIRT3-mediated deacetylation of TRAP1 modulates TRAP1 activity or its affinity for gamitrinib, though no direct evidence was reported [27]. SIRT3 knockdown was also shown to increase ROS levels, and SIRT3 overexpression reversed an increase in ROS caused by gamitrinib [27]. Interestingly, attenuation of SIRT3 specifically destabilized TRAP1 substrates NDUFA9 (CI) and SDHB (CII), but not SIRT3 substrates SOD2 and GDH, suggesting that SIRT3-mediated deacetylation of TRAP1 is important for TRAP1 chaperone activity [27]. Interestingly, these interactions were observed in glioblastoma (GBM) cancer stem cells (CSC), which showed a preference for mitochondrial respiration over glycolysis. This work provides a new paradigm for understanding the role of SIRT3 in cancer [110]. Given this context and the known role of both proteins in regulating mitochondrial metabolism, reciprocal regulation of SIRT3 and TRAP1 may provide a positive feedback mechanism that impacts the ability of TRAP1 to chaperone its dependent proteins.

#### *4.3. Nitrosylation*

The PTM S-nitrosylation (SNO) is the result of the covalent addition of -NO to the thiol group of cysteine residues [111]. SNO is enzymatically catalyzed by nitrosylases and reversed by the activity of denitrosylases, including S-nitrosoglutathione reductase (GSNOR) [112]. GSNOR is commonly deleted in hepatocellular carcinoma (HCC), and GSNOR-KO mice develop HCC, linking aberrant nitrosylation to cancer [113]. TRAP1- C501-SNO was identified by mass spectrometry [54,114] and this modification was found to decrease TRAP1 ATPase activity, modulate conformational rearrangement, and promote its proteasomal degradation [54,98]. TRAP1 degradation also led to increased SDH activity, in agreement with previous work [44], and sensitized cells to SDH inhibitors, identifying TRAP1-SNO as a predictor of tumor cell response to this class of drugs [54]. It follows that mutation of this residue to TRAP1-C501S provided protection from apoptosis in the presence of nitric oxide donors, demonstrating that disruption of TRAP1-SNO is essential for its anti-apoptotic role [98]. Curiously, however, TRAP1 is overexpressed in many cancers, allowing for the possibility that TRAP1-SNO is context-specific and perhaps also under temporal regulation.

Taken together, PTMs exert influence on TRAP1 through regulating the kinetics of ATP hydrolysis and associated conformational rearrangements, interaction with client proteins, and promoting TRAP1 degradation.

#### **5. Current State of TRAP1 Inhibitor Development**

Inhibition of cell metabolism is a re-emerging anti-cancer strategy [115]. TRAP1 control of cellular metabolic flux and mitochondrial apoptosis outlined herein identifies TRAP1 inhibition as a potential anti-cancer therapeutic target. Efforts towards the development of ATP-competitive inhibitors for cytosolic Hsp90 have provided lead compounds for optimization to address the dual challenges of mitochondrial localization and TRAP1 specificity. Conjugation to a chemical scaffold such as the mitochondrial-targeting moiety triphenylphosphonium (TPP) is necessary to provide mitochondrial penetrance [116,117]. Specificity for TRAP1 over Hsp90 may also be a necessary consideration, as well-established Hsp90 ATP-competitive inhibitors cannot differentiate between the ATP-binding pockets, potentially leading to off-target toxicity [33].

#### *5.1. Gamitrinibs*

The most widely used mitochondrial Hsp90 inhibitors are gamitrinibs (G), small molecules consisting of the Hsp90 inhibitor 17-allylamino-17-demethoxygeldanamycin (17-AAG) attached to a mitochondrial-targeting moiety such as cyclic guanidinium repeats or TPP (G-G1-4 and G-TPP, respectively) [118]. These gamitrinibs have demonstrably reduced the viability of prostate [91,119–122], colon [119,123], melanoma [119,124], cervix [122,125], ovary [122], breast [118,119,121,124,125], and glioma cancers [126], particularly glioblastomas [120,124,127–129]. Gamitrinibs disrupt the anti-apoptotic effects of TRAP1, as evidenced by decreased mitochondrial membrane potential and increased cytochrome *c* release in G-TPP-treated PC3 prostate cancer cells [119]. Furthermore, the stability of the sensitive cytosolic Hsp90 client proteins Akt and phospho-Y416-Src was impacted by 17-AAG treatment, but unaffected by G-TPP in PC3 cells, demonstrating the selective targeting of gamitrinibs to the mitochondria [119]. A further consideration is the potential for resistance development, as PC3 cells continuously incubated with 17-AAG eventually became resistant to G-TPP, but not G-G4 [118,119]. This finding potentially suggests that the choice of mitochondrial-targeting moiety may be critically important and not necessarily limited simply to drug transport. Overall, selective TRAP1 inhibition with ATP-competitive gamitrinib derivatives remains a challenge. Further, these data emphasize the importance of understanding effectors of TRAP1 for the identification of potential combinatorial therapeutic targets to augment inhibition of TRAP1-mediated signaling pathways.

#### *5.2. Purine-Scaffold Inhibitors*

In addition to 17-AAG, mitochondrial targeting of the purine-scaffold Hsp90 inhibitor PU-H71 has also demonstrated efficacy against TRAP1. A TPP-conjugated derivative of PU-H71 (SMTIN-P01) showed a remarkable ability to target mitochondria over non-conjugated PU-H71 and a slight improvement in cytotoxicity over gamitrinibs [130]. Interestingly, adjustments to the length of the TPP resulted in changes in inhibitor behavior. When the TPP was modified to have a 10-length carbon chain (as opposed to the standard 6 length carbon chain), this so-called SMTIN-C10 induced structural changes to TRAP1 and demonstrated increased inhibition of TRAP1 [52]. SMTIN-C10 was found to bind to an allosteric binding site at E115 in the N-terminal domain of TRAP1, in addition to binding to the ATP pocket, resulting in TRAP1 adopting a closed formation [52]. This long linker approach was adapted for other TRAP1 inhibitors as well, including Mitoquinone. TPP-Mitoquinone has shown utility and specificity by targeting the client-binding middle domain of TRAP1 [117]. Mitoquinone has been demonstrated to have protective properties in various animal models of neurological maladies, such as traumatic brain injury [131], Huntington's disease [132], amyotrophic lateral sclerosis (ALS) [133], and Alzheimer's

disease [134]. This finding is contradictory to the working model of TRAP1 function, especially considering that TRAP1 downregulation is observed in Alzheimer's disease patients [135] and its overexpression is protective against oxidative stress in ALS [62]. These results highlight the need to understand the disease-specific contexts of TRAP1 function to identify appropriate disease models for the evaluation of TRAP1 inhibitors.

#### *5.3. New Inhibitors*

Since their discovery, Hsp90 inhibitors have primarily targeted the ATP-binding pocket (Figure 5). This is the mechanism of the natural product geldanamycin (GA) [136–138] and its derivatives, as well as the first synthetic inhibitor of TRAP1, Shepherdin [139]. Shepherdin was designed by imitating the minimal Hsp90-binding sequence of Survivin (aa 79–87), an anti-apoptotic protein that binds to the N-domain of Hsp90 [140]. Consequently, Shepherdin was also found to disrupt Hsp90-ATP binding with 13 predicted sites of hydrogen bonding in the ATP pocket [139]. Modeling studies based on the structure of Shepherdin identified the small molecule 5-aminoimidazole-4-carboxamide-1-β-D-ribofuranoside (AICAR), a previously characterized AMPK activator [141,142], as a potential Hsp90 inhibitor, though its development as a scaffold for Hsp90 inhibition has not been pursued. *Biomolecules* **2022**, *12*, x FOR PEER REVIEW 11 of 21 Shepherdin was designed by imitating the minimal Hsp90-binding sequence of Survivin (aa 79*–*87), an anti-apoptotic protein that binds to the N-domain of Hsp90 [140]. Consequently, Shepherdin was also found to disrupt Hsp90-ATP binding with 13 predicted sites of hydrogen bonding in the ATP pocket [139]. Modeling studies based on the structure of Shepherdin identified the small molecule 5-aminoimidazole-4-carboxamide-1-β-D-ribofuranoside (AICAR), a previously characterized AMPK activator [141,142]*,* as a potential Hsp90 inhibitor, though its development as a scaffold for Hsp90 inhibition has not been pursued.

**Figure 5.** Structures of discussed TRAP1 inhibitors. ATP-competitive small molecules targeting the N-domain of TRAP1 (PDB: 6xg6) are labeled blue, while allosteric inhibitors, primarily targeting the TRAP1 middle domain, are labeled red. SMTIN-C10 is a bifunctional inhibitor, with elements of both ATP-competitive and allosteric inhibition. **Figure 5.** Structures of discussed TRAP1 inhibitors. ATP-competitive small molecules targeting the N-domain of TRAP1 (PDB: 6xg6) are labeled blue, while allosteric inhibitors, primarily targeting the TRAP1 middle domain, are labeled red. SMTIN-C10 is a bifunctional inhibitor, with elements of both ATP-competitive and allosteric inhibition.

Though ATP-competitive Hsp90 inhibitors are still widely used, an alternative approach in hopes of achieving TRAP1 specificity over other Hsp90 family members has emerged through allosteric targeting. One example of this strategy is honokiol bis-dichlo-Though ATP-competitive Hsp90 inhibitors are still widely used, an alternative approach in hopes of achieving TRAP1 specificity over other Hsp90 family members has

> roacetate (HDCA), which is able to specifically inhibit TRAP1 by binding to an allosteric pocket within the middle domain. This pocket has a surface landscape defined by a positively charged region sandwiched between two negatively charged regions that are

emerged through allosteric targeting. One example of this strategy is honokiol bis-dichloroacetate (HDCA), which is able to specifically inhibit TRAP1 by binding to an allosteric pocket within the middle domain. This pocket has a surface landscape defined by a positively charged region sandwiched between two negatively charged regions that are separated from each other by a large hydrophobic area. HDCA binds in this hydrophobic area and allosterically inhibits TRAP1 ATPase activity, but not that of Hsp90 [43]. *Biomolecules* **2022**, *12*, x FOR PEER REVIEW 12 of 21 separated from each other by a large hydrophobic area. HDCA binds in this hydrophobic area and allosterically inhibits TRAP1 ATPase activity, but not that of Hsp90 [43]. Further, computational methods by Sanchez-Martin et al. utilized the unique asym-

Further, computational methods by Sanchez-Martin et al. utilized the unique asymmetry of TRAP1 to identify an allosteric pocket on the straight protomer of the TRAP1 dimer that can serve as a TRAP1-specific inhibitor binding surface [42]. Inter-domain communication is essential to the ATPase cycle of TRAP1, and previous work has shown that inhibitor-bound TRAP1 stalls in the NTD dimerized phase [143]. In agreement, the computationally identified compounds (compounds 5–7) were hypothesized to inhibit TRAP1 by reducing the ability of the ATP-binding site to communicate with the client-binding region of the middle domain. In fact, several of these small molecules were shown to decrease TRAP1 ATPase activity to a degree comparable to that of 17-AAG, while not significantly interfering with Hsp90 ATPase activity, demonstrating specificity for TRAP1 [42]. Furthermore, allosterically inhibited TRAP1 bound approximately 30% less SDHA than its control and experienced a significant increase in succinate-coenzyme-Q reductase (SQR) activity. While the tested compound did not alter cell viability, it delayed cell proliferation over a 96 h observation [42]. The successful utilization of TRAP1 asymmetry to identify unique allosteric binding pockets provides a significant starting point for future inhibitor work. metry of TRAP1 to identify an allosteric pocket on the straight protomer of the TRAP1 dimer that can serve as a TRAP1-specific inhibitor binding surface [42]. Inter-domain communication is essential to the ATPase cycle of TRAP1, and previous work has shown that inhibitor-bound TRAP1 stalls in the NTD dimerized phase [143]. In agreement, the computationally identified compounds (compounds 5–7) were hypothesized to inhibit TRAP1 by reducing the ability of the ATP-binding site to communicate with the client-binding region of the middle domain. In fact, several of these small molecules were shown to decrease TRAP1 ATPase activity to a degree comparable to that of 17-AAG, while not significantly interfering with Hsp90 ATPase activity, demonstrating specificity for TRAP1 [42]. Furthermore, allosterically inhibited TRAP1 bound approximately 30% less SDHA than its control and experienced a significant increase in succinate-coenzyme-Q reductase (SQR) activity. While the tested compound did not alter cell viability, it delayed cell proliferation over a 96 h observation [42]. The successful utilization of TRAP1 asymmetry to identify unique allosteric binding pockets provides a significant starting point for future inhibitor work. **6. Future Perspectives** 

#### **6. Future Perspectives** The function of TRAP1 as a regulator of cellular metabolic flux and mitochondrial

The function of TRAP1 as a regulator of cellular metabolic flux and mitochondrial apoptosis underscores a duality in which cell fate decisions are determined (Figure 6). Normal cells demonstrate basal TRAP1 expression, facilitating oxidative metabolism and programmed cell death. Dysregulation of TRAP1 expression manifests in noted hallmarks of cancer, including cell death resistance and deregulation of cellular energetics [144]. A thorough delineation of the mechanism of TRAP1 function in these roles is essential to combatting diseases of mitochondrial dysfunction, including cancer and neurodegeneration. apoptosis underscores a duality in which cell fate decisions are determined (). Normal cells demonstrate basal TRAP1 expression, facilitating oxidative metabolism and programmed cell death. Dysregulation of TRAP1 expression manifests in noted hallmarks of cancer, including cell death resistance and deregulation of cellular energetics [144]. A thorough delineation of the mechanism of TRAP1 function in these roles is essential to combatting diseases of mitochondrial dysfunction, including cancer and neurodegeneration.

**Figure 6.** The multiple roles of TRAP1 in cancer cell mitochondria, revolving around evasion of apoptosis and metabolic reprogramming. TRAP1 acts as a chaperone for the Ca2+ binding protein Sorcin as well as Complexes II and IV of the ETC. Increased TRAP1 levels are associated with calcium tolerance, increased levels of the antioxidant glutathione, reduced levels of ROS, reduced levels of MIC60 ubiquitination, and in many cases, a shift towards the Warburg effect. TRAP1, along with Hsp90 and Hsp60, can form a complex with CypD to prevent opening of the PTP. TRAP1 is post-translationally modified by PINK1, Erk1/2, GSNOR, and SIRT3. **Figure 6.** The multiple roles of TRAP1 in cancer cell mitochondria, revolving around evasion of apoptosis and metabolic reprogramming. TRAP1 acts as a chaperone for the Ca2+ binding protein Sorcin as well as Complexes II and IV of the ETC. Increased TRAP1 levels are associated with calcium tolerance, increased levels of the antioxidant glutathione, reduced levels of ROS, reduced levels of MIC60 ubiquitination, and in many cases, a shift towards the Warburg effect. TRAP1, along with Hsp90 and Hsp60, can form a complex with CypD to prevent opening of the PTP. TRAP1 is post-translationally modified by PINK1, Erk1/2, GSNOR, and SIRT3.

Though our understanding of the cellular impact of TRAP1 is coming into focus, several outstanding questions remain that are essential to our comprehension of the full scope of TRAP1 biology. (1) Is TRAP1 ATPase activity, and by extension TRAP1 chaperone function, essential for its biological activity? ATP-competitive inhibitors of TRAP1 demonstrate efficacy in cell models of cancer, suggesting that TRAP1 function is coupled to its ATPase activity; however, catalytically inactive TRAP1 mutants are able to complement TRAP1 function and revert metabolic dysfunction [26]. Reconciling these disparate observations is an ongoing challenge. (2) What is the physiological impact of TRAP1 dimeric and tetrameric forms, and is transition between these states essential for its function? Cytosolic Hsp90s are well-established dimers, and though the domain architecture of TRAP1 is similar, it remains unclear whether the TRAP1 dimer is the primary biological unit. (3) Is specific targeting of TRAP1 in cancer essential? Many existing TRAP1 inhibitors are mitochondrially targeted Hsp90 inhibitors. Though strategic inhibition of cytosolic Hsp90 has yet to demonstrate clinical success, perhaps simultaneous disruption of TRAP1 and the mitochondrial Hsp90 pool will prove efficacious [145]. (4) Can TRAP1 be used as a biomarker in cancer? Previous work has demonstrated that circulating Hsp90 can potentially be used as a biomarker in certain conditions, however the presence of circulating TRAP1 has not been evaluated [146–148]. Similarly, TRAP1 expression and activity is dysregulated in cancer, potentially suggesting an ability to serve as a predictive indicator of disease state. (5) TRAP1 mutations have been implicated in several conditions, including congenital anomalies of the kidney and urinary tract (CAKUT), vertebral defects, anal atresia, cardiac defects, tracheo-esophageal fistula, renal anomalies, and limb abnormalities (VACTERL), Parkinson's disease, cardiac hypertrophy, and severe autoinflammation [55,149–151]. What is the structural basis for the impact of these mutations on TRAP1 function? Is mutant-TRAP1 association with these diseases a consequence of its role as a more general regulator of mitochondrial dynamics [152]? (6) Can differential PTM of TRAP1 in normal and disease states predict disease-associated phenotypes? Indeed, it has been shown that PTMs modulate TRAP1, however whether this necessarily predicts TRAP1 behavior in disease states remains to be tested. (7) Do TRAP1 PTMs compensate for a lack of dedicated co-chaperones? In the case of Hsp90, a single phosphorylation can functionally replace the loss of the yeast co-chaperone Hch1 [101]. The relevance of this mechanism for TRAP1 has not yet been investigated, however the reliance of cytosolic Hsp90 on co-chaperone interaction suggests that TRAP1 PTMs can recapitulate some co-chaperone activities. (8) Can these PTMs be specifically manipulated to alter TRAP1 function? Many cancers are associated with increased TRAP1 activity, and decreased TRAP1 activity or loss-of-function mutations contribute to the pathogenesis of some neurodegenerative diseases [153]. Previous work discussed here demonstrates that PTMs play a role in the regulation of TRAP1 stability, and TRAP1 PTMs are dysregulated in disease. High-throughput methods [154] as well as the study of cytosolic Hsp90s suggest that TRAP1 function will be regulated by a constellation of PTMs with differential incidence that correlates with disease state [3].

The literature reviewed here from several experimental systems demonstrates that in cancers that overexpress TRAP1, attenuation of TRAP1 expression or activity is sufficient to slow cell growth, and in some instances, induce apoptosis. Furthermore, nuanced studies of Hsp90 have demonstrated that PTM can modulate the efficacy of Hsp90 inhibitors [3], implying a similar framework for the application of TRAP1 inhibitors. The identification of predictive indicators of response to TRAP1 inhibition and potential targets for anti-cancer therapy in combination with TRAP1 inhibitors are two essential pieces of information that can be gained from decrypting the TRAP1 chaperone code.

**Author Contributions:** Conceptualization, M.R.W.; writing—original draft preparation, L.A.W. and M.R.W.; writing—review and editing, L.A.W., S.J.B., D.B., M.M. and M.R.W.; visualization, L.A.W., S.J.B. and M.R.W.; supervision, M.R.W.; project administration, M.R.W.; funding acquisition, D.B., M.M. and M.R.W. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was partly supported by funds from SUNY Upstate Medical University and The Upstate Foundation (M.M., D.B. and M.R.W.).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Acknowledgments:** We would like to thank Giorgio Colombo and Len Neckers for insightful scientific discussions. The figures were created using elements from Biorender.com on an institutional license from SUNY Upstate Medical University.

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

#### **Abbreviations**

17-AAG, 17-allylamino-17-demethoxygeldanamycin; AICAR, 5-aminoimidazole-4-carboxamide-1-β-D-ribofuranoside; ALS, amyotrophic lateral sclerosis; ANT, adenosine nucleotide translocator; ATP, adenosine triphosphate; ATPB, ATP synthase subunit beta; AMPK, AMP-activated protein kinase; CAKUT, congenital anomalies of the kidney and urinary tract, CDK1, cyclin-dependent kinase 1; COXII/IV, cytochrome *c* oxidase subunit 2/4; CSC, cancer stem cells; CypD, cyclophilin D; Erk1/2, extracellular signal-regulated kinase 1/2; ETC, electron transport chain; G, gamitrinib; G1-4, cyclic guanidinium 1-4; GA, geldanamycin; GDH, glutamate dehydrogenase; GMB, glioblastoma; GSH, reduced glutathione; GSNOR, S-nitrosoglutathione reductase; HCC, hepatocellular carcinoma; HDCA, honokiol bis-dichloroacetate; HIF1α, hypoxia inducible factor 1α; Hsp, heat shock protein; Hsp60, heat shock protein 60; Hsp90, heat shock protein 90; HTRA2, high-temperature requirement A2; IMM, inner mitochondrial membrane; KO, knockout; MAD2, mitotic arrest deficiency 2; MICOS, mitochondrial contact site and cristae organizing system; NDUFA9, NADH:ubiquinone oxidoreductase subunit A9; NRF2, nuclear factor erythroid 2-related factor 2; NTD, N-terminal domain; OMM, outer mitochondrial membrane; PD, Parkinson's disease; PINK1, phosphatase and tensin homolog (PTEN)-induced kinase 1; PPP, pentose phosphate pathway; PT, mitochondrial permeability transition; PTM, post-translational modification; PTP, permeability transition pore; ROS, reactive oxygen species; SDH, succinate dehydrogenase; SDHA/B, succinate dehydrogenase subunit a/b; SIRT3, sirtuin-3; SNO, S-nitrosylation; SOD2, superoxide dismutase 2; SQR, succinate-coenzyme-Q reductase; TBP7, TATA-Box binding protein 7; TCA, tricarboxylic acid cycle; TPP, triphenylphosphonium; TRAP1, tumor necrosis factor (TNF) receptor-associated protein-1; VACTERL, vertebral defects, anal atresia, cardiac defects, tracheo-esophageal fistula, renal anomalies, and limb abnormalities; VHL, Von Hippel Lindau.

#### **References**

