*Article* **The APE2 Exonuclease Is a Client of the Hsp70–Hsp90 Axis in Yeast and Mammalian Cells**

**Siddhi Omkar \*, Tasaduq H. Wani † , Bo Zheng † , Megan M. Mitchem and Andrew W. Truman \***

> Department of Biological Sciences, The University of North Carolina at Charlotte, Charlotte, NC 28223, USA; twani@uncc.edu (T.H.W.); bzheng1@uncc.edu (B.Z.); mmitch92@uncc.edu (M.M.M.) **\*** Correspondence: sparanj2@uncc.edu (S.O.); atruman1@uncc.edu (A.W.T.)

† These authors contributed equally to this work.

**Abstract:** Molecular chaperones such as Hsp70 and Hsp90 help fold and activate proteins in important signal transduction pathways that include DNA damage response (DDR). Previous studies have suggested that the levels of the mammalian APE2 exonuclease, a protein critical for DNA repair, may be dependent on chaperone activity. In this study, we demonstrate that the budding yeast Apn2 exonuclease interacts with molecular chaperones Ssa1 and Hsp82 and the co-chaperone Ydj1. Although Apn2 does not display a binding preference for any specific cytosolic Hsp70 or Hsp90 paralog, Ssa1 is unable to support Apn2 stability when present as the sole Ssa in the cell. Demonstrating conservation of this mechanism, the exonuclease APE2 also binds to Hsp70 and Hsp90 in mammalian cells. Inhibition of chaperone function via specific small molecule inhibitors results in a rapid loss of APE2 in a range of cancer cell lines. Taken together, these data identify APE2 and Apn2 as clients of the chaperone system in yeast and mammalian cells and suggest that chaperone inhibition may form the basis of novel anticancer therapies that target APE2-mediated processes.

**Keywords:** Hsp70; Hsp90; APE2; Apn2; cancer; chaperone inhibition

#### **1. Introduction**

The well-conserved Hsp70 and Hsp90 molecular chaperones are critical for the folding, maturation and activity of a large number of "client" proteins [1]. Client proteins are found in diverse cellular pathways, and consequently, chaperones support the maintenance of apoptotic signaling, angiogenesis, autophagy, senescence [1–3]. Although prokaryotes possess a single prototypical Hsp70 and Hsp90 (DnaK and HtpG, respectively), eukaryotes possess several paralogs that differ in their subcellular localization and expression profile [4–6]. In budding yeast, the main cytosolic forms of Hsp70 are Ssa1–4, which arose from multiple gene duplication events. Ssa1 and 2 are constitutively expressed at high levels, whereas Ssa3 and 4 are highly heat inducible [7–9]. The Ssa paralogs are semiredundant, evidenced by the fact that yeast remain viable as long as they have one paralog expressed at constitutively high levels [7–9]. Despite their relatedness, recent studies suggest that the Ssa paralogs have slightly different client binding profiles [4]. Similarly, humans encode 13 isoforms of Hsp70s from a multigene family with major cytosolic paralogs being HspA8 (constitutive) and HspA1A/HspA1L (inducible) [10–12]. Hsp90 also exists in various forms in cells. In mammalian cells, the inducible Hsp90a and constitutively expressed Hsp90b are the major species in the cytosol, equivalent to yeast Hsp82 and Hsc82, respectively [5,13]. A major stress that cells must deal with to survive are challenges to genome integrity in the form of DNA damage [14]. The sensing of DNA damage and its repair are mediated by an array of proteins that together form the DNA damage response (DDR) pathway [15]. While chaperones support many key signal transduction pathways in the cell, evidence is building to support a particularly critical role for chaperones in the detection and repair of DNA damage. Hsp70 and Hsp90 support DDR by activating and stabilizing a huge number of DDR proteins including p53, CHK1, FANCA, FANCD2,

**Citation:** Omkar, S.; Wani, T.H.; Zheng, B.; Mitchem, M.M.; Truman, A.W. The APE2 Exonuclease Is a Client of the Hsp70–Hsp90 Axis in Yeast and Mammalian Cells. *Biomolecules* **2022**, *12*, 864. https:// doi.org/10.3390/biom12070864

Academic Editor: Chrisostomos Prodromou

Received: 2 June 2022 Accepted: 18 June 2022 Published: 21 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/).

BRCA1/2, MRN and RNR complexes [16–18]. A common type of DNA damage is the loss of a base from genomic DNA, known as apurinic/apyrimidinic (AP) sites. The repair of such sites involves the recruitment of the related APE1 and APE2 exonucleases (Apn1 and Apn2 in yeast) [19–24]. Although APE1 and APE2 display functional overlap, APE2 possesses an extra C-terminal domain that is absent in APE1 and lacks any redox activity [22]. A recent study examined global protein abundance and epigenetic changes in response to Hsp90 inhibition. Several DDR proteins were among those found to decrease upon ganetespib and AUY922 treatment, including XRCC1, XPC and APE2 [25]. While APE1 becomes associated with Hsp70 during DNA repair to augment endonuclease activity, no such mechanistic connection between chaperones and APE2 has been identified [26]. In this study, we demonstrate a novel interaction between APE2/Apn2 and the Hsp70–Hsp90 system in yeast and mammalian cells. Although there appears to be no preference for which Hsp90 or Hsp70 paralog APE2/Apn2 bind, yeast Apn2 is destabilized in yeast lacking Ssa2, 3 and 4. Inhibition of Hsp90 via ganetespib or Hsp70 via JG-98 triggered a surprisingly rapid reduction of APE2 in a range of cancer cell lines. Understanding the intricacies of chaperone–endonuclease interactions could lead to more targeted and less toxic cancer therapeutics that exploit the genomic instability often seen in tumor cells.

#### **2. Materials and Methods**

#### *2.1. Yeast Strains and Growth Conditions*

Yeast cultures were grown in either YPD (1% yeast extract) US Biological Life Sciences, Swampscott, MA, USA, 2% glucose (VWR, Radnor, PA, USA), 2% peptone (Thermo Fisher Scientific, Waltham, MA, USA) or in SD (0.67% yeast nitrogen base without amino acids and carbohydrates (US Biological Life Sciences), 2% glucose), supplemented with the appropriate nutrients to select for plasmids and tagged genes. *Escherichia coli* DH5α was used to propagate all plasmids. *E. coli* cells were cultured in Luria broth medium (1% Bacto tryptone, 0.5% Bacto yeast extract, 1% NaCl) and transformed to ampicillin resistance by standard methods. Hsp70 isoform plasmids were transformed into yeast strain *ssa1–4*∆ [27] using PEG/lithium acetate. After restreaking onto media lacking leucine, transformants were streaked again onto media lacking leucine and containing 5-fluoro-orotic acid (5-FOA) (US Biological Life Sciences), resulting in yeast that expressed Hsp70 paralogs as the sole cytoplasmic Hsp70 in the cell. For a full description of yeast strains see Table 1 and for plasmids see Table 2.

#### *2.2. Purification of HA-Tagged Apn2 from Yeast*

The protocol followed for HA-IP was taken from [28] with slight modifications. Cells transformed with control pRS316 plasmid or the plasmid-expressing HA-tagged Apn2 [26] were grown overnight in SD-URA media and then re-inoculated into a larger culture of selectable media and grown to an OD<sup>600</sup> of 0.800. Cells were harvested, and HA-tagged proteins were isolated as follows. Protein was extracted via bead beating in 500 µL protein extraction buffer (50 mM Na-phosphate pH 8.0, 300 mM NaCl, 0.01% Tween-20). Then, 1000 µg of protein extract was incubated with 25 µL anti-HA magnetic beads (Thermo Fisher Scientific) at 30 ◦C for 30 min. Anti-HA beads were collected by magnet and then washed 3 times with TBS-T and 2 times with protein extraction buffer. After the final wash, the buffer was aspirated, and beads were incubated with 75 µL protein extraction buffer, and 25 µL 5× SDS-PAGE sample buffer sample was denatured for 5 min at 95 ◦C and boiled for 10–15 min. Next, the beads were collected via magnet, and the supernatant-containing purified HA-Apn2 was transferred to a fresh tube. Then, 20 µL of each sample was analyzed on SDS-PAGE.

#### *2.3. Mammalian Cell Culture and Drug Treatment*

The protocol used for transfection and drug treatment was taken from [22] with slight modifications. HEK293T cells were cultured in Dulbecco's modified Eagle's minimal essential medium (DMEM; Invitrogen, Carlsbad, CA, USA) supplemented with 10% fetal bovine

serum (FBS; Invitrogen), 100 U/mL penicillin (Invitrogen) and 100 µg/mL streptomycin (Invitrogen). L-GlutaMAX nutrient mixture (Gibco, Waltham, MA, USA, Cat#31765-035) (10% FBS, 100 units of penicillin and 100 units of streptomycin) was used to culture PC3, RPMI 1640 based medium (10% FBS, 100 units of penicillin and 100 units of streptomycin, 1% L-GlutaMAX-I) for LNCaP and DMEM-based medium (10% FBS, 100 units of penicillin and 100 units of streptomycin, 1% L-GlutaMAX-I) for MCF7. All cell lines were incubated at 37 ◦C in a 5% CO<sup>2</sup> containing atmosphere. Cells were seeded in 6-well plates at <sup>1</sup> <sup>×</sup> <sup>10</sup>6/2 mL per well one day prior to transfection. Cells were transfected by APE2 expression plasmid pcDNA-APE2-HA-BCP [29] with Lipofectamine3000 transfection kit (Invitrogen, Cat#L3000-015), and 2.5 µg of DNA and 7.5 µL of Lipofectamine3000 were used for each well. Briefly, diluted Lipofectamine3000 and DNA plus P3000 with Opti-MEM I (Gibco, Cat#31985-070) were mixed and incubated at room temperature for 15 min and then added to cell culture dropwise. The cells were treated for 0, 2, 4, 8 and 16 h post 48 h transfection with 10 µM JG-98, which is a Hsp70 inhibitor or 10 µM ganetespib (STA-9090, Selleckchem, Houston, TX, USA, Cat#S1159) for Hsp90 inhibition.

#### *2.4. Transfections and Co-Immunoprecipitation in Mammalian Cells*

The protocol used for transfection and drug treatment was adapted from [28] with slight modifications. HEK293T cells or specific cancer cells such as PC3, LNCaP and MCF7 were either untransfected (mock) or transfected with plasmids for expression of HA-tagged and/or V5-tagged proteins for constitutive HSPA8 and inducible HSPA1L and HSPA1A using Lipofectamine 3000 (Thermo Fisher Scientific). After 48 h, the cells were washed with 1X PBS, and total cell extract was prepared from the cells using M-PER (Thermo Fisher Scientific) containing EDTA-free protease and phosphatase inhibitor cocktail (Thermo Fisher Scientific) according to the manufacturer's recommended protocol. Protein was quantitated using the Bradford Assay. HA-tagged proteins were purified as follows. First, 200 µg of protein extract was incubated with 25 µL anti-HA magnetic beads (Thermo Fisher Scientific) at 30 ◦C for 30 min. Anti-HA beads were collected by magnet and then washed 3 times with TBS-T and 2 times with protein extraction buffer. After the final wash, the buffer was aspirated, and beads were incubated with 75 µL protein extraction buffer, and 25 µL 5× SDS-PAGE sample buffer sample was denatured for 5 min at 95 ◦C and boiled for 10–15 min. Next, the beads were collected via magnet, and the supernatant-containing purified HA-APE2 was transferred to a fresh tube. Finally, 20 µL of each sample was analyzed on SDS-PAGE.

#### *2.5. Western Blotting*

First, 20 µg of protein was separated by 4–12% NuPAGE SDS-PAGE (Thermo Fisher Scientific). Proteins were detected using the following antibodies; anti-HA tag (Thermo Fisher Scientific), Anti-FLAG tag (Sigma-Aldrich, St. Louis, MO, USA, USA #F1365), anti-PGK (Thermo Fisher Scientific, #MA5-15738), anti-Ydj1 (Stressmarq Biosciences Inc., Victoria, BC, Canada, #SMC-166D), anti-HDJ2 (Thermo Fisher Scientific, #MA512748). Blots were imaged on a ChemiDoc MP imaging system (Bio-Rad, Hercules, CA, USA). After treatment with Super Signal West Pico Chemiluminescent Substrate (GE Healthcare, Piscataway, NJ, USA). Blots were stripped and reprobed with the relevant antibodies using Restore Western Blot Stripping Buffer (Thermo Fisher Scientific).

#### **3. Results**

#### *3.1. Apn2 Interacts with Ydj1, Hsp82 and Ssa1 in Yeast*

Previous studies suggested that inhibition of Hsp90 may lead to loss of APE2 in bladder cancer [25]. To determine whether there was a connection between yeast APE2 (Apn2) and chaperones, we purified HA-tagged Apn2 from yeast and probed the complex with anti-HA, anti-Hsp82, anti-Ssa1, and anti-Ydj1 antibodies. We observed a clear association with Ssa1, Hsp82 and Ydj1 (Figure 1A). There are four cytosolic Hsp70s in yeast, Ssa1, 2, 3 and 4, which are highly similar to the amino acid sequence that arose from multiple

yeast gene duplication events [4]. While these paralogs have clear functional overlap, they also display differential client preferences [4]. To determine whether all Ssa paralogs can interact with Apn2, we performed co-immunoprecipitation experiments in WT BY4742 yeast cells (Table 1) expressing plasmids-HA-Apn2 and exogenous Flag-Ssa1, 2, 3 or 4 (Figure 1B). In this context, Apn2 bound equally to all Ssa paralogs (Figure 1B). To query whether all four Ssa paralogs could support Apn2 stability, we examined the levels of constitutively expressed HA-Apn2 in *ssa1–4*∆ yeast, expressing only one of the four Ssa proteins (Table 1). The levels of Apn2 were significantly decreased in yeast-expressing Ssa1 as the sole Ssa paralog in the cell (Figure 1C,D). Co-chaperones of Hsp70 play an important role in regulating chaperone activity and specificity [30]. We wondered whether Ydj1, a major co-chaperone of Ssa1–4, may support Apn2 levels in a similar way to its chaperoning of the ribonucleotide complex [28]. To test this possibility, we compared the abundance of Apn2 in WT yeast and those lacking Ydj1 (Table 1). In contrast to the regulation of RNR, the lack of Ydj1 had minimal impact on Ape2 levels (Figure 1E,F). with Ssa1, Hsp82 and Ydj1 (Figure 1A). There are four cytosolic Hsp70s in yeast, Ssa1, 2, 3 and 4, which are highly similar to the amino acid sequence that arose from multiple yeast gene duplication events [4]. While these paralogs have clear functional overlap, they also display differential client preferences [4]. To determine whether all Ssa paralogs can interact with Apn2, we performed co‐immunoprecipitation experiments in WT BY4742 yeast cells (Table1) expressing plasmids‐HA‐Apn2 and exogenous Flag‐Ssa1, 2, 3 or 4 (Fig‐ ure 1B). In this context, Apn2 bound equally to all Ssa paralogs (Figure 1B). To query whether all four Ssa paralogs could support Apn2 stability, we examined the levels of constitutively expressed HA‐Apn2 in *ssa1–4*∆ yeast, expressing only one of the four Ssa proteins (Table 1). The levels of Apn2 were significantly decreased in yeast‐expressing Ssa1 as the sole Ssa paralog in the cell (Figure 1C,D). Co‐chaperones of Hsp70 play an important role in regulating chaperone activity and specificity [30]. We wondered whether Ydj1, a major co‐chaperone of Ssa1–4, may support Apn2 levels in a similar way to its chaperoning of the ribonucleotide complex [28]. To test this possibility, we compared the abundance of Apn2 in WT yeast and those lacking Ydj1 (Table 1). In contrast to the regulation of RNR, the lack of Ydj1 had minimal impact on Ape2 levels (Figure 1E,F).

and chaperones, we purified HA‐tagged Apn2 from yeast and probed the complex with anti‐HA, anti‐Hsp82, anti‐Ssa1, and anti‐Ydj1 antibodies. We observed a clear association

*Biomolecules* **2022**, *12*, x 4 of 13

**Figure 1.** Apn2 interacts with Hsp82, Hsp70 and Ydj1 in yeast. (**A**) Yeast cells expressing Apn‐HA were grown to mid‐log phase at 30 °C. Lysate from these cells were analyzed by Western blotting with an anti‐HA, anti‐Ssa1, anti‐Ydj1 and anti‐Hsp82 antibody. Pgk1 was used as a loading control. Immunoprecipitation was performed using anti‐HA magnetic beads, and the interaction was stud‐ ied. (**B**) WT cells were co‐transformed with Apn2‐HA and individual Ssa isoforms. Yeast cells were grown to mid‐log phase at 30 °C. Lysates were analyzed by Western blotting with HA and FLAG specific antibody. Immunoprecipitation was performed using anti‐HA magnetic beads, and inter‐ action between FLAG‐Ssa and Apn2‐HA was checked using anti‐HA and anti‐FLAG antibodies on Western blot. (**C**) Yeast expressing the indicated FLAG‐Ssa (on a constitutive promoter) in a *ssa1– 4*∆ background transformed with Apn2‐HA were grown to mid‐log phase at 30 °C. Lysates were analyzed by Western blotting with HA‐ and FLAG‐specific antibodies. (**D**) Relative abundance of Apn2‐HA was quantitated by taking the ratio of Apn2‐HA/PGK1. Data are the mean and SD of three replicate experiments and compared to Ssa2 (\*\* *p* < 0.001) (**E**) WT BY4742 and Ydj1∆ cells, were **Figure 1.** Apn2 interacts with Hsp82, Hsp70 and Ydj1 in yeast. (**A**) Yeast cells expressing Apn-HA were grown to mid-log phase at 30 ◦C. Lysate from these cells were analyzed by Western blotting with an anti-HA, anti-Ssa1, anti-Ydj1 and anti-Hsp82 antibody. Pgk1 was used as a loading control. Immunoprecipitation was performed using anti-HA magnetic beads, and the interaction was studied. (**B**) WT cells were co-transformed with Apn2-HA and individual Ssa isoforms. Yeast cells were grown to mid-log phase at 30 ◦C. Lysates were analyzed by Western blotting with HA and FLAG specific antibody. Immunoprecipitation was performed using anti-HA magnetic beads, and interaction between FLAG-Ssa and Apn2-HA was checked using anti-HA and anti-FLAG antibodies on Western blot. (**C**) Yeast expressing the indicated FLAG-Ssa (on a constitutive promoter) in a *ssa1–4*∆ background transformed with Apn2-HA were grown to mid-log phase at 30 ◦C. Lysates were analyzed by Western blotting with HA- and FLAG-specific antibodies. (**D**) Relative abundance of Apn2-HA was quantitated by taking the ratio of Apn2-HA/PGK1. Data are the mean and SD of three replicate experiments and compared to Ssa2 (\*\* *p* < 0.001) (**E**) WT BY4742 and Ydj1∆ cells, were transformed with HA-Apn2 plasmid. Transformants were grown to mid-log phase at 30 ◦C. Lysate from these cells was analyzed by Western blotting with an anti-HA and anti-Ydj1 antibody. (**F**) Relative abundance of Apn2-HA was quantitated by taking the ratio of Apn2-HA/PGK1. Data are the mean and SD of three replicate experiments and compared to WT.

#### *3.2. Apn2 Interacts with Both Hsp82 and Is a Client of Hsp90 in Yeast 3.2. Apn2 Interacts with Both Hsp82 and Is a Client of Hsp90 in Yeast*

Our previous results suggested that Apn2 may also be a direct client of Hsp90. To test this hypothesis, we examined Apn2 in yeast expressing a well-characterized temperature sensitive point mutation in Hsp90 [31]. Cells expressing HA-Apn2 in either Hsp82G170D (Table 1) or WT (Table 1) were grown at 25 ◦C until early mid-log phase and were split into two flasks, one of which was shifted to 39 ◦C. Cells were lysed after 90 min, and HA-Apn2 levels were examined by Western blot. Incubation at 39 ◦C caused a significant decrease in HA-Apn2 levels in Hsp82G170D cells, while HA-Apn2 levels remained unchanged in WT cells, confirming Apn2 as a client of Hsp90 (Figure 2A). There are two Hsp90 paralogs in yeast, the heat-inducible Hsp82 and constitutive Hsc82 (Table 1). To assess Apn2 binding preferences for the two Hsp90s, we purified Apn2 from yeast expressing tagged versions of Hsp82 or Hsc82 using anti-HA magnetic beads. Consistent with our results in Figure 1B (above), the binding of Apn2 was equal to both heat-inducible Hsp82 and constitutive Hsc82 (Figure 2C). Our previous results suggested that Apn2 may also be a direct client of Hsp90. To test this hypothesis, we examined Apn2 in yeast expressing a well‐characterized temper‐ ature sensitive point mutation in Hsp90 [31]. Cells expressing HA‐Apn2 in either Hsp82G170D (Table1) or WT (Table1) were grown at 25 °C until early mid‐log phase and were split into two flasks, one of which was shifted to 39 °C. Cells were lysed after 90 min, and HA‐Apn2 levels were examined by Western blot. Incubation at 39 °C caused a signif‐ icant decrease in HA‐Apn2 levels in Hsp82G170D cells, while HA‐Apn2 levels remained un‐ changed in WT cells, confirming Apn2 as a client of Hsp90 (Figure 2A). There are two Hsp90 paralogs in yeast, the heat‐inducible Hsp82 and constitutive Hsc82 (Table 1). To assess Apn2 binding preferences for the two Hsp90s, we purified Apn2 from yeast ex‐ pressing tagged versions of Hsp82 or Hsc82 using anti‐HA magnetic beads. Consistent with our results in Figure 1B (above), the binding of Apn2 was equal to both heat‐induci‐ ble Hsp82 and constitutive Hsc82 (Figure 2C).

*Biomolecules* **2022**, *12*, x 5 of 13

mean and SD of three replicate experiments and compared to WT.

transformed with HA‐Apn2 plasmid. Transformants were grown to mid‐log phase at 30 °C. Lysate from these cells was analyzed by Western blotting with an anti‐HA and anti‐Ydj1 antibody. (**F**) Rel‐ ative abundance of Apn2‐HA was quantitated by taking the ratio of Apn2‐HA/PGK1. Data are the

**Figure 2.** Apn2 interacts with Hsc82 and Hsp82. (**A**) Yeast G170D and P82a cells expressing Apn2‐ HA were grown to mid‐log phase at 30 °C. Cells were stressed at 39 °C, and lysates from unstressed and heat shocked cells were analyzed for Apn2 levels using Western blot with anti‐HA antibodies. Pgk1 was used as a loading control. (**B**) Relative abundance of Apn2‐HA was quantitated by taking the ratio of Apn2‐HA/PGK. Data are the mean and SD of three replicate experiments, and further, unstressed cells were compared to heat shocked cells (\*\* *p* < 0.001). (**C**) Hsc82‐Glu and Hsp82‐His yeast cells were transformed with Apn2‐HA. Cells were grown to mid‐log phase at 30 °C. Lysate from these cells was analyzed by SDS‐PAGE and Western blotting using anti‐HA and yeast anti‐ Hsc82‐specific antibodies. Pgk1 was used as a loading control. Immunoprecipitation was performed using anti‐HA magnetic beads, and the interaction was studied. **Figure 2.** Apn2 interacts with Hsc82 and Hsp82. (**A**) Yeast G170D and P82a cells expressing Apn2-HA were grown to mid-log phase at 30 ◦C. Cells were stressed at 39 ◦C, and lysates from unstressed and heat shocked cells were analyzed for Apn2 levels using Western blot with anti-HA antibodies. Pgk1 was used as a loading control. (**B**) Relative abundance of Apn2-HA was quantitated by taking the ratio of Apn2-HA/PGK. Data are the mean and SD of three replicate experiments, and further, unstressed cells were compared to heat shocked cells (\*\* *p* < 0.001). (**C**) Hsc82-Glu and Hsp82-His yeast cells were transformed with Apn2-HA. Cells were grown to mid-log phase at 30 ◦C. Lysate from these cells was analyzed by SDS-PAGE and Western blotting using anti-HA and yeast anti-Hsc82 specific antibodies. Pgk1 was used as a loading control. Immunoprecipitation was performed using anti-HA magnetic beads, and the interaction was studied.

#### *3.3. Mammalian APE2 Interacts with the Hsp90–Hsp70 Chaperone System 3.3. Mammalian APE2 Interacts with the Hsp90–Hsp70 Chaperone System*

Mammalian APE2 plays a variety of roles in key cellular processes involving the response to a multitude of stressors, including DNA single- and double-strand breaks, base excision repair, and oxidative stress, leading to the activation of DDR complexes and pathways, including ATR and Chk1 [16,18]. The abundance of several DDR proteins, including APE2, decreased in bladder cancer cells treated with Hsp90 inhibitors [25]. To determine if there was a physical interaction between chaperones and APE2, we took a similar approach to that of Figure 1. HEK293 cells were grown to mid-confluence and were transfected with a construct expressing HA-APE2 (Table 2). After 48 h, cells were lysed, and APE2 complexes were purified using anti-HA magnetic beads. SDS-PAGE

analysis and Western blotting of APE2 complexes revealed the presence of Hsp70 and Hsp90, which were not observed in the immunoprecipitation from cells lacking HA-APE2 (Figure 3A). Despite the robust interaction of APE2 with the chaperones, the major DNAJA1 co-chaperone was not observed in the APE2 complex (Figure 3A). by SDS‐PAGE/Western blotting (Figure 3B). Consistent with our results in yeast, APE2 binding was observed between both the constitutive and heat‐inducible expressed HSPs in mammalian cells (Figure 3B).

There are a variety of Hsp70 family members expressed in mammalian cells. Alt‐ hough they are highly conserved, they vary in their client selectivity, cellular localization and expression pattern in tissues [11,12,32]. Our previous results suggested that APE2 interacts with HSPA8, the major constitutively expressed isoform of Hsp70 in cells. To determine whether APE2 might be able to bind other HSPA family members, we co‐trans‐ fected HEK293 cells with plasmids (Table 2) expressing HA‐APE2 and V5‐tagged HSPA family members that included inducible HSPA1A, HSPA1L and non‐inducible HSPA8. After 48 h, we purified HA‐APE2 from these cells and subjected the complex to analysis

Mammalian APE2 plays a variety of roles in key cellular processes involving the re‐ sponse to a multitude of stressors, including DNA single‐ and double‐strand breaks, base excision repair, and oxidative stress, leading to the activation of DDR complexes and path‐ ways, including ATR and Chk1 [16,18]. The abundance of several DDR proteins, including APE2, decreased in bladder cancer cells treated with Hsp90 inhibitors [25]. To determine if there was a physical interaction between chaperones and APE2, we took a similar ap‐ proach to that of Figure 1. HEK293 cells were grown to mid‐confluence and were trans‐ fected with a construct expressing HA‐APE2 (Table 2). After 48 h, cells were lysed, and APE2 complexes were purified using anti‐HA magnetic beads. SDS‐PAGE analysis and Western blotting of APE2 complexes revealed the presence of Hsp70 and Hsp90, which were not observed in the immunoprecipitation from cells lacking HA‐APE2 (Figure 3A). Despite the robust interaction of APE2 with the chaperones, the major DNAJA1 co‐chap‐

*Biomolecules* **2022**, *12*, x 6 of 13

erone was not observed in the APE2 complex (Figure 3A).

**Figure 3.** Mammalian APE2 interacts with the Hsp90–Hsp70 chaperone system. (**A**) HEK293 cells were grown to mid‐confluence and were transfected with a construct expressing HA‐APE2 from a constitutive CMV promoter. After 48 h, cells were lysed, and APE2 complexes were purified using anti‐HA‐magnetic beads. Lysates from these cells were analyzed by SDS‐PAGE and Western blot‐ ting using anti‐HA, anti‐Hsp70, anti‐DNAJA1 and anti‐Hsp90 specific antibodies. Beta‐actin was used as a loading control. Immunoprecipitation was performed using HA beads, and the interaction was studied. (**B**) HEK293 cells were co‐transfected with V5‐tagged Hsp70 and APE2‐HA. **Figure 3.** Mammalian APE2 interacts with the Hsp90–Hsp70 chaperone system. (**A**) HEK293 cells were grown to mid-confluence and were transfected with a construct expressing HA-APE2 from a constitutive CMV promoter. After 48 h, cells were lysed, and APE2 complexes were purified using anti-HA-magnetic beads. Lysates from these cells were analyzed by SDS-PAGE and Western blotting using anti-HA, anti-Hsp70, anti-DNAJA1 and anti-Hsp90 specific antibodies. Beta-actin was used as a loading control. Immunoprecipitation was performed using HA beads, and the interaction was studied. (**B**) HEK293 cells were co-transfected with V5-tagged Hsp70 and APE2-HA. Immunoprecipitation was performed using anti-HA-magnetic beads, and the interaction was studied using anti-V5 and anti-HA antibody.

There are a variety of Hsp70 family members expressed in mammalian cells. Although they are highly conserved, they vary in their client selectivity, cellular localization and expression pattern in tissues [11,12,32]. Our previous results suggested that APE2 interacts with HSPA8, the major constitutively expressed isoform of Hsp70 in cells. To determine whether APE2 might be able to bind other HSPA family members, we co-transfected HEK293 cells with plasmids (Table 2) expressing HA-APE2 and V5-tagged HSPA family members that included inducible HSPA1A, HSPA1L and non-inducible HSPA8. After 48 h, we purified HA-APE2 from these cells and subjected the complex to analysis by SDS-PAGE/Western blotting (Figure 3B). Consistent with our results in yeast, APE2 binding was observed between both the constitutive and heat-inducible expressed HSPs in mammalian cells (Figure 3B).

The stability of APE2 in epithelial cells is dependent on Hsp70 and Hsp90 function. Molecular chaperones regulate the folding, maturation and stability of their client proteins [33]. Our previous results implied that APE2 may be a bona fide client of the Hsp90–Hsp70 system. To examine this possibility, we assessed the impact of chaperone inhibition on APE2 abundance. HEK293 cells expressing HA-APE2 were treated with either an inhibitor of Hsp90 (ganetespib) or Hsp70 (JG-98). Cells were harvested at the indicated time points, and APE2 abundance was determined by Western blotting. HEK293

cells treated with ganetespib showed a decrease in APE2 abundance after only 8 h of treatment (Figure 4A). Even more impressive was the rapid decrease in APE2 levels after only 2 h of treatment of JG-98 (Figure 4B). We queried whether this dependence extended to other cancer cell lines including breast cancer (MCF-7) as well as androgen-dependent and androgen-independent prostate cancer (LNCaP and PC-3, respectively). As with our previous experiments, these cell lines were treated with ganetespib, and APE2 levels were assessed through Western blotting at 2 h intervals. In the case of PC-3, MCF7 and LNCaP, the APE2 levels significantly decreased after 2 h of treatment of JG-98 (Figure 5A–F). To similarly understand whether Hsp70 contributed toward APE2 stability, we treated MCF-7, LNCaP and PC-3 cells with the Hsp90 inhibitor and measured APE2 abundance via Western blotting. APE2 levels started to decline significantly after 2 h of treatment with maximum inhibition seen at 16 h (Figure 6A–F). ment of JG‐98 (Figure 4B). We queried whether this dependence extended to other cancer cell lines including breast cancer (MCF‐7) as well as androgen‐dependent and androgen‐ independent prostate cancer (LNCaP and PC‐3, respectively). As with our previous ex‐ periments, these cell lines were treated with ganetespib, and APE2 levels were assessed through Western blotting at 2 h intervals. In the case of PC‐3, MCF7 and LNCaP, the APE2 levels significantly decreased after 2 h of treatment of JG‐98 (Figure 5A–F). To similarly understand whether Hsp70 contributed toward APE2 stability, we treated MCF‐7, LNCaP and PC‐3 cells with the Hsp90 inhibitor and measured APE2 abundance via Western blot‐ ting. APE2 levels started to decline significantly after 2 h of treatment with maximum inhibition seen at 16 h (Figure 6A–F).

*Biomolecules* **2022**, *12*, x 7 of 13

ied using anti‐V5 and anti‐HA antibody.

Immunoprecipitation was performed using anti‐HA‐magnetic beads, and the interaction was stud‐

Molecular chaperones regulate the folding, maturation and stability of their client proteins [33]. Our previous results implied that APE2 may be a bona fide client of the Hsp90– Hsp70 system. To examine this possibility, we assessed the impact of chaperone inhibition on APE2 abundance. HEK293 cells expressing HA‐APE2 were treated with either an in‐ hibitor of Hsp90 (ganetespib) or Hsp70 (JG‐98). Cells were harvested at the indicated time points, and APE2 abundance was determined by Western blotting. HEK293 cells treated

4A). Even more impressive was the rapid decrease in APE2 levels after only 2 h of treat‐

The stability of APE2 in epithelial cells is dependent on Hsp70 and Hsp90 function.

**Figure 4.** Inhibition of Hsp90 or Hsp70 promote a rapid reduction in APE2 levels. (**A**) HEK293 cells expressing HA‐APE2 were treated with either an inhibitor of Hsp90 (ganetespib) or (**C**) Hsp70 (JG‐ 98). Cells were harvested at the indicated time points, and APE2 abundance was determined by SDS‐PAGE and Western blotting using anti‐HA antibody. Beta‐actin was used as a loading control. (**B**,**D**) The relative abundance of APE2‐HA was quantitated by taking the ratio of Apn2‐HA/Beta‐ actin from 3 replicate experiments and compared to untreated HEK293 cells. Data are the mean and **Figure 4.** Inhibition of Hsp90 or Hsp70 promote a rapid reduction in APE2 levels. (**A**) HEK293 cells expressing HA-APE2 were treated with either an inhibitor of Hsp90 (ganetespib) or (**C**) Hsp70 (JG-98). Cells were harvested at the indicated time points, and APE2 abundance was determined by SDS-PAGE and Western blotting using anti-HA antibody. Beta-actin was used as a loading control. (**B**,**D**) The relative abundance of APE2-HA was quantitated by taking the ratio of Apn2-HA/Betaactin from 3 replicate experiments and compared to untreated HEK293 cells. Data are the mean and SD of three replicate experiments and are compared to untreated. Statistical significance is indicated as (\*\* *p* < 0.001) (\* *p* < 0.05).

*Biomolecules* **2022**, *12*, x 8 of 13

as (\*\* *p* < 0.001) (\* *p* < 0.05).

as (\*\* *p* < 0.001) (\* *p* < 0.05).

**Figure 5.** Stability of APE2 in a range of cancer cell lines is dependent on Hsp90 and Hsp70 function. (**A**) PC3 (**C**) MCF7 and (**E**) LnCAP cells expressing HA‐APE2 were treated with an inhibitor of Hsp70 (JG‐98). Cells were harvested at the indicated time points, and APE2 abundance was deter‐ mined by SDS‐PAGE and Western blotting using anti‐HA antibody. Beta‐actin was used as a load‐ ing control. (**B**,**D**,**F**) The relative abundance of APE2‐HA was quantitated by taking the ratio of APE2‐HA/Beta‐actin. Data are the mean and SD of three replicate experiments and are compared to untreated (\*\* *p* < 0.001). **Figure 5.** Stability of APE2 in a range of cancer cell lines is dependent on Hsp90 and Hsp70 function. (**A**) PC3 (**C**) MCF7 and (**E**) LnCAP cells expressing HA-APE2 were treated with an inhibitor of Hsp70 (JG-98). Cells were harvested at the indicated time points, and APE2 abundance was determined by SDS-PAGE and Western blotting using anti-HA antibody. Beta-actin was used as a loading control. (**B**,**D**,**F**) The relative abundance of APE2-HA was quantitated by taking the ratio of APE2-HA/Betaactin. Data are the mean and SD of three replicate experiments and are compared to untreated (\*\* *p* < 0.001). **Figure 5.** Stability of APE2 in a range of cancer cell lines is dependent on Hsp90 and Hsp70 function. (**A**) PC3 (**C**) MCF7 and (**E**) LnCAP cells expressing HA‐APE2 were treated with an inhibitor of Hsp70 (JG‐98). Cells were harvested at the indicated time points, and APE2 abundance was deter‐ mined by SDS‐PAGE and Western blotting using anti‐HA antibody. Beta‐actin was used as a load‐ ing control. (**B**,**D**,**F**) The relative abundance of APE2‐HA was quantitated by taking the ratio of APE2‐HA/Beta‐actin. Data are the mean and SD of three replicate experiments and are compared to untreated (\*\* *p* < 0.001).

SD of three replicate experiments and are compared to untreated. Statistical significance is indicated

SD of three replicate experiments and are compared to untreated. Statistical significance is indicated

**Figure 6.** Stability of APE2 in a range of cancer cell lines is dependent on Hsp90 function. (**A**) PC3 (**C**) MCF7 and (**E**) LnCAP cells expressing HA‐APE2 were treated with an inhibitor of Hsp90 (gan‐ etespib). Cells were harvested at the indicated time points, and APE2 abundance was determined by SDS‐PAGE and Western blotting. Beta‐actin was used as a loading control. (**B**,**D**,**F**) The relative abundance of APE2‐HA was quantitated by taking the ratio of APE2‐HA/Beta‐actin. Data are the mean and SD of three replicate experiments and are compared to untreated (\*\* *p* < 0.001). **4. Discussion Figure 6.** Stability of APE2 in a range of cancer cell lines is dependent on Hsp90 function. (**A**) PC3 (**C**) MCF7 and (**E**) LnCAP cells expressing HA‐APE2 were treated with an inhibitor of Hsp90 (gan‐ etespib). Cells were harvested at the indicated time points, and APE2 abundance was determined by SDS‐PAGE and Western blotting. Beta‐actin was used as a loading control. (**B**,**D**,**F**) The relative abundance of APE2‐HA was quantitated by taking the ratio of APE2‐HA/Beta‐actin. Data are the mean and SD of three replicate experiments and are compared to untreated (\*\* *p* < 0.001). **4. Discussion Figure 6.** Stability of APE2 in a range of cancer cell lines is dependent on Hsp90 function. (**A**) PC3 (**C**) MCF7 and (**E**) LnCAP cells expressing HA-APE2 were treated with an inhibitor of Hsp90 (ganetespib). Cells were harvested at the indicated time points, and APE2 abundance was determined by SDS-PAGE and Western blotting. Beta-actin was used as a loading control. (**B**,**D**,**F**) The relative abundance of APE2-HA was quantitated by taking the ratio of APE2-HA/Betaactin. Data are the mean and SD of three replicate experiments and are compared to untreated (\*\* *p* < 0.001).

#### **4. Discussion**

The ability of cells to repair and maintain their genome is critical for their survival. The response to DNA damage is highly complex and relies on several different signaling cascades comprising multiple proteins [14,15]. The Hsp70–Hsp90 chaperone system binds and regulates several important proteins in this process, including APE1 and P53. Recent efforts in understanding the role of chaperones in DDR have included large-scale proteomics analysis, such as that of Li et al., which examined the abundance of proteins in 5637 bladder cancer cells after treatment with the Hsp90 inhibitors ganetespib (STA9090), or luminespib (AUY-922) [25]. In that study, over 800 proteins were downregulated, including XRCC1, XPC, RAD50, 53BP1 and notably, APE2 [25]. In this study, we have identified a role for the Hsp70 and Hsp90 chaperones in regulating the activity of the APN2/Ape2 exonuclease in yeast and mammalian cells.

#### *4.1. APE2 and Apn2 Display Binding Preferences for Chaperone and Co-Chaperone Paralogs*

An unresolved question in chaperone biology is why cells express many highly similar chaperone proteins. In yeast, the four Ssa proteins are highly conserved with over 80% similarity in amino acids sequence [4]. Ssa1 and Ssa2 represent the major cytosolic Hsp70s present under basal conditions, while Ssa3 and Ssa4 are highly heat induced. Several studies have suggested that these chaperone paralogs have overlapping but unique interactomes [34]. Recently, work using the model substrate ribonucleotide reductase (RNR) showed a clear preference for this client in binding Ssa1 and Ssa2 [35]. Although Apn2 binds cytosolic Hsp70 and Hsp90 paralogs equally, cells expressing Ssa1 as their sole cytosolic Ssa1 are unable to support WT levels of Ape2 as depicted by compromised stability in Figure 1B,C. The difference in Apn2 abundance in Ssa2 vs. Ssa1-expressing yeast is particularly interesting considering how similar the two proteins are. However, previous studies have shown that even a single divergent amino acid between Ssa1 and Ssa2 can produce differences in their ability to modulate prion propagation and protein degradation [36]. A recent study observed a parallel defect in septin levels in Ssa1-expressing yeast [37]. Future research, possibly involving a comparative interactome study of Ssa proteins, may shed light on this issue [34].

Cells express a variety of co-chaperones that are critical for stimulation of chaperone ATPase activity and for loading clients onto chaperones for folding [3,30,38]. We show here that Apn2 co-purifies with Ydj1, a major Hsp70 co-chaperone (Figure 1A). In contrast to ribonucleotide reductase whose stability depends on Ydj1 function, loss of Ydj1 does not impact Apn2 stability [28]. It is possible that Apn2 stability in yeast is additionally regulated by other semi-redundant co-chaperones such as Sis1, which has similar yet distinct roles in the cell as Ydj1 [39–42]. This may also explain why in our studies, DNAJA1 the mammalian homologue of Ydj1 does not appear to interact with APE2 (Figure 3A). Going forward, it would be interesting to identify and understand the major co-chaperones responsible for regulating APE2 and Apn2 function in mammalian and yeast cells, respectively.

#### *4.2. Novel Anticancer Strategies Based on Fine-Tuning Chaperone Function*

Molecular chaperones are required for the stability and activity of many proteins, including oncoproteins that are critical for cancer progression [43–46]. Recently, APE2 has been revealed to be an important player in regulating genome integrity and cancer progression [20,22,23,29,47,48]. Our study suggests that targeting APE2 activity through inhibition of chaperone function may be a viable anticancer therapy. While in vitro studies such as those presented here clearly show the value of manipulating chaperone function, studies in vivo suggest that complete abolishment of Hsp70 or Hsp90 results in severe toxicity for patients [25,49]. Several alternative approaches to bypass the toxicity issue are currently being pursued [49–51]. The first has been to identify key co-chaperones that regulate oncogenic clients and to develop drugs that inhibit them, such as 116-9e and C-86 [52–54]. While DNAJA1 is not observed in complex with APE2, it is possible that drugs such as 116-9e and C-86 may have a broad enough specificity to be target

regulatory co-chaperones of APE2 in cancer. An alternative method for fine-tuning of chaperones may be to manipulate their post-translational modifications (PTMs) [55–57]. Future studies examining the Hsp70/Hsp90-APE2 structure may allow for specific targeting of this interaction via small molecules that bind the interaction interface or alter critical PTMs required for chaperone–exonuclease interaction.


**Table 1.** Yeast strains used in this study.

#### **Table 2.** Plasmids used in this study.


Overall, this work identifies a new client of the Hsp70–Hsp90 axis, the Apn2/APE2 exonuclease. The rapid loss of APE2 in cancer cells upon inhibition of either Hsp90 or Hsp70 provides a path forward for novel therapies that jointly target chaperones and the DNA damage response.

**Author Contributions:** Conceptualization, A.W.T. and S.O.; methodology, A.W.T.; formal analysis, S.O., B.Z. and T.H.W.; investigation, S.O., B.Z. and T.H.W.; resources, A.W.T.; data curation, S.O.; writing—original draft preparation, S.O. and M.M.M.; writing—review and editing, A.W.T., S.O., T.H.W. and M.M.M.; funding acquisition, A.W.T. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the National Institutes of Health, grant numbers R01GM139885 and R15GM139059.

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

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** All data from this study can be found in the main manuscript.

**Acknowledgments:** We would like to thank Mehdi Mollapour, Daniel Durocher, and Nayun Kim for kindly providing materials for this study.

**Conflicts of Interest:** The authors declare no competing interests.

### **References**


## *Review* **The Role of Hsp90 in Retinal Proteostasis and Disease**

**Kalliopi Ziaka and Jacqueline van der Spuy \***

UCL Institute of Ophthalmology, 11-43 Bath Street, London EC1V 9EL, UK; kalliopi.ziaka.16@ucl.ac.uk **\*** Correspondence: j.spuy@ucl.ac.uk

**Abstract:** Photoreceptors are sensitive neuronal cells with great metabolic demands, as they are responsible for carrying out visual phototransduction, a complex and multistep process that requires the exquisite coordination of a large number of signalling protein components. Therefore, the viability of photoreceptors relies on mechanisms that ensure a well-balanced and functional proteome that maintains the protein homeostasis, or proteostasis, of the cell. This review explores how the different isoforms of Hsp90, including the cytosolic Hsp90α/β, the mitochondrial TRAP1, and the ER-specific GRP94, are involved in the different proteostatic mechanisms of photoreceptors, and elaborates on Hsp90 function when retinal homeostasis is disturbed. In addition, several studies have shown that chemical manipulation of Hsp90 has significant consequences, both in healthy and degenerating retinae, and this can be partially attributed to the fact that Hsp90 interacts with important photoreceptor-associated client proteins. Here, the interaction of Hsp90 with the retinaspecific client proteins PDE6 and GRK1 will be further discussed, providing additional insights for the role of Hsp90 in retinal disease.

**Keywords:** chaperone; co-chaperone; heat shock protein; Hsp90; inherited retinal disease; photoreceptor; phototransduction; proteostasis

#### **1. Phototransduction and Protein Folding in Photoreceptors (PR)**

Photoreceptor cells are highly specialized sensory neurons in the retina, and are essential for converting light into a neural signal, a fundamental process which initiates vision. In the mammalian retina, there are two types of photoreceptor cells, the rods and the cones. Both rods and cones are adjacent to the retinal pigment epithelium (RPE), a monolayer of pigmented cells which is vital for the normal function and survival of photoreceptors [1]. Morphologically, photoreceptors consist of a synaptic terminal, a nuclear region, and an inner segment (IS) and outer segment (OS) which are connected by a connecting cilium (CC). The OS of both cell types consists of closely spaced membranous discs containing photopigment molecules, called opsins, which are coupled to a lightabsorbing chromophore (retinal, an aldehyde of vitamin A). Opsins are responsible for tuning the absorption of light to a specific wavelength of the light spectrum. The rod OS contains the rod-specific photopigment rhodopsin, whereas the cone OS contains one of the three cone-opsins, S-opsin, M-opsin, or L-opsin. Rhodopsin, with a peak absorption (*λ*max) of ~500 nm, functions during dim light conditions allowing scotopic vision, whereas cone opsins are responsible for processing wavelengths ranging from ~350 to 560 nm, thus allowing colour vision [2]. Rods and cones share the same cellular mechanism of light detection, a process known as phototransduction.

Phototransduction is a complex mechanism in which light is converted into an electrical signal through the sequential activation of signalling proteins. In rods, phototransduction is activated by the photoisomerization of the rhodopsin-bound chromophore 11-*cis*retinal to all-*trans* retinal, inducing a conformational change in rhodopsin to its activated form metarhodopsin II. Metarhodopsin II stimulates the trimeric G-protein transducin by catalysing the exchange of GDP for GTP on the α-subunit. The GTP-associated α-subunit of transducin dissociates from the β and γ subunits and activates PDE6, a phosphodiesterase

**Citation:** Ziaka, K.; van der Spuy, J. The Role of Hsp90 in Retinal Proteostasis and Disease. *Biomolecules* **2022**, *12*, 978. https://doi.org/ 10.3390/biom12070978

Academic Editor: Yongzhang Luo

Received: 20 June 2022 Accepted: 8 July 2022 Published: 12 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/).

that hydrolyses cGMP. The decreased concentration of cGMP in the OS results in the closure of cGMP-gated channels in the plasma membrane, and the cessation of sodium and calcium influx, which, in turn, leads to the hyperpolarisation of the rod cell and the inhibition of glutamate release at the synaptic terminal [3]. A series of biochemical reactions is required for photoreceptors to return to their inactive state. This involves another network of proteins which restore the various activated components to their inactive state. G-protein-coupled receptor kinase 1 (GRK1) phosphorylates metarhodopsin II, inducing a conformational change that enables the binding of arrestin, leading to its inactivation. PDE6 is inactivated upon GTP hydrolysis of the transducin α-subunit, a process that is facilitated by the GTPase-activating protein (GAP) complex, consisting of RGS9 (regulator of G-proteinsignalling isoform 9) and G-protein β-subunit or G-protein β-subunit-like protein [4]. As a result, free cGMP concentration returns to normal levels (depolarised state) due to the activation of guanylyl cyclase activating proteins (GCAP) and the cGMP-synthesizing enzyme guanylate cyclase (GC). The phototransduction cascade in cone photoreceptors is similar to that in rods and is mediated by homologous phototransduction proteins [5,6]. However, while rods generate a detectable single photon response for maximal sensitivity in dim light conditions, cones are less sensitive than rods and require the simultaneous activation of tens to hundreds of opsin molecules in bright light conditions to generate a detectable response. The high spatial and temporal resolution of cone-mediated vision is made possible by the rapid kinetics of activation and inactivation, the trade-off of which is low amplification and sensitivity. In contrast, the trade-off of the high amplification gain of rods in dim light is their slow kinetics [5,6]. The adaption of rods and cones that shift their dynamic range towards dim and bright light detection respectively places a great metabolic demand on photoreceptors, as the visual cycle requires high amounts of energy for the phototransduction components to coordinate and function together. Moreover, constant triggering of phototransduction causes photooxidative stress to the OS components which need to be constantly replaced to avoid permanent damage. This is achieved by synthesis of new OS disks at the base of the OS and shedding of the OS tips which are phagocytosed by the RPE [7]. Interestingly, RPE cells have the highest phagocytic activity in the body, highlighting the intense metabolic demands of OS renewal [8,9]. To maintain this ability while performing their normal biological function, photoreceptors rely on high levels of protein synthesis, and the correct folding, assembly, trafficking, and degradation of various protein components. High levels of protein synthesis of the phototransduction components occur in the IS which must be continuously translocated through the connecting cilium to their site of action in the OS. The balance of these processes in the photoreceptor cell is called protein homeostasis or "proteostasis".

#### **2. The Importance of Hsp90 Isoforms in Retinal Proteostasis**

Proteostasis is maintained and controlled by an extensive network of molecular chaperones, proteolytic systems, and their regulators, termed the proteostasis network (PN). To ensure the correct folding and degradation of misfolded proteins, the PN includes sophisticated protein quality control (PQC) mechanisms, of which the chaperone Hsp90 is a vital component. Hsp90 is an ATP-dependent protein that is universally found in various cellular compartments, such as the cytosol, the ER, and the mitochondria. There are five Hsp90 members, which according to published guidelines for HSP nomenclature are categorised under the HSPC family and include the cytosolic HSPC1 (HSP90AA1), HSPC2 (HSP90AA2), HSPC3 (HSP90AB1), the endoplasmic reticulum (ER) HSPC4 (GRP94), and the mitochondrial HSPC5 (TRAP1) isoforms [10]. These isoforms participate in different PQC systems in the various compartments of the cell with a common aim to support the folding or refolding and stability of client proteins. Hsp90 functions as a dimer, with each protomer within the Hsp90 dimer comprising an N-terminal ATP-binding domain, a middle domain, and a constitutively dimerised C-terminal domain. Hsp90 chaperone activity is coupled with ATP hydrolysis, wherein ATP binds to an open conformation of Hsp90, which induces the transient dimerisation of the N-terminal domains and ATP hydrolysis with

subsequent release of the client protein (reviewed by [11,12]). The Hsp90 cycle facilitates the folding, maturation, or assembly of near-native client proteins, of which there are several hundred (https://www.picard.ch/downloads/Hsp90interactors.pdf accessed on 3 July 2022). Hsp90 co-chaperones interact non-covalently with Hsp90 to modulate the Hsp90 cycle or specifically target client proteins to Hsp90 (reviewed by [11,12]).

#### *2.1. Cytosolic Hsp90*

The cytosolic isoforms of Hsp90 participate in protein folding as a part of the heat shock response (HSR) and the Hsp90/Hsp70 protein folding machinery (Figure 1). The HSR is an orchestrated process that leads to the rapid transcription of selective genes encoding cytosolic molecular chaperones, also known as heat shock proteins (HSPs). The transcriptional activation of HSPs is regulated by transcription factors known as heat shock factors (HSFs) [13]. HSF1 is the key regulator of the HSR leading to HSP induction in response to stress. In the absence of stress, monomeric HSF1 is maintained in an inactive state by interaction with molecular chaperones in the cytosol, including Hsp90 [14]. In the presence of stress, HSF1 is converted from an inactive monomer to an active DNA-binding trimer, and this trimerization process involves the dissociation of Hsp90 and co-chaperones from its regulatory domain [15]. The trimerized HSF1 translocates to the nucleus and binds to heat shock elements (HSE) in the promoters of target genes that promote the transcription of Hsp90 and other HSPs [16].

In addition to its role in the HSR, cytosolic Hsp90 also functions as part of the Hsp90/Hsp70 protein folding machinery, in which Hsp90 targets client proteins, early folding intermediates in a near native state, and in concert with Hsp70, facilitates the thermodynamically favourable maturation of these clients [17]. In mammalian cells, including photoreceptors, there are two major cytosolic Hsp90 isoforms, the stress-inducible Hsp90α (HSPC1) and the constitutively expressed Hsp90β (HSPC3), which share 85% sequence identity [18] (Figure 1). The less abundant Hsp90α A2 (HSPC2) isoform is identical to Hsp90α with the exception of an N-terminal extension in Hsp90α. A recent study showed that Hsp90α deficiency in mice can cause rhodopsin retention in the IS and eventually lead to retinal degeneration. Further investigation revealed that microtubule-associated protein 1B (MAP1B), which is important for microtubule stabilization, was associated with Hsp90α and significantly reduced in Hsp90-deficient mice by proteasomal degradation. The authors suggested that Golgi organisation and vesicle transportation, which both rely on stable microtubules, are disrupted and this could be the underlying cause of photoreceptor degeneration [19].

#### *2.2. ER-Associated GRP94*

The ER has its own network of molecular chaperones which ensure that correctly folded proteins are produced and exit from the organelle for further processing. The glucose-regulated protein 94 (GRP94) (HSPC4) is a key regulator of the ER quality control mechanism and its residence in the ER is facilitated by its distinct C-terminal sequence KDEL, which serves as an ER retrieval signal for the KDEL receptor [20] (Figure 1). GRP94, together with BiP (GRP78), are two of the most abundant proteins in the ER [21] and play a significant role in regulating the ER unfolded protein response (UPR). The UPR involves the activation of a well synchronised set of signalling pathways directed by ER-resident transmembrane proteins that include inositol-requiring protein-1 (IRE1), the protein kinase RNA (PKR)-like ER kinase (PERK), and the activating transcription factor 6 (ATF6) [22] (Figure 1). During stress overload, the UPR branches respond by stimulating the expression of UPR-targeted genes which encode proteins, such as molecular chaperones, folding catalysts, subunits of the translocation machinery (Sec61 complex), ER-associated degradation (ERAD) molecules, and antioxidants. This activation leads to the upregulation of the protein folding and degradation capacity and the inhibition of protein synthesis in order to alleviate the stress and restore the equilibrium in the ER. Specifically in photoreceptors, GRP94 has been shown to be involved in opsin quality control as it forms a complex with

mutant opsins and other chaperones (BiP) [23]. Apart from protein folding, ER quality control involves ERAD, a mechanism to detect misfolded proteins and tag them for proteasomal degradation (Figure 1). Evidence from Christianson et al. (2008) [24] showed that GRP94 actively participates in ERAD, when α1-antitrypsin, an ERAD substrate, failed to degrade in GRP94-depleted cells. Misfolded rhodopsin is also subjected to ERAD [25,26], and it has been suggested that GRP94 and BiP might be involved in the recognition of the non-glycosylated ER-retained misfolded opsins [26]. Another important feature of GRP94 is its ability to bind calcium and maintain calcium homeostasis in the ER. The ER quality control machinery is coupled to the storage and utilization of calcium [27]. Most calcium in the ER is stored bound to proteins and GRP94 is one of the most important calcium-binding proteins [20]. A reduction in total calcium levels can strongly affect protein folding and change the molecular chaperone selection in the ER [28–30].

#### *2.3. Mitochondrial TRAP1*

The tumour necrosis factor receptor associated protein 1 (TRAP1) is the mitochondriaspecific Hsp90 isoform (HSPC5) and has distinct structural and functional properties (Figure 1). Structurally, TRAP1 is similar to the cytosolic Hsp90 isoforms, with the exception of a cleavable N-terminal mitochondrial localization signal and an N-terminal extension or 'strap' that provides stability in its 'closed' conformation [31]. Functionally, TRAP1 participates in the maintenance of mitochondrial integrity, protein folding and response to proteotoxic stress in mitochondria, and protection from oxidative stress damage [31,32]. Similarly to the ER, mitochondria are particularly vulnerable to the disturbance of proteostasis due to their high intrinsic protein folding demands. Hence, they have developed their own protective mechanism to overcome proteotoxic stress, known as the mitochondrial unfolded protein response (UPRmt) [33]. Similar to the HSR and ERspecific UPR, UPRmt is a transient transcriptional change in response to proteotoxic stress that is regulated by mitochondria. It was first described by Martinus et al. (1996) [34] as the transcriptional activation of mitochondria-specific chaperones and depends on mitochondrial-nuclear communication. The activation of UPRmt promotes protein folding, limits protein import, and reduces the translation of mitochondrial proteins. The UPRmt has been most extensively studied in *C. elegans* and there is only a small number of studies that have attempted to characterise it in mammalian cells [34–36]. Similarly to the ER-specific UPR, studies have shown that UPRmt elicits a multi-axis response that is regulated by several proteins and leads to distinct molecular outcomes (reviewed by [37]). However, the exact molecular mechanisms and regulators of the different UPRmt pathways remain largely unexplored, especially in disease. A study in *Drosophila* showed that modulation of TRAP1 expression led to the nuclear translocation of the transcription factor Dve which induced the expression of the mitochondrial chaperonin Hsp60, mitochondrial Hsp70, and a putative protease, CG5045, suggesting that TRAP1 is able to activate the UPRmt . The same study found that TRAP1 modulation could significantly improve health span, potentially by activation of the UPRmt [38]. In recent years, a functional interaction between ER and mitochondria during stress has been the focus of scientific interest. TRAP1 has been shown to have an important role in this ER–mitochondria interaction since it can potentially regulate the ER-associated UPR [39]. In photoreceptors, the elongated mitochondria extend almost the entire length of the IS, and are critical in meeting the high energy demands of photoreceptors for protein synthesis and phototransduction in the OS as well as serving as a calcium store. The photoreceptor TRAP1 is therefore likely to play a critical role in the maintenance of photoreceptor homeostasis and the UPRmt (Figure 1).

viewed by [11,12]).

*2.1. Cytosolic Hsp90* 

transcription of Hsp90 and other HSPs [16].

**Figure 1.** The roles of Hsp90 in photoreceptor proteostasis. (**A**) PDE6 and GRK1 are important components of phototransduction activation and deactivation, respectively. PDE6 and GRK1 are synthesised in the photoreceptor inner segment and translocated to the outer segment via the connecting cilium. Both PDE6 and GRK1 are Hsp90 client proteins. (**B**) The mitochondrial isoform TRAP1 is involved in protein folding and the mitochondrial unfolded protein response (UPRmt). (**C**) The cytosolic isoforms Hsp90α/Hsp90β participate in protein folding in association with the Hsp70 folding machinery, and as a part of the heat-shock response (HSR) by regulating the activation of heat-shock transcription factor 1 (HSF1). During stress, Hsp90 together with other chaperones dissociate from HSF1, which then trimerizes and is activated via phosphorylation. Activated HSF1 translocates to the nucleus and stimulates the expression of molecular chaperones. (**D**) ER-associated GRP94 is important in protein folding and ER protein quality control mechanisms, the UPR, and ER-associated degradation (ERAD). The UPR involves three signalling pathways mediated via ER transmembrane protein folding sensors, the inositol-requiring protein-1 (IRE1), the protein kinase RNA (PKR)-like ER kinase (PERK), and the activating transcription factor 6 (ATF6). Activation of the UPR branches leads to the increased expression of proteins, such as molecular chaperones, folding catalysts, subunits of the translocation machinery (Sec61 complex), ERAD molecules, and antioxidants. Created with BioRender.com.

middle domain, and a constitutively dimerised C-terminal domain. Hsp90 chaperone activity is coupled with ATP hydrolysis, wherein ATP binds to an open conformation of Hsp90, which induces the transient dimerisation of the N-terminal domains and ATP hydrolysis with subsequent release of the client protein (reviewed by [11,12]). The Hsp90 cycle facilitates the folding, maturation, or assembly of near-native client proteins, of which there are several hundred (https://www.picard.ch/downloads/Hsp90interactors.pdf accessed on 3 July 2022). Hsp90 co-chaperones interact non-covalently with Hsp90 to modulate the Hsp90 cycle or specifically target client proteins to Hsp90 (re-

The cytosolic isoforms of Hsp90 participate in protein folding as a part of the heat shock response (HSR) and the Hsp90/Hsp70 protein folding machinery (Figure 1). The HSR is an orchestrated process that leads to the rapid transcription of selective genes encoding cytosolic molecular chaperones, also known as heat shock proteins (HSPs). The transcriptional activation of HSPs is regulated by transcription factors known as heat shock factors (HSFs) [13]. HSF1 is the key regulator of the HSR leading to HSP induction in response to stress. In the absence of stress, monomeric HSF1 is maintained in an inactive state by interaction with molecular chaperones in the cytosol, including Hsp90 [14]. In the presence of stress, HSF1 is converted from an inactive monomer to an active DNA-binding trimer, and this trimerization process involves the dissociation of Hsp90 and co-chaperones from its regulatory domain [15]. The trimerized HSF1 translocates to the nucleus and binds to heat shock elements (HSE) in the promoters of target genes that promote the

#### **3. The Role of Hsp90 in Retinal Disease**

#### *3.1. Hsp90 and the Stress Response in Retinal Disease*

Evidently, Hsp90 is of high importance in the retina because of the many vital roles it has in the different PQC mechanisms of proteostasis. Its importance in retinal homeostasis can be further highlighted by exploring its role in retinal disease paradigms. Currently, 280 genes (316 genes and loci) are associated with inherited retinal degeneration (IRD) (https://sph.uth.edu/retnet/home.htm, accessed on 3 July 2022). The inheritance of IRDs

can be autosomal recessive, autosomal dominant, or X-linked, and IRDs can furthermore be progressive or stationary, as well as non-syndromic or part of a wider syndrome. Nonsyndromic retinal dystrophies are further classified as macular dystrophies, cone and cone– rod dystrophies, rod–cone dystrophies, or chorioretinopathies [40]. Retinitis pigmentosa (RP) describes a group of retinal degenerative rod–cone dystrophies that are primarily characterized by the loss of rod photoreceptors, as well as the subsequent degeneration of cones. Mutations in rhodopsin are the most common cause of autosomal dominant retinitis pigmentosa (adRP) [2]. The photoreceptor stress machinery has been found to be induced in various models of rhodopsin misfolding. For example, the upregulation of the UPR and of the HSR, including increased levels of Hsp90 and Hsp70, has been observed in the P23H-1 transgenic rat, in which mutant rhodopsin Pro23His (P23H) is misfolded and retained in the ER [41]. Upon treatment with arimoclomol, an HSR co-inducer, Hsp90 and Hsp70 levels were further elevated, and this was associated with decreased rhodopsin aggregation, photoreceptor rescue, and improved visual responses [41]. The role of Hsp90 in the upregulation of the UPR and HSR is likely to be important in other retinal diseases caused by protein misfolding, in which the maintenance and regulation of the vast array of structural and functional proteins is critical for normal photoreceptor homeostasis.

#### *3.2. Hsp90 Inhibition in Retinal Disease*

Additional evidence of the significance of Hsp90 in retinal dystrophies, including RP, arise from the consequences of its pharmacological inhibition. Hsp90 inhibition can elicit a dual effect, leading to the proteasome-mediated degradation of its client proteins or the disruption of the chaperone complex with HSF1 and the activation of the HSR, leading to the upregulation of molecular chaperones. Therefore, the potential of Hsp90 inhibitors to manipulate the photoreceptor stress machinery has been explored in various studies. A list of Hsp90 inhibitors that have been used in models of retinal disease is summarized in Table 1. The Hsp90 inhibitor 17-*N*-allylamino-17-demethoxygeldanamycin (17-AAG), also known as tanepsimycin, has been shown to protect against rhodopsin aggregation and toxicity in a cell model of P23H rhodopsin [42]. Two other Hsp90 inhibitors, geldanamycin (GA) and radicicol, also showed a similar effect on alleviating the toxic gain-of-function mechanisms of P23H rhodopsin in vitro, although this effect was less potent compared to 17-AAG [42]. The amelioration of P23H rhodopsin aggregation was not observed in mouse embryonic fibroblasts from HSF-1 knock-out mice, suggesting that the protection depends on HSF1 and the activation of the HSR [43]. In accordance with these findings, the systemic administration of 2-amino-7,8-dihydro-6H-pyrido[4,3- D]pyrimidin-5-one NVPHSP990 (HSP990), a blood brain barrier permeable Hsp90 inhibitor, activated HSF-1 and induced the upregulation of molecular chaperones in the retina of P23H transgenic rats [43]. This HSP990-mediated stimulation of the stress machinery was associated with reduced rhodopsin aggregation and mislocalisation, improved visual function, and photoreceptor survival several weeks after a single drug dose [43]. Hsp90 inhibition has also been reported to be protective in another form of adRP, RP10, which is caused by mutations in the inosine 50 -monophosphate dehydrogenase type 1 (IMPDH1) gene [44]. Systemic delivery of 17-AAG, facilitated by the RNA interference-mediated modulation of the inner blood–retina barrier, protected against photoreceptor degeneration in the Arg224Pro (R224P) mutant IMPDH mouse model, by promoting the expression of HSPs, including Hsp90, which, in turn, reduced the formation of IMPDH aggregates [45]. These studies highlight the potential neuroprotective effects of Hsp90 inhibition in retinal protein misfolding disorders via upregulation of the HSR.

It has also been shown, however, that prolonged inhibition of Hsp90 in the retina may also play a detrimental role in photoreceptor proteostasis as a consequence of the degradation of key Hsp90 client proteins. The rhodopsin mutant Arg135Leu (R135L) is hyperphosphorylated and constitutively bound to arrestin, thereby disrupting vesicular traffic in photoreceptors. 17-AAG enhanced the vectorial transport of R135L rhodopsin to the OS by suppressing the endocytosis defect that characterises this mutation, thereby

restoring R135L rhodopsin localization to the WT phenotype in rat retinae [43]. In an in vitro cell model of R135L rhodopsin, Hsp90 inhibition by 17-AAG similarly blocked the recruitment of arrestin to R135L rhodopsin and led to a reduction in the aberrant endocytosis of R135L rhodopsin [43]. Interestingly, this effect was HSF1-independent as 17-AAG rescued the intracellular accumulation of R135L rhodopsin and restored the cytosolic localization of arrestin in HSF-1 knock-out mouse embryonic fibroblasts [43]. It was hypothesized that Hsp90 inhibition may instead mediate its effect on R135L rhodopsin by client-mediated degradation. Indeed, prolonged Hsp90 inhibition with HSP990 in vivo led to a post-translational reduction in GRK1 and PDE6 protein levels, identifying them as Hsp90 clients. Hsp90 inhibition in cells led to the rapid proteasomal degradation of newly synthesised GRK1 confirming a requirement for Hsp90 for GRK1 maturation and function. The effect of Hsp90 inhibition on R135L rhodopsin was therefore attributed to the fact that GRK1 was identified as an Hsp90 client protein, and Hsp90 inhibition decreased GRK1 levels resulting in reduced R135L phosphorylation and subsequently, reduced arrestin binding.


**Table 1.** List of Hsp90 inhibitors used in models of retinal degeneration.

#### *3.3. Ocular Toxicities in Clinical Trials of Hsp90 Inhibition*

These findings have important implications for the pharmacological manipulation of molecular chaperones as a therapeutic approach for retinal disease. Despite the plethora of evidence that Hsp90 inhibition can provide protection in the diseased retina as described above, reports from clinical trials in oncology highlight ocular toxicities that have emerged as an important clinical concern (Table 2). Hsp90 N-terminal inhibitors, including ansamycin derivatives (17-dimethylaminoethylamino-17-demethoxygeldanamycin (17- DMAG)), resorcinol derivatives (AT13387, AUY922), and benzamide derivatives (SNX-5422 (PF-04929113)), have been associated with visual disturbances, such as blurred vision, photopsia, night blindness, photophobia, and retinopathy [47–57]. In addition, some preclinical studies have reported that severe retinal degeneration occurred in rats and

beagle dogs after treatment with Hsp90 inhibitors [46,58,59].The oral administration of the Hsp90 scaffold N-terminal inhibitor CH5164840 led to a loss of pupillary light reflex, abnormal electroretinographic (ERG) responses, and histological changes in the photoreceptor outer nuclear layer, including photoreceptor degeneration, in beagle dogs [58]. Similarly, intravenous administration of 17-DMAG or AUY922 promoted photoreceptor cell death in Sprague Dawley (SD) rats in addition to the upregulation of the HSR [46], and AUY92-induced abnormal ERG responses, and photoreceptor OS disorganization in Brown Norway and Wistar rats [59]. However, while ocular effects have been widely reported in preclinical and clinical studies of certain Hsp90 inhibitors, visual disturbances have not been reported for all Hsp90 inhibitors, including 17-AAG and the resorcinol derivative ganetespid. Preclinical studies comparing the ocular toxicity of 17-DMAG, 17-AAG, AUY922, and ganetespid suggest that the extent of ocular toxicity correlates with the retinal biocompatibility and clearance rate of the compound, with high levels of accumulation and prolonged inhibition of Hsp90 in the retina, leading to photoreceptor cell death [46]. Therefore, the inhibition of Hsp90 as a therapeutic approach in the retina is clearly a doubleedged sword, whereby Hsp90 inhibition can both induce a neuroprotective response but also lead to ocular toxicity upon prolonged retinal accumulation. The mechanism of retinal toxicity as a consequence of Hsp90 inhibition is poorly understood; however, a possible explanation is that the ocular toxicity observed upon prolonged Hsp90 inhibition might be mediated by the disruption caused to important Hsp90 client proteins in the retina. As described previously, GRK1 biosynthesis requires Hsp90, and prolonged Hsp90 inhibition via systemic administration of HSP990 reduced GRK1 and PDE6 levels post-translationally, suggesting that the Hsp90 client list includes important components of the phototransduction cascade. More recently, Transient Receptor Potential cation channel subfamily M member 1 (TRPM1) has been identified as another potential Hsp90 client protein in the retina [57]. TRPM1 is a constitutively open calcium entry channel primarily expressed in skin melanocytes and retinal ON-bipolar cells in the inner nuclear layer. The treatment of mice with AUY922 resulted in increased apoptosis in the photoreceptor outer nuclear layer, disorganization of the photoreceptor outer segments, disruption of RPE cells, and a dose-dependent decrease in TRPM1 via disruption of the interaction with Hsp90 [57].

**Table 2.** List of Hsp90 inhibitors used in clinical trials in oncology and their ocular effects. \* The observed ocular effects were transient or resolved after treatment discontinuation.



#### **4. Hsp90 Client Proteins in the Retina**

Whilst PDE6, GRK1, and TRPM1 have been identified as Hsp90 client proteins in the retina, only PDE6 and GRK1 are specifically expressed in photoreceptor cells and are important components of the phototransduction cascade. An in-depth understanding of the mechanisms underlying the specific recruitment of PDE6 and GRK1 to Hsp90 is crucial to understand the biogenesis of these important phototransduction proteins, not only in the healthy retina but also in retinal diseases associated with these Hsp90 clients, and the review will henceforth focus on mechanistic and structural insights into PDE6 and GRK1 as Hsp90 client proteins. Whilst the Hsp90-PDE6 chaperone complex has been investigated in depth, less is known regarding GRK1 as a specific client for Hsp90 and the role of this association in disease.

#### *4.1. The Hsp90-PDE6 Chaperone Complex*

PDE6, a member of the class I family of phosphodiesterases [60], is a heterotetrametric complex, which in rod photoreceptors, comprises the catalytic PDE6α and PDE6β subunits together with two inhibitory PDE6γ subunits. Cone PDE6 comprises two catalytic PDE6α' subunits and two inhibitory PDE6γ' subunits. In the phototransduction cascade, activated transducin relieves the inhibition of the PDE6 catalytic subunits imposed by the inhibitory subunits, leading to cGMP hydrolysis. Mutations in rod PDE6α, PDE6β, and PDE6γ cause autosomal recessive RP and mutations in PDE6β can also cause autosomal dominant congenital stationary night blindness (https://sph.uth.edu/retnet/, accessed on 3 July 2022). In contrast, mutations in cone PDE6α' and PDE6γ' are associated with autosomal recessive cone or cone–rod dystrophy, or achromatopsia (https://sph.uth.edu/retnet/, accessed on 3 July 2022). Interestingly, whilst mutations in the PDE6 subunits cause relatively milder forms of inherited retinal degeneration, mutations in the reported cochaperone for PDE6, the photoreceptor-specific aryl hydrocarbon receptor interacting protein-like 1 (AIPL1), cause Leber congenital amaurosis (LCA), a severe early onset and rapidly progressive disease leading to photoreceptor degeneration and the loss of vision within the first few years of life [61]. An early observation in *Aipl1* knockout and knockdown mice was the post-transcriptional loss of all three subunits of rod PDE6 prior to the onset of retinal degeneration [62,63]. Cone PDE6 levels were also substantially reduced in cone photoreceptors lacking AIPL1 [64]. In the absence of AIPL1, the rod PDE6

subunits were stably synthesized but subsequently misassembled and targeted for rapid proteasomal degradation [65]. Similarly, whilst the loss of AIPL1 had no effect on the synthesis of the cone PDE6 subunits, the translated subunits were unstable and could not assemble into the holoenzyme [66]. These studies confirmed that AIPL1 is important for the post-translational stability and assembly of both rod and cone PDE6. *Biomolecules* **2022**, *12*, x FOR PEER REVIEW 10 of 20

#### 4.1.1. AIPL1 Structure 4.1.1. AIPL1 Structure

AIPL1 was first identified as a possible Hsp90 co-chaperone due to its homology to the Hsp90 tetratricopeptide repeat (TPR) domain co-chaperone aryl hydrocarbon receptor interacting protein (AIP), with which it shares 49% identity and 69% similarity [61]. AIPL1 and AIP comprise a C-terminal TPR domain and a N-terminal FK506 binding protein (FKBP)-like domain, similar to larger members of the FKBP family of immunophilins, including the Hsp90 TPR domain co-chaperones FKBP51 and FKBP52. The AIPL1 TPR domain consists of three consecutive TPR motifs, and the crystal structure of the human AIPL1 TPR domain revealed that, similar to other Hsp90 TPR domain co-chaperones, the AIPL1 TPR domain adopts a typical TPR fold [67] (Figure 2). Each TPR motif consists of a pair of anti-parallel α-helices such that the consecutive TPR motifs form a series of six antiparallel α-helices connected by short loops followed by a seventh α-helix, which all together forms a right-handed amphipathic groove. The AIPL1 FKBP-like domain shares the typical FKBP fold comprising a five stranded β sheet forming a half β-barrel surrounding a short α helix and creating a hydrophobic cavity (Figure 2). However, unlike other members of the FKBP family, the FKBP-like domain of AIP and AIPL1 lack peptidyl prolyl isomerase activity and cannot bind immunosuppressant drugs [68,69]. Moreover, the FKBP-like domain of both AIP and AIPL1 uniquely include an extensive insert region linking the last two β strands in the FKBP-like domain [69–71]. In AIP, the insert regions consist of a 19 residue long helical segment followed by a mostly random coil structure and an αhelix [69]. In contrast, the crystal structure of the human AIPL1 FKBP-like domain revealed that the insert region in AIPL1 (residues 90–146) is well structured and comprises three consecutive α-helices (α2, α3 and α4) connected by short loops [71] (Figure 2). Additional differences between the AIP and AIPL1 FKBP-like domains include the absence of an N-terminal α-helix in AIPL1 that is thought to structurally stabilize the AIP FKBP-like fold; and a loop between β4 and α1 that adopts a 'looped-out' conformation in AIPL1 but a 'looped-in' conformation in AIP, wherein a critical hinge residue Trp72 is either flipped in or out, respectively, thus modulating access to a hydrophobic cavity [71] (Figure 2). Finally, the AIPL1 TPR domain is followed by a C-terminal 56 amino acid proline rich domain (PRD), an unstructured random coil that is imperfectly conserved in primates and absent in non-primates [61,72,73] (Figure 2). AIPL1 was first identified as a possible Hsp90 co-chaperone due to its homology to the Hsp90 tetratricopeptide repeat (TPR) domain co-chaperone aryl hydrocarbon receptor interacting protein (AIP), with which it shares 49% identity and 69% similarity [61]. AIPL1 and AIP comprise a C-terminal TPR domain and a N-terminal FK506 binding protein (FKBP)-like domain, similar to larger members of the FKBP family of immunophilins, including the Hsp90 TPR domain co-chaperones FKBP51 and FKBP52. The AIPL1 TPR domain consists of three consecutive TPR motifs, and the crystal structure of the human AIPL1 TPR domain revealed that, similar to other Hsp90 TPR domain co-chaperones, the AIPL1 TPR domain adopts a typical TPR fold [67] (Figure 2). Each TPR motif consists of a pair of anti-parallel α-helices such that the consecutive TPR motifs form a series of six antiparallel α-helices connected by short loops followed by a seventh α-helix, which all together forms a right-handed amphipathic groove. The AIPL1 FKBP-like domain shares the typical FKBP fold comprising a five stranded β sheet forming a half β-barrel surrounding a short α helix and creating a hydrophobic cavity (Figure 2). However, unlike other members of the FKBP family, the FKBP-like domain of AIP and AIPL1 lack peptidyl prolyl isomerase activity and cannot bind immunosuppressant drugs [68,69]. Moreover, the FKBP-like domain of both AIP and AIPL1 uniquely include an extensive insert region linking the last two β strands in the FKBP-like domain [69–71]. In AIP, the insert regions consist of a 19 residue long helical segment followed by a mostly random coil structure and an α-helix [69]. In contrast, the crystal structure of the human AIPL1 FKBP-like domain revealed that the insert region in AIPL1 (residues 90–146) is well structured and comprises three consecutive α-helices (α2, α3 and α4) connected by short loops [71] (Figure 2). Additional differences between the AIP and AIPL1 FKBP-like domains include the absence of an N-terminal α-helix in AIPL1 that is thought to structurally stabilize the AIP FKBP-like fold; and a loop between β4 and α1 that adopts a 'looped-out' conformation in AIPL1 but a 'looped-in' conformation in AIP, wherein a critical hinge residue Trp72 is either flipped in or out, respectively, thus modulating access to a hydrophobic cavity [71] (Figure 2). Finally, the AIPL1 TPR domain is followed by a C-terminal 56 amino acid proline rich domain (PRD), an unstructured random coil that is imperfectly conserved in primates and absent in non-primates [61,72,73] (Figure 2).

**Figure 2.** Model of AIPL1. The TPR domain (PDB 6PX0, cyan) and the FKBP-like domain (PDB 5U9A, magenta, yellow, and green) of AIPL1 were superimposed onto FKBP51 (1KT0). The PRD (brown) of AIPL1 was modelled using the I-TASSER server [74,75]. Helices 2, 3, and 4 of the unique **Figure 2.** Model of AIPL1. The TPR domain (PDB 6PX0, cyan) and the FKBP-like domain (PDB 5U9A, magenta, yellow, and green) of AIPL1 were superimposed onto FKBP51 (1KT0). The PRD (brown) of

insert region of the AIPL1 FKBP51-like domain (yellow) and the loop between β4 and α1 that adopts a 'looped-out' conformation in AIPL1 (green) are shown. The PRD is required for the intrinsic chaperone activity of AIPL1. The TPR domain mediates the interaction with Hsp90/Hsp70 and the PDE6 inhibitory subunits, whilst Hsp90 may also make contact with the α3 helix in the unique insert re-

of C. Prodromou, Genome Damage and Stability Centre, University of Sussex.

AIPL1 was modelled using the I-TASSER server [74,75]. Helices 2, 3, and 4 of the unique insert region of the AIPL1 FKBP51-like domain (yellow) and the loop between β4 and α1 that adopts a 'looped-out' conformation in AIPL1 (green) are shown. The PRD is required for the intrinsic chaperone activity of AIPL1. The TPR domain mediates the interaction with Hsp90/Hsp70 and the PDE6 inhibitory subunits, whilst Hsp90 may also make contact with the α3 helix in the unique insert region. The FKBP-like domain constitutes a ligand-binding site for isoprenyl groups. Figure courtesy of C. Prodromou, Genome Damage and Stability Centre, University of Sussex.

#### 4.1.2. The Interaction of AIPL1 with Hsp90

Hidalgo-de-Quintana et al. (2008) first provided experimental evidence for the TPRmediated interaction of full length human AIPL1 with both Hsp90 and Hsp70, with preferential binding to Hsp90 [76]. The TPR consensus residues required for the packing of adjacent α-helices in the TPR motifs and residues involved in tight electrostatic interactions with the C-terminal EEVD TPR acceptor sites of Hsp90 and Hsp70 are conserved in AIPL1. The deletion of the Hsp90 MEEVD pentapeptide or the Hsp70 TIEEVD heptapeptide significantly reduced the interaction of AIPL1 with Hsp90 and Hsp70, respectively, and the MEEVD peptide competitively reduced the interaction of AIPL1 with Hsp90 in quantitative binding assays [76,77]. Moreover, the mutation of lysine 265 to alanine (K265A), a carboxylate clamp residue critical for the tight electrostatic interaction of TPR domain co-chaperones with the C-terminal EEVD motif, significantly reduced the interaction of AIPL1 with Hsp90 and Hsp70 [76,78]. The AIPL1 TPR domain alone can interact with Hsp90 in the absence of the FKBP-like domain, and the disruption of the TPR domain by LCA-associated missense mutations, deletions, insertions, duplications, or C-terminal truncations significantly reduced or abolished the interaction with Hsp90 [77,79]. Therefore, the TPR domain is critical for the interaction of AIPL1 with Hsp90 and features directing the prototypical core TPR domain co-chaperone–chaperone interaction are conserved in AIPL1. Accordingly, Sacristan-Reviriego et al. (2017) showed that human AIPL1 preferentially interacts with Hsp90 in the nucleotide-bound closed conformation and that this interaction is reduced by both apyrase treatment or HSP990 inhibition, indicating that productive Hsp90 ATPase cycles are required for efficient AIPL1 interaction [77]. Moreover, AIPL1 stabilized rod PDE6α to proteasomal degradation in the cytosol and this function was significantly reduced by Hsp90 inhibition with HSP990, GA, or 17-AAG [77]. Similarly, biolayer interferometry (BLI) binding assays recently reported the preferential binding of mouse AIPL1 to adenylyl-imidodiphosphate (AMP-PNP)-bound Hsp90 in a 1:2 stoichiometry [78]. In this study, DMAG treatment significantly impacted the ability of AIPL1 to chaperone cone PDE6α' activity. Altogether, the data point to the preferential interaction of AIPL1 with Hsp90 in the closed conformation and the importance of a functional AIPL1-Hsp90 interaction for PDE6 stability and activity.

In addition to the role of the core TPR domain contacts in mediating the interaction of AIPL1 with the chaperone TPR acceptor site, additional requirements for this interaction have been investigated. In the case of FKBP51 and FKBP52, a region C-terminal to the TPR domain comprising a seventh α-helical extension (α-helix 7) mediates differential binding to Hsp90 and the truncation of FKBP51 and FKBP52 within this α-helical extension at Asn404 and Asn406, respectively, largely abrogated Hsp90 interaction [80]. Interestingly, the removal of the α-helical extension C-terminal to the core TPR domain of human AIPL1 by truncation at the topologically equivalent residue, Glu317 (AIPL1 1-317), reduced but did not abolish the interaction of AIPL1 with Hsp90 and Hsp70 [76]. Similarly, the truncation of the 12 C-terminal residues of mouse AIPL1 (AIPL1 1-316) did not abrogate the binding of AIPL1 to Hsp90 but moderately reduced the affinity for Hsp90 in BLI assays [78]. Notably, this region was however critical for the ability of AIPL1 to chaperone PDE6 in an in vitro heterologous assay for cone PDE6α' function. This suggests that in addition to the core TPR domain contacts, residues within the TPR α-helical extension may be important for functional chaperone complex assembly, although several residues thought to mediate

contact of the α-helical extension of FKBP51 with Hsp90 are missing or not conserved in mouse or human AIPL1.

Other regions implicated in the interaction of AIPL1 with Hsp90 include the primatespecific PRD and the α3 helix in the unique insert region of the FKBP-like domain. The deletion of the PRD, whilst having no effect on the structure or thermostability of AIPL1 [68,73], was reported to modestly increase the interaction of AIPL1 with Hsp90 in surface plasmon resonance (SPR) spectroscopy assays. On the other hand, the interaction of AIPL1 with Hsp90 following deletion of the PRD was reported to be comparable to that of full length AIPL1, although a significantly increased interaction was observed with the TPR domain alone in the absence of the PRD [79]. Finally, disease-associated mutations in the AIPL1 PRD had no effect on the interaction with Hsp90 [79]. Whilst the PRD may therefore not play a significant role in the interaction with Hsp90, it does appear to play a critical role in the intrinsic chaperone activity of AIPL1 [68]. AIPL1 was first shown to efficiently suppress the formation of intracellular inclusions comprising misfolded fragments of the AIPL1 interacting partner, NUB1, in a concentration-dependent manner [81]. AIPL1 also suppressed the thermal aggregation of citrate synthase (CS) and protected CS from thermal inactivation, and this effect was lost upon the deletion of the PRD [68]. The AIPL1 suppression of aggregation of the NUB1 fragments was not dependent on Hsp90, as GA had no effect in this assay, but was additive with Hsp70 dependent on AIPL1 C-terminal sequences [76]. Overall, the data suggest that the PRD is critical for AIPL1 intrinsic chaperone activity in association with Hsp70.

A number of studies have also investigated the contribution of the AIPL1 FKBP-like domain to Hsp90 interaction. The AIPL1 FKBP-like domain and TPR domain expressed alone can each fold stably to acquire the native conformation [67,70,71,82]. It has been reported that the AIPL1 FKBP-like domain alone, however, cannot interact with Hsp90 in the absence of the TPR domain [77]. Indeed, the LCA-associated patient mutation, Glu163Stop, which leads to the loss of the entire TPR domain and PRD, completely abolished the interaction of AIPL1 with Hsp90, confirming the critical role of the TPR domain in Hsp90 interaction [77]. However, patient-associated mutations in the FKBP-like domain, including missense mutations and in-frame deletions, diminished the interaction of AIPL1 with Hsp90 and impacted rod PDE6 activity in an indirect assay of cGMP hydrolysis [77,79], suggesting that whilst the FKBP-like domain alone cannot bind Hsp90, it is important for stable ternary chaperone complex formation with full length AIPL1. Interestingly, a very weak but transient interaction of Hsp90 with the N-terminal FKBP-like domain of AIP has been reported, and this interaction was reduced by the deletion of the FKBP-like domain unique insert region [69]. Similarly, the replacement of the α3 helix in the AIPL1 FKBP-like unique insert region with five glycine residues modestly affected the interaction with Hsp90, but critically impacted the activity of cone PDE6α' [78]. A model of the Hsp90-AIPL1 complex based on the cryo-EM structure of the Hsp90-FKBP51 complex placed the α3 helix of the insert region in close proximity to Hsp90, suggesting a moderate contribution of the insert region to the AIPL1–Hsp90 interface [78]. As the TPR acceptor site of Hsp90 can competitively bind a multitude of TPR domain co-chaperones, it has been suggested that contacts with the α3 helix may contribute to the specificity of the interaction of AIPL1 with Hsp90.

#### 4.1.3. The AIPL1-Mediated Targeting of PDE6 to Hsp90

The binding interface between the PDE6 client and Hsp90 has not been investigated. However, several mechanisms have been proposed wherein AIPL1 could specifically target PDE6 to Hsp90. Ramamurthy et al. (2003) first reported that AIPL1 could interact with and facilitate the processing of farnesylated proteins [83]. Notably, the PDE6 catalytic subunits are isoprenylated at the cysteine residue of their C-terminal CAAX box, with the identity of the CAAX box C-terminal residue suggesting that rod PDE6α is farnesylated whilst rod PDE6β and cone PDE6α' are geranylgeranylated. A general role for AIPL1 in protein farnesylation was suggested, since several interactors in a Y2H screen were farnesylated,

and mutation of the CAAX box cysteine to induce the loss of farnesylation or promote geranylgeranylation led to the loss of these interactions [83]. Accordingly, AIPL1 was found to interact with rod PDE6α in the mouse retina, and the AIPL1 interaction with PDE6β was reported to be dependent on that with PDE6α [65]. FRET assays with an AMCA conjugated farnesylated cysteine probe, S-farnesyl-L-cysteine methyl ester, revealed a high affinity interaction with the purified FKBP-like but not the TPR domain [70]. Mutation of Cys89 or Leu147 flanking the unique FBKP insert region or the deletion of the insert region (residues 96-143) abolished the interaction with the probe. Similarly, FRET assays confirmed a potent interaction of AIPL1 with an AMCA-conjugated peptide mimic of the PDE6α C-terminus with the cysteine residue modified by S-farnesylation and carboxymethylation [73]. Interestingly, competition assays with an excess of N-acetyl-S-geranylgeranyl-L-cysteine suggested for the first time that AIPL1 may also bind geranylgeranyl. Indeed, the crystal structure of the AIPL1 FKBP-like domain (residues 2-161) in the apo state and in the presence of either S-farnesyl-L-cysteine methyl ester or geranylgeranyl pyrophosphate confirmed the interaction of AIPL1 with these isoprenoid moieties that bind deep within the hydrophobic cavity [71]. There were no significant differences between the apo and isoprenoid-bound structures, with isoprenoid binding inducing only minor conformational changes in the ligand binding domain. Molecular dynamics simulations supported a model wherein the β4-α1 loop adopts a 'looped-in' conformation in the apo structure with the critical Trp72 residue thus occluding the hydrophobic ligand-binding pocket, which then rotates to the 'flipped-out' conformation upon isoprenoid binding [71]. The α2 side chains of the insert region were found to contribute significantly to isoprenoid binding [71], explaining the previous observation that the deletion of the insert region abrogated interaction with a farnesyl probe [70]. Moreover, the mutation of residues in the β4-α1 loop also markedly attenuated isoprenoid binding [71]. These studies thus confirmed the direct interaction of the AIPL1 FKBP-like domain with either farnesyl or geranylgeranyl, suggesting that AIPL1 might specifically target PDE6 to Hsp90 through these interactions. Notably, the mutation of the PDE6α' CAAX-box cysteine to favour either farnesylation or geranylgeranylation had no impact on the ability of AIPL1 to chaperone functional cone PDE6 in in vitro heterologous activity assays, suggesting that the role of AIPL1 is indiscriminate with respect to the identity of the isoprenoid moiety [84]. More recently, a PDE6α Cys857Ser knockin mouse model has been generated that abrogates the farnesylation of rod PDE6α [78]. Interestingly, the levels and targeting of PDE6α and PDE6β to the photoreceptor OS, as well as both the basal and maximal PDE6 activity, were comparable to control mice, in addition to which there was no change in either ERG or optical coherence tomography (OCT) measurements. Moreover, the deletion of the C-terminal 28 residues of cone PDE6α', including the CAAX motif or the loss of isoprenylation by Cys855Ser mutation, had no effect in in vitro assays of AIPL1 chaperoned PDE6 activity. Finally, steric occlusion of the AIPL1 prenyl binding site in an Ile61Phe/Ile151Phe double mutant had no effect on the ability of AIPL1 to chaperone PDE6α' or the Cys855Ser mutant in the heterologous assay. Hence, overall, whilst it is clear that the interaction of the AIPL1 FKBP-like domain with the PDE6 isoprenoid moieties contributes to the formation of the ternary chaperone complex, the exact role of this interaction in PDE6 biogenesis remains unclear. It is noteworthy that whilst the FKBP-like domain of AIPL1 appears to bind either farnesyl or geranylgeranyl groups indiscriminately, only PDE6 is affected in the *Aipl1* knockout and knockdown mouse models (in addition to soluble retinal guanylate cyclase in *Aipl1* knockout cones), despite the wide range of phototransduction components that are isoprenylated, thus suggesting that features other than the interaction of AIPL1 with the PDE6 isoprenoid groups must facilitate the specific recruitment of PDE6 to Hsp90 by AIPL1.

One such possibility is the interaction of AIPL1 with the inhibitory subunits of rod and cone PDE6. The rod PDE6γ subunit was reported to interact with AIPL1 using FRET assays [73]. Subsequently, a conserved C-terminal peptide of rod PDE6γ and cone PDE6γ' was shown to bind the AIPL1 TPR domain but not the AIPL1 FKBP-like domain with association and dissociation kinetics consistent with a 1:1 binding model [67]. Moreover, molecular modelling suggested that the C-terminal 25 residues of the rod and cone inhibitory subunits encompass most, if not all, of the contact with the TPR domain overlapping with the Hsp90 binding site, such that the inhibitory subunits and Hsp90 bind in a mutually exclusive manner [67]. This suggests a model in which the interaction of the PDE6γ/PDE6γ' subunits with the AIPL1 TPR domain impart specificity toward the PDE6 client, as the AIPL1 FKBP-like domain can bind isoprenyl moieties indiscriminately and the Hsp90 TPR acceptor site is bound competitively by TPR domain co-chaperones. This has important implications for modelling the role of AIPL1 and Hsp90 in PDE6 biogenesis in retinal photoreceptors, though it is noteworthy that AIPL1 failed to interact with either rod PDE6γ or cone PDE6γ' in co-immunoprecipitation assays of AIPL1 from mouse retinal explants [65,66].

Overall, these studies highlight the structural and mechanistic basis of PDE6 recruitment to Hsp90 via the PDE6-specific Hsp90 co-chaperone AIPL1. Misfolded PDE6 subunits that cause autosomal recessive retinal disease likely undergo unproductive folding cycles with Hsp90, leading to their post-translational degradation and loss of function.

#### *4.2. The Hsp90-GRK1 Chaperone Complex*

In comparison to the Hsp90-PDE6 chaperone complex, the interaction and maturation of GRK1 with Hsp90 as a client protein in retinal photoreceptors is poorly characterised. Mutations in GRK1 are associated with the Oguchi subtype of recessive congenital stationary night blindness (https://sph.uth.edu/retnet/, accessed on 3 July 2022). In the retina, GRK1 is expressed in rod photoreceptors whilst GRK7 is expressed in cone photoreceptors. Both GRK1 and GRK7 are members of the G protein-coupled receptor kinase (GRK) family, serine/threonine-specific protein kinases that mediate the agonist-dependent phosphorylation of G protein-coupled receptors.

GRK1 and GRK7 specifically target the activated state of the G protein-coupled receptors rhodopsin and cone opsin, respectively, and play a key role in the deactivation of the phototransduction cascade and photorecovery after light onset. GRK1 and GRK7 are tethered to the photoreceptor OS phospholipid membranes in close proximity to their substrate by C-terminal isoprenylation. In the dark, GRK1 and GRK7 are bound to and inhibited by recoverin. The activation of the phototransduction cascade leads to the release of calcium from recoverin, which induces a conformational change involving the rotation of a 'myristoyl switch' that results in the calcium-dependent dissociation of recoverin from the membrane and release from GRK, thus enabling GRK to phosphorylate rod and cone opsin [85]. The GRK-mediated phosphorylation of the opsins induces a conformational change in the receptor that in turn allows the binding of arrestin and receptor deactivation through sterically blocking the binding of transducin, effectively switching off the cascade [85].

The GRK family, in addition to the visual kinase subfamily (GRK1, GRK7), includes the β-adrenergic receptor (β-AR) kinase subfamily (GRK2, GRK3) and the GRK4 subfamily (GRK4, GRK5, GRK6). All GRK family members are composed of a short highly conserved ~16 residue N-terminal element unique to this family of kinases [86]. This is followed by a regulator of G protein signalling homology domain (RH) that is interrupted by a highly conserved serine/threonine kinase domain. The catalytic domain of the GRK family, including that of GRK1 and GRK7, has a highly conserved architecture comprising a small N-terminal lobe and a large C-terminal lobe connected by a flexible hinge region that forms a deep nucleotide binding cleft between them, followed by a C-terminal extension (C-tail). The N-terminal lobe comprises a five-stranded β-sheet with a conserved αC helix, whereas the C-terminal lobe comprises six α-helices. A loop within the C-tail forms an active site tether that contributes to the ATP binding site. The substrate mainly interacts with the surface of the C-terminal lobe. The RH domain folds into a bi-lobed helical bundle that bridges the small and large kinase domains [86]. Kinase activation involves a conformational change in which the N-terminal lobe moves towards C-terminal lobe

to form a closed state, with the αC-β4 loop thought to act as a hinge point for inter-lobe movement, enabling the rotation between open and closed conformations [87].

Hsp90 has been shown to play a role in the maturation and stabilization of the GRK family members GRK1, GRK2, GRK3, GRK5, and GRK6, with the inhibition of Hsp90 by GA or 17-AAG, leading to rapid proteasomal degradation of the newly synthesized GRKs [43,88]. Similar to GRK1, cone GRK7 is also likely to be an Hsp90 client protein. Hsp90 is known to bind directly to the kinase catalytic domain with Hsp90 binding determinants widely distributed in both lobes. The kinase catalytic domain is one of the most abundant structures in the human proteome, present in more than 500 protein kinases, and has a highly conserved architecture common to serine/threonine and tyrosine kinases. The kinase domain is thus regarded as a universal acceptor site mediating kinase interaction with Hsp90, and it is highly likely, therefore, that the kinase domain of GRK1 and GRK7 similarly directs the interaction with Hsp90. Considerable effort has been invested in identifying the specific features that mediate the recognition of client kinases by Hsp90 [89–94]. Interestingly, no global sequence determinants have been identified for the interaction of kinases with Hsp90, despite the high level of conservation in the kinase domain. Instead, a consensus has emerged in which the intrinsic stability of the kinase domain is an important determinant for Hsp90 interaction. Hsp90 kinase clients were reported to be more thermodynamically unstable than non-clients, with the small-molecule stabilization of the kinase domain reducing the client interaction and mutation of the kinase domain, leading to stronger Hsp90 client binding [89,90,93,94].

Finally, it is well known that cell division cycle 37 (Cdc37) is a ubiquitous kinasededicated co-chaperone that is universally employed to direct kinase clients to Hsp90, thereby providing selective recognition of the kinase family [93,94]. Cdc37 has been shown to directly bind the kinase catalytic site, overlapping with the Hsp90 binding. Cdc37 binds to kinase clients in the absence of Hsp90, whereas Hsp90 interacts only weakly without Cdc37. Therefore, Hsp90 and Cdc37 are thought to act in concert in chaperoning client kinases with Hsp90-mediated maturation of kinases strictly dependent on the Cdc37 dependent recruitment of the kinase to Hsp90 [93,94]. Whilst not experimentally tested, it is highly likely that the co-chaperone for GRK1 and GRK7 is Cdc37. Experimental investigation of GRK1 and GRK7 interaction with Hsp90 and Cdc73 will provide further insights and evidence that the features directing client kinase assembly with Cdc73 and Hsp90 are conserved amongst the visual GRKs.

#### **5. Conclusions**

In summary, retinal photoreceptors are amongst the most metabolically active cells in the human body. Consequently, high levels of reactive oxygen species accumulate in the photoreceptors, leading to membrane and protein damage. Approximately 10% of the photoreceptor outer segments are turned over daily to replace damaged membranes and proteins. There is therefore an extremely high demand on protein synthesis and turnover in retinal photoreceptor cells requiring high levels of proteostasis. There are currently 280 genes associated with inherited retinal disease, many of which code for proteins of the visual cycle and phototransduction cascade that require high levels of protein synthesis in the photoreceptor inner segment and translocation to the outer segment. Protein quality control is therefore of vital importance in photoreceptor cells and many inherited retinal diseases are protein misfolding disorders. Hsp90 is centrally important to protein folding and quality control in the retinal photoreceptors, including in the cytosol, ER, and in the mitochondria. Indeed, the induction of the HSR via short-term Hsp90 inhibition has been shown to be neuroprotective in in vitro and in vivo models of inherited retinal disease. However, longer term inhibition of Hsp90 in the retina may in fact be detrimental due to the resultant degradation of specific Hsp90 client proteins in the photoreceptors, including PDE6 and GRK1. Mutations in these Hsp90 client proteins themselves lead to retinal disease. Whilst the precise role of Hsp90 in the folding, maturation, or assembly of these retina-specific client proteins is not fully elucidated, this raises the possibility that small

molecule manipulation of the Hsp90 cycle to promote the favourable maturation of these client proteins may be a potential therapeutic approach for diseases associated with these clients. In addition, the induction of the HSR in the absence of Hsp90 inhibition might be another favourable avenue for treating protein misfolding disorders in the retina.

**Author Contributions:** Conceptualization, K.Z. and J.v.d.S.; writing—original draft preparation, K.Z. and J.v.d.S.; writing—review and editing, K.Z. and J.v.d.S.; funding acquisition, J.v.d.S. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Medical Research Council, UK; grant number [MR/P02582X/1].

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

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

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

#### **Abbreviations**

17-DMAG, 17-Dimethylaminoethylamino-17-demethoxygeldanamycin; 17-AAG, 17-*N*-allylamino-17-demethoxygeldanamycin; ATF6, activating transcription factor 6; AMP-PNP, adenylylimidodiphosphate; AIP, aryl hydrocarbon receptor interacting protein; AIPL1, aryl hydrocarbon receptor interacting protein-like 1; adRP, autosomal dominant retinitis pigmentosa; Cdc37, cell division cycle 37; CS, citrate synthase; CC, connecting cilium; ER, Endoplasmic reticulum; ERAD, ER-associated degradation; FKBP, FK506 binding protein; GRK1, G-protein-coupled receptor kinase 1; GA, geldanamycin; GRP94, glucose-regulated protein 94; GC, guanylate cyclase; GCAP, guanylyl cyclase activating proteins; HSE, heat shock element; HSF, heat shock factor; HSP, heat shock protein; HSR, heat shock response; HSP990, 2-amino-7,8-dihydro-6H-pyrido[4,3-D]pyrimidin-5-one NVPHSP990; IS, inner segment; IMPDH1, inosine 50 -monophosphate dehydrogenase type 1; IRE1, inositol-requiring protein-1; LCA, Leber congenital amaurosis; MAP1B, microtubule-associated protein 1B; UPRmt, mitochondrial unfolded protein response; OCT, optical coherence tomography; OS, outer segment; PRD, proline rich domain; PERK, protein kinase RNA (PKR)-like ER kinase; PQC, protein quality control; PDE6, retinal phosphodiesterase; RPE, retinal pigment epithelium; RP, Retinitis pigmentosa; RGS9, regulator of G-protein-signalling isoform 9; SPR, surface plasmon resonance; TPR, tetratricopeptide repeat; TRAP1, tumor necrosis factor receptor associated protein 1; UPR, unfolded protein response.

#### **References**

