Polypharmacology-based kinome screen identifies new regulators of KSHV reactivation

Kaposi’s sarcoma-associated herpesvirus (KSHV) causes several human diseases including Kaposi’s sarcoma (KS), a leading cause of cancer in Africa and in patients with AIDS. KS tumor cells harbor KSHV predominantly in a latent form, while typically <5% contain lytic replicating virus. Because both latent and lytic stages likely contribute to cancer initiation and progression, continued dissection of host regulators of this biological switch will provide insights into fundamental pathways controlling the KSHV life cycle and related disease pathogenesis. Several cellular protein kinases have been reported to promote or restrict KSHV reactivation, but our knowledge of these signaling mediators and pathways is incomplete. We employed a polypharmacology-based kinome screen to identifiy specific kinases that regulate KSHV reactivation. Those identified by the screen and validated by knockdown experiments included several kinases that enhance lytic reactivation: ERBB2 (HER2 or neu), ERBB3 (HER3), ERBB4 (HER4), MKNK2 (MNK2), ITK, TEC, and DSTYK (RIPK5). Conversely, ERBB1 (EGFR1 or HER1), MKNK1 (MNK1) and FRK (PTK5) were found to promote the maintenance of latency. Mechanistic characterization of ERBB2 pro-lytic functions revealed a signaling connection between ERBB2 and the activation of CREB1, a transcription factor that drives KSHV lytic gene expression. These studies provided a proof-of-principle application of a polypharmacology-based kinome screen for the study of KSHV reactivation and enabled the discovery of both kinase inhibitors and specific kinases that regulate the KSHV latent-to-lytic replication switch.


Abstract 10
Kaposi's sarcoma-associated herpesvirus (KSHV) causes several human diseases including 11 Kaposi's sarcoma (KS), a leading cause of cancer in Africa and in patients with AIDS. KS tumor 12 cells harbor KSHV predominantly in a latent form, while typically <5% contain lytic replicating 13 virus. Because both latent and lytic stages likely contribute to cancer initiation and progression, 14 continued dissection of host regulators of this biological switch will provide insights into 15 fundamental pathways controlling the KSHV life cycle and related disease pathogenesis. Several 16 cellular protein kinases have been reported to promote or restrict KSHV reactivation, but our 17 knowledge of these signaling mediators and pathways is incomplete. We employed a 18 polypharmacology-based kinome screen to identifiy specific kinases that regulate KSHV 19 reactivation. Those identified by the screen and validated by knockdown experiments included 20 several kinases that enhance lytic reactivation: ERBB2 (HER2 or neu), ERBB3 (HER3), ERBB4 21 (HER4), MKNK2 (MNK2), ITK, TEC, and DSTYK (RIPK5). Conversely, ERBB1 (EGFR1 or 22 HER1), MKNK1 (MNK1) and FRK (PTK5) were found to promote the maintenance of latency. 23 Mechanistic characterization of ERBB2 pro-lytic functions revealed a signaling connection 24 between ERBB2 and the activation of CREB1, a transcription factor that drives KSHV lytic gene 25 expression. These studies provided a proof-of-principle application of a polypharmacology-based 26 kinome screen for the study of KSHV reactivation and enabled the discovery of both kinase 27 inhibitors and specific kinases that regulate the KSHV latent-to-lytic replication switch. 28

Author Summary 29
Kaposi's sarcoma-associated herpesvirus (KSHV) causes Kaposi's sarcoma, a cancer 30 particularly prevalent in Africa. In cancer cells, the virus persists in a quiescent form called latency, 31 in which only a few viral genes are made. Periodically, the virus switches into an active replicative 32 cycle in which most of the viral genes are made and new virus is produced. What controls the 33 switch from latency to active replication is not well understood, but cellular kinases, enzymes that 34 control many cellular processes, have been implicated. Using a cell culture model of KSHV 35 reactivation along with an innovative screening method that probes the effects of many cellular 36 kinases simultaneously, we identified drugs that significantly limit KSHV reactivation, as well as 37 specific kinases that either enhance or restrict KSHV replicative cycle. Among these were the 38 ERBB kinases which are known to regulate growth of cancer cells. Understanding how these and 39 other kinases contribute to the switch leading to production of more infectious virus helps us 40 understand the mediators and mechanisms of KSHV diseases. Additionally, because kinase 41 inhibitors are proving to be effective for treating other diseases including some cancers, identifying 42 ones that restrict KSHV replicative cycle may lead to new approaches to treating KSHV-related 43 diseases. 44 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted February 1, 2023.

Introduction 45
Kaposi's sarcoma-associated herpesvirus (KSHV) is the etiologic agent of Kaposi's sarcoma 46 (KS), a leading cause of cancer in Africa and a substantial health concern for AIDS patients 47 worldwide [1][2][3]. KSHV causes three other less prevalent diseases: primary effusion lymphoma in 48 B cells, multicentric Castleman disease, and a KSHV inflammatory cytokine syndrome. The main 49 proliferating tumor cell of KS is the spindle cell. In KS spindle cells, the virus exists predominantly 50 in a latent state in which only a few of its ~90 genes are expressed. Approximately 5% of spindle 51 cells express markers of the lytic replicative cycle [4][5][6][7] representing a relatively infrequent switch 52 from latency to lytic replication in tumors. While considerable effort has been devoted to 53 characterizing cellular and viral factors that support the maintenance of latency or the induction 54 of lytic replication, our knowledge of the complex signaling involved in this replicative switch 55 remains incomplete. Because many latent and lytic KSHV genes have oncogenic properties and 56 are involved in disease progression [4,[8][9][10][11][12][13], understanding the regulators of this replicative 57 switch is of fundamental importance for understanding KSHV disease pathogenesis and has 58 potential relevance for new therapeutic interventions.. 59 The human kinome comprises 518 protein kinases known to regulate myriad host and viral 60 processes, including KSHV latency and the switch to lytic replication [14][15][16]. Due to the essential 61 regulatory roles of kinases, dysregulation of their catalytic activity causes many types of cancers 62 and other diseases. Viruses also usurp cellular kinases or encode their own kinases to modulate 63 the signaling of the host cell to promote specific virus lifecycle stages or replicative functions. For 64 KSHV, both the virus-encoded kinase, ORF36 [17], and cellular kinases are necessary for lytic 65 replication [18][19][20][21][22]. Prior reports identified several kinases with roles in KSHV reactivation using 66 various screening approaches, including a kinase cDNA overexpression screen [23], phospho-site 67 antibody microarray [18], and proteome analysis [24] following KSHV primary infection or after 68 induction of lytic replication. From these and other studies, fewer than a dozen kinases have been 69 validated as aiding latency or facilitating reactivation [18,24,25]. By completing a more 70 comprehensive investigation of protein kinase regulators of KSHV reactivation, we might identify 71 FDA-approved kinase inhibitors that could be repurposed with the aim of reducing KSHV reservoirs 72 and/or treating KSHV-associated cancers and lymphoproliferative diseases. 73 Recently, kinase centric polypharmacology-based screens have been developed to identify both 74 kinase inhibitors and their targeted kinases that regulate cell death, cancer cell migration and 75 other cancer cell phenotypes [26][27][28]. These screens have also been used in Plasmodium 76 infected cells and to evaluate kinase roles during virus-induced cytokine production [29,30]. This 77 innovative approach employs broadly acting kinase inhibitors as tools that exploit built-in 78 redundancy from their shared kinase targets, when used in Kinase Inhibitor Regularization (KiR) 79 analyses [26,27,31]. Specifically, the polypharmacology-based kinome screen and KiR platform 80 uses a small set of computationally-derived kinase inhibitors to restrict the catalytic activity of 81 multiple endogenously expressed kinases. Data derived from testing the phenotypic effects of 82 these inhibitors, coupled with known drug specificities and potencies for each kinase target, allows 83 for a network-based, machine-learning analysis that initially predicts the impact of untested kinase 84 inhibitors. Refinements by iterative screening of additional kinase inhibitors curates with high 85 accuracy, single kinases predicted to have significant regulatory potential for the system 86 evaluated. This method is attractive compared to alternative kinome screening methods due to 87 the high-throughput nature, built-in redundancy for enhanced accuracy, and two screen outputs, 88 kinase inhibitors and specific kinases. 89 Herein, we describe the adaption of this kinome screening approach to study KSHV reactivation 90 in an epithelial cell system commonly used to study KSHV reactivation. From this approach, we 91 discovered two drugs, lestaurtinib and K252a, as potent kinase inhibitors of induced lytic 92 replication and an initial six kinases not previously associated with reactivation. Among these 93 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted February 1, 2023. ; https://doi.org/10.1101/2023.02.01.526589 doi: bioRxiv preprint

Generation of a recombinant KSHV with infection and lytic replication indicators. 112
To enable precise measurement of the initial transition from KSHV latent-to-lytic replication, we 113 constructed a new recombinant virus called lytic replication indicator KSHV (KSHV LRI ). This virus is 114 derived from the KSHV bacterial artificial chromosome 16 (BAC16) that expresses GFP 115 constitutively, enabling identification of infected cells. For detecting lytic replication, we introduced 116 an expression cassette containing the KSHV polyadenylated long non-coding RNA promoter 117 (PrPAN) driving expression of a streptavidin-binding peptide fused to a truncated low-affinity nerve 118 growth factor receptor (SBP-ΔLNGFR) and nuclear localized mCherry (mCherry-NLS) ( Fig 1A). The 119 coding regions for these two genes are separated by a 2A self-cleaving peptide (P2A) sequence. 120 The SBP-ΔLNGFR was designed for selection by streptavidin binding of cells undergoing lytic 121 replication. The NLS on mCherry enables accurate, high-throughput quantification of individual 122 cells containing lytic replicating virus by fluorescence imaging. We introduced KSHV LRI into iSLK 123 cells, a human renal carcinoma epithelial cell line that encodes doxycycline (DOX)-inducible 124 KSHV replication and transcription activator (RTA). As expected, the KSHV LRI infected cells 125 expressed GFP ( Fig 1B) and LANA (Figs S1A and S1B), and addition of lytic inducing agents, 126 DOX or DOX plus sodium butyrate (NaB), resulted in mCherry-NLS ( Fig 1B) and SBP-ΔLNGFR 127 (Figs S1C and S1D) expression. Importantly, the engineered fluorescent constructs allowed us to 128 monitor total KHSV LRI infected cells by GFP area per image and virus reactivation by mCherry-129 NLS positive cells per image in infected iSLK cells using high-throughput, quantitative 130 fluorescence imaging. 131 Polypharmacology-based kinome screen for enhancers or repressors of KSHV 132 reactivation. 133 To identify kinases important for KSHV reactivation, we employed a polypharmacology-based 134 kinome screen using the kinase inhibitor regularization (KiR) pipeline as previously described [31]. 135 First, we incubated KSHV LRI latently infected iSLK cells with the vehicle (DMSO) control or each 136 of the 29 computationally-selected kinase inhibitors (KI 1-29) at four concentrations (31nM, 137 125nM, 500nM, 2µM) and concurrently induced lytic replication by adding DOX plus NaB to the 138 cell media (Fig 2A). Collectively, these 29 KIs targeted a broad range of kinases (296 out of 360 139 kinase profiled). At 72h post-treatment, we quantified KSHV reactivation by counting mCherry-140 NLS positive cells per image and cell survival by measuring GFP area per image using the 141 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted February 1, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023 Incucyte Imaging System. A representative dose-response curve for two kinase inhibitors (cmpd 142 #1: CAS no. 252003-65-9; cmpd #2: CAS no. 1221153-14-5) that either enhanced or restricted 143 KSHV reactivation at 500nM concentration is shown in Fig 2B. We found that seven of the KIs 144 were toxic to cells under these conditions and these were removed from further analyses. Data 145 from the remaining 22 KIs at 500nM with DOX plus NaB provided the initial "training set" for 146 machine learning-based analyses [33] with a range of 6 -30% total cells reactivated. For the 147 DMSO control, 20% total cells were reactivated and set to 100 ( Fig 2C). While there were changes 148 in the induction of lytic replication following treatment with only the KIs (without DOX or NaB), the 149 range of lytic replication (0 -0.06% total cells) was too small for the KiR analysis to generate 150 kinase inhibitor predictions ( Fig S2). These data suggest that broad kinase inhibition alone cannot 151 efficiently activate the switch from latency to lytic replication in this system, but following RTA 152 expression and release of some epigenetic constraints by NaB, kinases do measurably regulate 153 KSHV reactivation. 154 For the lytic inducing condition (DOX + NaB) that had a greater dynamic range for KSHV 155 reactivation, we produced an initial KiR model from the 22 KI "training set" phenotype dataset and 156 a separate dataset of > 400 KI's effects on 298 recombinant human protein kinases [33]. Leave-157 one-out-cross validation (LOOCV) was used to evaluate the accuracy of the model. We then 158 tested an additional 12 KIs predicted by the KiR model to impact KSHV reactivation ( Fig 2D, Table  159 S1). The resulting responses were iteratively included in the training set to improve the prediction 160 accuracy of the model. Based on the final KiR model, the correlation between predicted and 161 observed response was >0.8. The final list of 428 predicted kinase inhibitors is shown in Fig 2E  162 and Table S2. Of note, two broadly acting kinase inhibitors, lestaurtinib and K252a, were predicted 163 to be regulators of reactivation based on the initial KiR model. Following testing of these drugs 164 ( Fig 2D) and inclusion of the KSHV reactivation data into the final model, these two drugs 165 remained at the top of the list of kinase inhibitors predicted to regulate KSHV reactivation (Table  166 S2). 167

Kinases validated as cellular regulators of KSHV reactivation in epithelial cells. 168
Based on the KiR model, we selected 13 kinases with the highest average coefficient across all 169 alpha values (Table S3) which corresponds to likely regulators of KSHV reactivation. We 170 evaluated these predictions using siRNAs targeting each of these kinases. Individual depletion of 171 six of these kinases significantly altered KSHV reactivation under lytic inducing conditions without 172 having significant effects on cell viability (Fig 3). Knocking down ERBB4 (HER4), MKNK2 (MNK2), 173 and DSTYK (RIPK5) as well as two TEC family kinases, ITK and TEC, reduced reactivation, 174 indicating that these kinases are pro-lytic factors. In contrast, knocking down FRK (RAK or PTK5) 175 caused a slight but statistically significant increase in reactivation, suggesting that it may be a pro-176 latency factor. 177 Our finding that knocking down ITK reduced reactivation (Fig 3)  ERBB signaling cascades are complex [38,39]. Therefore, we measured activation of a panel of 218 substrates ( Fig 6A) using a high-throughput, reverse-phase protein array approach under both 219 latent and induced lytic replication conditions in control and ERBB2 depleted cells. Depletion of 220 ERBB2 in latently infected cells significantly attenuated the activation of ERBB1, as measured by 221 phosphorylation at Tyr 1173 ( Fig 6B). This result suggests that in latent cells, ERBB1:ERBB2 222 heterodimers form and ERBB2 transphosphroylates ERBB1 to activate downstream signaling.

223
Treatment with lytic inducing agents (in control siRNA cells) attenuated ERBB1 phosphorylation 224 to a similar extent as in latently infected cells with ERBB2 depletion. ERBB2 knockdown in cells 225 treated with lytic inducing agents did not further reduce ERBB1 activation. These results suggest 226 that the ERBB1:ERBB2 heterodimer is disrupted during induction of lytic replication and are 227 consistent with our finding that ERBB1 is a pro-latent factor that restricts lytic gene expression 228 (Fig. 5). 229 In our panel of substrates, we also evaluated phosphorylation of proteins involved in parallel 230 signaling or crosstalk with EBBB2, as well as downstream signaling intermediates, and substrates 231 of the selected signaling intermediates (Fig 6A). For proteins involved in crosstalk signaling, we 232 measured the activation of plasma membrane receptors MET [40][41][42] and PDGFRβ [43] during 233 lytic reactivation in control and ERBB2 knockdown cells. Similar to ERBB1, we found that MET 234 activation decreased during lytic induction as compared to latent infection ( Fig 6C), however, 235 PDGFR-β was unchanged ( Fig S5A). Furthermore, ERBB2 knockdown did not affect MET nor 236 PDGFR-β activation, indicating that neither are regulated by ERBB2 during latency or reactivation. 237 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted February 1, 2023. ; https://doi.org/10. 1101/2023 In testing downstream signaling intermediates, we found that depletion of ERBB2 significantly 238 reduced AKT activation during latency ( Fig 7A). Activation of SRC and ERK1/2 showed a similar 239 trend but was not statistically significant. Strikingly, the phosphorylation of all these kinases were 240 similarly decreased following reactivation. Depleting ERBB2 during reactivation did little to alter 241 these trends, although it did slightly increase ERK1/2 phosphorylation as compared to control 242 lytically induced cells (Fig 7A). No significant trend was observed for pan PKC phosphorylation 243 ( Fig S5B). The reduced phosphorylation of AKT, SRC and ERK1/2 mimicked the decreased 244 ERBB1 phosphorylation phenotypes and suggest that ERBB1:ERBB2 signaling through AKT and 245 to a lessor degree SRC and ERK1/2 is ERBB2 dependent in latent cells and disrupted during 246 early stages of lytic replcation. 247 The janus kinase family (JAKs) are also signaling intermediates downstream of the ERBBs. The 248 JAK3 Inhibitor VI, predicted from the initial KiR screen to affect reactivation, was confirmed to 249 regulate lytic reactivation ( Fig 2D). To further investigate the role of JAK signaling, we tested 250 another JAK inhibitor, tofacitinib that has greater specificity for the JAK proteins ( Fig 7B) and 251 found that it did not affect reactivation ( Fig 7C). Since tofacitinib inhibits the catalytic activity of all 252 JAK kinases while the JAK3 Inhibitor VI is more specific for JAK3 and TYK2 JAKs and several 253 other unrelated kinases (Fig 7B), we wondered if different JAK family members may have 254 counteracting effects on reactivation as we observed for the MKNK family protein kinases (Fig 4).

255
To test this hypothesis, we used specific siRNAs to deplete JAK1, JAK2, JAK3, or TYK2 and 256 performed assays under two lytic induction conditions. Under full lytic inducing conditions (DOX 257 + NaB), none of these individual depletions impacted reactivation significantly ( Fig 7C). We also 258 conducted experiments in uninduced and DOX-only induced cells and observed that depletion of 259 JAK1 increased reactivation 2.4-fold for the DOX condition as compared to the control (Fig S5C), 260 suggesting that JAK1 is a pro-latency factor. Knockdown efficiency and specificity were shown 261 for siRNAs targeting each JAK1, JAK2 and TYK2 using RT-qPCR ( Fig S6A). JAK3 RNA 262 abundance was below the level of detection for RT-qPCR. To assess the knockdown efficiency 263 of the siRNAs targeting JAK3, we transiently overexpressed JAK3 by transfecting HeLa cells with 264 a JAK3 expression plasmid and then transfected cells with targeting siRNAs followed by 265 immunoblot analysis of JAK3 protein levels. Exogenously expressed JAK3 was almost completely 266 depleted in this system ( Fig S6B). These findings illustrate that the JAK1 protein kinase may have 267 pro-latent activity which could be regulated like AKT, SRC and ERK1/2, by ERBB1:ERBB2 268 heterodimeric signaling or by another upstream receptor that is activated in response to RTA 269 expression. 270 ERBB2 phosphorylates CREB1, STAT1, and STAT3 transcription factors during KSHV 271 reactivation. 272 Finally, we tested the role of ERBB2 on the activation of proteins downstream of the selected 273 signaling intermediates, including myristoylated alanine-rich C-kinase substrate (MARCKS), S6 274 ribosomal protein, and several transcription factors ( Fig 6A). Phosphorylation of the MARKS 275 substrate, which is downstream of PKC, was increased during induction of lytic replication (Fig  276  S5D). ERBB2 knock down lessened the induction but not to a statistically significant extent. We 277 detected only small and mostly insignificant effects of lytic induction and ERBB2 knock down on 278 phosphorylation of S6, NFκB and β-catenin (Figs S5E and F). Despite these findings, we did 279 observe a striking phenotype for three other transcription factors. Specifically, the cyclic AMP-280 responsive element-binding protein 1 (CREB1), signal transducer and activator of transcription 1 281 (STAT1), and STAT3 all exhibited a spike in activation under lytic inducing conditions as 282 compared to latency and the effect was attenuated by depletion of ERBB2 in lytically induced 283 cells, supporting a signaling function of ERBB2 to activate these transcription factors (Fig 8). In 284 KSHV infected cells, phosphorylation of CREB1 Ser 133 by MSK1/2 was reported to enhance 285 KSHV lytic gene expression [18]. Thus, our results highlight the role of ERBB2 upstream of 286 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made appear to act by regulation of latency-associated nuclear antigen (LANA) phosphorylation, 295 activation of CREB1 lytic gene transcription factor, and cell proliferation and survival, respectively. 296 Kinases with pro-latency roles include CDK6 which interacts with KSHV v-cyclin and nucleophosmin 297 to restrict lytic replication [48], and AMPKα1 which restricts lytic replication following primary 298 infection through an unknown mechanism [46]. While the role of kinases as regulators of KSHV 299 reactivation has been established, the previously published screens have limitations. For example, 300 kinase overexpression screens can result in artificial kinase catalytic activity, localization, or protein 301 interactions, surveys of protein phosphorylation can implicate activated signaling pathways with 302 substrates shared by many kinases, and proteome analyses can inform on kinase abundance but 303 not on catalytic activity or function [18,23,24]. Additionally, screening of KIs can result in many off-304 target effects due to the inherent broad activity of these drugs [25]. The KiR kinome screening 305 approach that we employed is not without limitations but takes a unique approach as compared to 306 these previously published screens to build on our understanding of kinases and KSHV reactivation. 307 A key advantage of our method is the use of data characterizing the varying potencies of KIs for 308 multiple targets to enable predictions of kinase inhibitors and specific kinases that regulate 309 reactivation. As revealed, our report of this approach can provide new insights into kinase regulation 310 of KSHV latency and reactivation. 311 Using the polypharmacology-based KiR screening method, we were able to predict and validate 312 both kinase inhibitors and specific kinases that regulate KSHV latency and reactivation. Two of 313 the top kinase inhibitors included lestaurtinib, a broadly acting tyrosine kinase inhibitor and K252a, 314 a staurosporine analog that also has broad inhibitor potency. Neither of these drugs has been 315 tested for effects on the KSHV latent-to-lytic replication switch, although lestaurtinib restricts 316 multiple stages of adenovirus replication in cell culture [51] and K252a impedes EBV lytic 317 replication [52]. None of the six initially validated kinases, ERBB4, MKNK2, ITK, TEC, DSTYK, 318 and FRK ( Fig. 3), had been specifically characterized previously as regulators of KSHV 319 reactivation. The kinases that were predicted by the KiR screen but did not validate by siRNA 320 knock down included several MAPKs (MAP3K8, MAP2K2, MAP4K4), CAMK2G, and LRRK2, all 321 of which have functional paralogs. It is possible that redundancies in the signaling pathways would 322 require knocking down more than one of these kinases to reveal a reactivation phenotype. of six kinases validated these as newly described regulators of KSHV reactivation. 335 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted February 1, 2023. 9 The two validated screen hits with the strongest impacts on reactivation, ERBB4 and MKNK2, are 336 from kinase families with members that share high sequence homology and functional overlap.

337
ERBB family members have well-characterized roles in cancer, cell proliferation, and cell survival, 338 and they can form heterodimers to regulate the signaling pathways for these biological processes. 339 MKNK1, like MKNK2, can phosphorylate the cap-binding protein eIF4E to regulate translation [55].

340
MKNK2 has also been shown to phosphorylate KSHV LANA latent protein in in vitro kinase assays 341 [56]. To determine if these closely related kinases have pro-lytic functions like ERBB4 and MKNK2, 342 we tested their effect on KSHV reactivation. Indeed, knocking down either ERBB2 (also called 343 HER2 or neu) or ERBB3 inhibited reactivation, even somewhat more strongly than ERBB4 ( Fig 4A).

344
In contrast, knocking down ERBB1 (Figs 5A and 5B) or MKNK1 ( Fig 4B) promoted certain stages 345 of lytic replication, suggesting that they positively contribute to latency maintenance. These studies 346 illustrate that kinase paralogs with some shared signaling pathways, nonetheless, can have 347 counteracting roles in regulating the KSHV latent-to-lytic replication switch. 348 We selected ERBB2 for further investigation because it plays a unique function in ERBB complexes 349 as with activated lytic replication ( Fig 6B). These data, along with the finding that ERBB1 restricts lytic 357 gene expression (Fig 5), support a model in which ERBB1 restriction of lytic gene expression occurs 358 via ERBB1:ERBB2 heterodimer signaling during latent infection but during lytic replication, this 359 interaction is disrupted (Fig. 9). Intriguingly, this dissociation event coincides with the expression of 360 ERBB3 and ERBB4 which could facilitate ERBB2 partner switching, especially because ERBB2 361 has a stronger affinity for ERBB3 binding as compared to the other ERBBs [36]. 362 To determine which of the many signaling pathways regulated by the ERBB kinases are ERBB2-363 dependent during latency and lytic replication, we assayed for the phosphorylation of key residues 364 that are indicative of activation for a subset of signaling factors. The selected proteins included 365 plasma membrane proteins involved in crosstalk with ERBB kinases, downstream signaling 366 intermediates of ERBB kinases, and downstream substrates of the signaling intermediates ( Fig 6A). 367 Activation of plasma membrane receptors may contribute to the ERBB2 pro-lytic mechanism via 368 receptor crosstalk pathways. For example, the mesenchymal epithelial transition (MET) oncogene 369 can activate ERBB1 in some cancer cells and in others it is activated by ERBB1:ERBB2 370 heterodimers [40,42]. Also, signaling by the platelet derived growth factor receptor beta (PDGFRβ) 371 overlaps with ERBB signaling intermediates [43]. Our data show that MET signaling decreases with 372 reactivation through an ERBB2-independent mechanism, while PDGFRβ phosphorylation was 373 unchanged under all tested conditions (Figs 6C and S5A). These findings do not support a role for 374 MET or PDGFRβ as downstream factors of ERBB1:ERBB2 signaling. 375 From the investigation of signaling intermediates, we found further evidence consistent with 376 ERBB1:ERBB2 signaling during latency. In cells containing latent virus, AKT activation was reduced 377 by ERBB2 depletion and other intracellular kinases, SRC and ERK1/2, showed similar although not 378 statistically significant trends (Fig. 7A). Contrary to the latent state, during lytic replication ERBB2 379 does not appear to activate these signaling intermediates. We also assayed for the role of JAK 380 family kinases as these kinases are intermediates of ERBB signaling cascades and the JAK3 381 Inhibitor VI restricted reactivation ( Fig 2D). The individual JAK family kinases did not have pro-latent 382 or pro-lytic phenotypes under DOX-induced RTA plus NaB conditions ( Fig 7C), but we did observe 383 a moderate pro-latent phenotype for JAK1 under DOX-induced RTA alone conditions (Fig S5C). 384 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted February 1, 2023. ; https://doi.org/10.1101/2023.02.01.526589 doi: bioRxiv preprint Our investigation of ERBB2 sigaling revealed that ERBB1:ERBB2 heterodimer activates AKT and 385 to a lessor extent SRC and ERK1/2 during latency. JAK1 may also be activated by these receptors 386 or another factor to promote latency. But none of these signaling intermediates likely contribute to 387 the ERBB2-driven promotion of lytic replication. 388 The last category of factors that we tested are downstream substrates of the selected signaling 389 intermediates, although we recognize the caveat that other factors may regulate these substrates. 390 We found that ERBB2 was required to fully activate CREB1 (Fig 8), a transcription factor that 391 promotes KSHV lytic gene expression [18]. ERBB2 also activated STAT1 and STAT3 during lytic 392 replication (Fig 8), but only pro-latency roles for STAT1 and STAT3  the role of ERBB2 upstream of these other kinases will be of interest. The expression of another 409 pro-lytic kinase validated from our screen, ITK, is induced following KSHV reactivation in iSLK cells 410 (Table S4, [34]). This kinase and other TEC family kinases can interact with ERBB3 cytoplasmic tail 411 following ERBB2-mediated phosphorylation [61,62] suggesting that TEC family kinases may 412 function as signaling intermediates during the transition between latency and lytic replication. 413 Continued dissection of the roles of kinases during KSHV lytic replication will provide insight into 414 the overlapping or parallel pathways at work to coordinate this replicative switch of KSHV. 415 Together, these experiments confirm the utility of a polypharmacology-based kinome screen to 416 study KSHV reactivation regulators. The iSLK system provided a convenient cell line for these proof-417 of-principle studies with relatively high levels of induced-lytic replication, which is not achievable in 418 most KSHV latent cell culture systems, though conducting this screening approach in other cell 419 types and conditions is of interest. The translational potential of this research is most evident with 420 the new connection identified between KSHV latent-to-lytic replication switch and ERBB signaling. counteracting roles of the ERBBs to coordinate the critical KSHV latent-to-lytic replication switch. 426 Continued mechanistic probing of these factors will enhance our understanding of the intricacies of 427 this viral switch and application to other cell types and systems may inform on the therapeutic 428 potential of targeting these kinases to affect KSHV-driven diseases. 429 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

Kinase inhibitor Regularization (KiR) modeling 507
KiR models for KSHV reactivation were generated as previously described [31,33] ). A set of 29 508 inhibitors were tested on KSHV LRI latently infected iSLK cells as described above, with the end 509 result being a single response for each drug that represents the change in KSHV reactivation (as 510 % DMSO control) at the profiled dose of the inhibitor. The kinase inhibition profiles of each inhibitor 511 and the quantitative responses to those inhibitors were used as the explanatory and response 512 variables, respectively, for elastic net regularized multiple linear regression models [76]. Custom 513 R scripts (available at https://github.com/FredHutch/KiRNet-Public) employing the glmnet 514 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted February 1, 2023. ; https://doi.org/10.1101/2023.02.01.526589 doi: bioRxiv preprint 13 package were used to generate the final models .
[77] Leave-one-out cross-validation (LOOCV) 515 was used to select the optimal value for the penalty scaling factor λ. Models were computed for 516 11 evenly spaced values of α (the relative weighting between LASSO and Ridge regularization) 517 ranging from 0 to 1.0 inclusive. Kinases with positive coefficients in at least one of these models 518 (with the exception of α = 0, which always has non-zero coefficients for every kinase) were 519 considered hits. 520 siRNA or plasmid transfections 521 KSHV LRI latently infected iSLK cells were seeded at 4x10 5 cells/well into a 6-well plate. The next 522 day, cells were transfected with 100nM non-targeting control or individual kinase targeting On-523 TARGETplus SMARTPool siRNAs (horizon; Table S6) including four unique target sequences 524 and 5µl RNAiMAX (ThermoFisher #13778150) in Opti-MEM (ThermoFisher #31985088) per well 525 following manufacturer's instructions. For transfection with the JAK3 expressing plasmid, HeLa 526 cells were seeded into a 6-well plate at 5x10 5 cells/well. The next day, cells were transfected with 527 siRNAs as described above and 2 days following cell seeding, cells were transfected with 1.5µg 528 pcDNA3.1-JAK3 and 3µl lipofectamine 2000 (ThermoFisher #11668027) per well in Opti-MEM 529 following manufacturer's recommendations. 530

Reverse transcription qPCR 531
Knockdown efficiencies of targeting siRNAs were evaluated by reverse transcription qPCR 3-days 532 after siRNA transfection. For relative quantification of lytic gene expression, cells transfected with 533 siRNAs were treated with 1 µg/ml DOX plus 1 mM NaB at 3-days post transfection and harvested 534 at 24h following lytic induction for mCherry and ORF10 and at 48h for K8. System C1000 Touch Thermal Cycler using primers listed in table 5. Relative mRNA levels were 540 normalized to tubulin control and calculated using the ∆∆Ct method for experimental conditions 541 as compared to control conditions. 542

Reverse-phase protein arrays (RPPA) 543
Control or siRNA treated KSHV LRI iSLK cells were harvested 4-days following siRNA transfection 544 and 24h following KSHV lytic induction with 1 µg/ml DOX and 1 mM NaB for protein lysate 545 microarray analysis. Sample preparations and protein array analyses were performed as detailed 546 in Luckert et al. [78]. In brief, cells were rinsed then lysed in 50 mM Tris-HCl, 2% sodium dodecyl 547 sulfate, 5% glycerol, 5 mM ethylenediaminetetraacetate, and 1 mM sodium fluoride, 1X Complete 548 Protease Inhibitor Cocktail (1 tablet per 10 ml, Roche), 1X Pierce protease plus phosphatase 549 inhibitor tablet (Thermo Scientific #A32959), 10 mM β-glycerol phosphate, 1 mM PMSF, 1 mM 550 sodium orthovanadate, and 1 mM dithiothreitol. After filtering through a 0.2-µm filter plate, the 551 lysates were printed onto a nitrocellulose-coated slide using Aushon 2470 arrayer. Primary 552 antibodies listed in table S7 were diluted 1:100 and incubated with slide for 24h on an orbital 553 shaker at 4°C. IRDye secondary antibodies listed in table 7 were diluted to 1:1000 and incubated 554 with slide for 1h, shaking at room temperature. The microarray slides were scanned in 680-nm 555 and 800-nm channels with an Odyssey imager. Protein quantitation for siCtrl, uninduced KSHV LRI 556 iSLK samples were set to 100 and samples were normalized to these controls. Six independent 557 siRNA transfections were completed for the six RPPA experiments. 558 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted February 1, 2023. ; https://doi.org/10.1101/2023.02.01.526589 doi: bioRxiv preprint

Immunoblot assays 569
Lysates of uninfected, KSHV LRI latently infected, or infected and 3-day lytically induced iSLK cells 570 were separated on 8 or 10% polyacrylamide gels. For ERBB2 and paired actin blots, KSHV LRI 571 latently infected iSLK cells were harvested 2 days after siRNA transfection. For the JAK3 and 572 paired actin control blots, HeLa cells were harvested 3 days after siRNA transfection and 2 days 573 after pcDNA3.1-JAK3 transfection. For all gels, the protein was transferred onto a polyvinylidene 574 difluoride (PVDF) membrane (Millipore), and proteins were detected by probing with specific 575 antibodies (Table S7) using the Western Star chemiluminescent detection system (Applied 576 Biosystems) according to the manufacture's recommendations. 577

Kinase activity profiles for JAK inhibitors 591
The kinase activity profiles for tofacitinib and JAK3 inhibitor VI (Fig 7B) were taken from the 592 publicly available Kinhibition website (https://kinhibition.fredhutch.org/). Tofacitinib data indicates 593 specific restriction of all JAK kinases and two other kinases, LRRK2 and PKN1. The JAK3 VI 594 inhibitor restricts JAK3, TYK2 and 15 other kinases including Pim-1 and Pim-3 pro-lytic kinases 595 and two kinases predicted from the kinome screen, CLK1 and MAP4K4. A heatmap of kinase 596 activities during treatment with 500nM of either tofacitinib or JAK3 inhibitor VI was generated in 597 GraphPad Prism 8.0.1. 598 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made  binding peptide fused to a truncated low-affinity nerve growth factor receptor (SBP-ΔLNGFR), a 616 P2A self-cleaving peptide sequence, and then mCherry containing a nuclear localizing signal 617 (

mCherry-NLS). (B) Fluorescence images of uninfected iSLK and KSHV LRI infected cells collected 618
using an Incucyte Imaging System. Infected cells express the virus encoded GFP under control of 619 the cellular EF-1α promoter (top row). mCherry-nls (bottom row) detected after inducing lytic 620 replication with 1µg/ml DOX (middle column) or DOX plus 1mM NaB (right column) for 72h. 621 Reactivation data for DOX plus NaB treatment alone (black bar and dotted black line set to 100, 638 represents ~20% total cells) or combined with KIs (red bars, 0.5 µM) were calculated as a percent 639 of DOX plus NaB control. (D) Twelve additional KIs predicted from the initial model were tested 640 for KSHV reactivation (pink bars) and cell viability (grey bars). The iSLK KSHV LRI cells were 641 untreated ("-" DMSO only) or treated with DOX plus NaB alone (DMSO) or in combination with 642 KIs at 125nM, 500nM, or 2μM. Reactivation (pink bars) for each KI condition was measured by 643 mCherry expression and calculated as a percent of DOX plus NaB with DMSO control. Cell 644 viability (grey bars) was determined by cell confluence as a percent of DOX plus NaB with DMSO 645 control. PF: PF-477736; AZD: AZD3463; GSK: GSK-650394; ASP: ASP-3026; Gefit: Gefitinib; 646 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made data on kinase inhibitors as the training set (C). This data was used to predict the response of 649 potential kinase inhibitors. A subset of these predictions was then tested as a validation set (D), 650 and the results were used to refine the model's predictions. The No KI Ctrl (yellow star) represents 651 the DOX plus NaB with DMSO control condition. 652 Fig S2: Polypharmacology-based kinome screen in the absence of KSHV lytic inducing 653 agents. KSHV LRI reactivation phenotypes were obtained from 20 of the 29 pre-selected kinase 654 inhibitors that did not cause cellular toxicity. KSHV reactivation for control (black bar and dotted 655 black line) and kinase inhibitor treatment (red bars) were calculated as a percent of DOX plus NaB 656 treated cells set to 100 from data in Fig 2C. In this graph, 1.0 represents ~0.2% total cells and the 657 dotted line represents spontaneous reactivation, ~0.02% total cells. 658 For each knockdown, the efficiencies were averaged and listed below the corresponding kinase 665 target as % KD. P-value * < 0.05 and ** < 0.01. 666 iSLK cells using RT-qPCR from total RNA harvested at 3-days post transfection with siRNAs. 677

Fig S4: Knockdown efficiencies and specificity for ERBB and MKNK family members. (A) 678
Knockdown efficiency of ERBB2 targeting siRNA was evaluated by immunoblot for ERBB2 protein 679 at 2 days following siRNA transfection of KSHV LRI latently infected iSLK cells. Knockdown specificity 680 for siRNAs targeting (B) ERBB or (C) MKNK family kinases were evaluated in KSHV LRI latently 681 infected iSLK cells using RT-qPCR from total RNA harvested at 3-days post transfection with 682 siRNAs. 683

684
KSHV LRI latently infected iSLK cells were transfected with siRNA control or siRNAs targeting 685 individual ERBB family kinases and then 3-days later treated with DOX plus NaB for 24h or 48h. 686 Transcript levels for (A) PrPAN-mCherry (B) ORF10 and (C) K8.1 were quantified using target 687 specific primers and RT-qPCR of total RNA. 688 siRNA control or siRNAs targeting ERBB2 and then 3-days later untreated or treated with DOX plus 691 NaB for 24h. Cells were harvested, and protein lysates were analyzed using a reverse-phase 692 protein array (RPPA) for ERBB1 phosphorylation at Tyr 1173 or (C) for MET phosphorylation at 693 Tyr 1349 . P-value * ≤ 0.05. 694 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted February 1, 2023. ; https://doi.org/10.1101/2023.02.01.526589 doi: bioRxiv preprint

Fig 7: ERBB2 phosphorylation of intermediate kinases. (A) KSHV LRI latently infected iSLK cells 695
were transfected with siRNA control or siRNAs targeting ERBB2 and then 3-days later untreated or 696 treated with DOX plus NaB for 24h. Cells were harvested, and protein lysates were analyzed using 697 a RPPA for the phosphorylation of the indicated signaling intermediates downstream of ERBB2. (B) 698 Kinase inhibition profiles for tofacitinib and JAK3 inhibitor VI from the Kinhibition database 699 (https://kinhibition.fredhutch.org/). JAK3 VI inhibitor restricts JAK3, TYK2 and 15 other kinases. (C) 700 Cell viability (grey bars) and KSHV reactivation (red bars) were measured for KSHV LRI latently 701 infected iSLK cells untreated or tofacitinib treated cells in combination with lytic inducing agents 702 DOX plus NaB for 72h and for cells transfected with siRNAs targeting individual JAK family kinases 703 and 3-days later treated with DOX plus NaB for 72h. Control DMSO or control siRNA transfected 704 cells treated with DOX plus NaB (dotted black lines) were set to 100 and data for each condition 705 was calculated as a percent of the control. Kinase knockdown efficiencies at 3-days following siRNA 706 transfection were determined before addition of lytic inducing drugs and graphed in Fig S6. For 707 each knockdown, the efficiencies were averaged and listed below the corresponding kinase target 708 as % KD. P-values * ≤ 0.05 and ** ≤ 0.01. 709 or siRNAs targeting ERBB2 and then 3-days later untreated or treated with DOX plus NaB for 24h. 726 Cells were harvested, and protein lysates were analyzed using a RPPA for transcription factor 727 phosphorylation of CREB1 at Ser 133 , STAT1 at Tyr 701 , and STAT3 at Tyr 705 . P-values * ≤ 0.05, ** ≤ 728 0.01, and *** ≤ 0.001. 729 ERBB1:ERBB2 heterodimer signaling activates AKT and this trend applies to S6, SRC and ERK1/2 738 signaling intermediates to promote the latent state. JAK1 in some contexts also promotes latency 739 and may function downstream of the ERBB1:ERBB2 heterodimer. This signaling is repressed 740 during lytic replication as ERBB2 forms a heterodimer with newly expressed ERBB3, a switch 741 facilitated by high affinity binding of ERBB2 to ERBB3, and ERBB2-dependent signaling activates 742 lytic gene transcription factor, CREB1. 743 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted February 1, 2023. ; https://doi.org/10.1101/2023.02.01.526589 doi: bioRxiv preprint Supporting Information (SI) Captions 744 S1 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted February 1, 2023. ; https://doi.org/10.1101/2023.02.01.526589 doi: bioRxiv preprint . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted February 1, 2023. ;https://doi.org/10.1101https://doi.org/10. /2023