The dual specificity kinase DYRK1A modulates the levels of cyclin L2 to control HIV replication in macrophages.

HIV continues to be a major public health problem worldwide, with over 36 million people living with the virus. Although antiretroviral therapy (ART) can control the virus, it does not provide cure. The virus hides in the genomes of long-lived cells, such as resting CD4+ T cells and differentiated macrophages. To get a cure for HIV, it is important to identify and characterize the cellular factors that control HIV multiplication in these reservoir cells. Previous work showed that cyclin L2 is required for HIV replication in macrophages. However, how cyclin L2 is regulated in macrophages is unknown. Here we show that the protein DYRK1A interacts with and phosphorylates cyclin L2. Phosphorylation makes cyclin L2 amenable to cellular degradation, leading to restriction of HIV replication in macrophages.

minimal effect in dividing cells. This increase in HIV replication upon DYRK1A inhibition is dependent on intact cyclin L2. We found that depletion of DYRK1A increases cyclin L2 levels, thus increasing HIV replication. We present evidence to show that DYRK1A enhances the degradation of cyclin L2 through phosphorylation to restrict HIV replication in macrophages.

RESULTS
Cyclin L2 interacts with DYRKIA through its RS domain. To determine whether cyclin L2 interacts with DYRK1A, we transfected HeLa cells with green fluorescent protein (GFP)-DYRK1A and immunostained for endogenous cyclin L2. As shown in Fig.  1A, most of the cyclin L2 in the nucleus colocalized with DYRK1A in nuclear speckles (39). We confirmed these findings in macrophage-like cells by performing coimmunoprecipitations for the two endogenous proteins in differentiated THP-1 cells. Figure 1B shows that cyclin L2 coimmunoprecipitated with DYRK1A in cells transfected with control small interfering RNA (siRNA) but not in DYRK1A knockdown cells or IgG controls. These results show that the two proteins specifically interact. Next, we determined which domain of cyclin L2 is responsible for the interaction. Cyclin L2 has two main domains, the cyclin box in the N terminus and an arginine-and serine-rich (RS) domain in the C terminus (Fig. 1C). To find which domain is critical for interacting with DYRK1A, we constructed three Myc-tagged clones consisting of the full-length cyclin L2, cyclin L2ΔRS (residues 1 to 286), and cyclin L2Δc-box (residues 287 to 520). We transfected 293T cells with these constructs, pulled down Myc with a monoclonal antibody, and immunoblotted for endogenous DYRK1A. Full-length cyclin L2 interacted with DYRK1A, as expected (Fig. 1D). In addition, the mutant containing the RS domain (cyclin L2Δc-box) coimmunoprecipitated with DYRK1A. However, the mutant lacking the RS domain (cyclin L2ΔRS) could not pull down DYRK1A, indicating that the RS domain is required for the interaction between the two proteins. Knockdown of DYRK1A disproportionately increases HIV replication in nondividing cells. In dividing cells, previous work showed that depletion of DYRK1A results in less than a 2-fold increase in HIV replication (48). Given that the action of cyclin L2 on HIV replication is restricted to macrophages and the two proteins interact, we wondered what effect DYRK1A would have on HIV replication in nondividing cells. Therefore, we knocked down DYRK1A in HeLa and THP-1 monocytic cells to compare degrees of HIV replication in the two cell types ( Fig. 2A). As previously reported (48), there was about a 1.5-fold increase in HIV infection in HeLa cells (Fig. 2B). To test the idea that DYRK1A may be more important for HIV replication in monocytes/macrophages, we used stable THP-1 cells carrying either control or DYRK1A small hairpin RNA (shRNA). Since THP-1 cells can be differentiated into macrophage-like cells with phorbol 12-myristate 13-acetate (PMA), they provide a good model to compare the roles of DYRK1A in HIV infection in dividing and nondividing cells. Knockdown of DYRK1A with shRNA in undifferentiated THP-1 monocytes resulted in about a 3-fold increase in HIV infection (Fig. 2C). However, when the THP-1 cells were differentiated (dTHP-1) before infection, HIV replication increased to 8-fold (Fig. 2D). To confirm these results, we used siRNA to knock down DYRK1A in primary monocyte-derived macrophages (MDMs) isolated from three different HIV-negative donors (Fig. 2E). As shown in Fig. 2F to H, in DYRK1A knockdown cells, HIV infection increased 9.7-to 12.7-fold depending on the donor. The increases in HIV replication were not due to proliferation of THP-1 cells or MDMs, as shown by the cell proliferation assays whose results are shown in Fig. 2I to K. Next, we used macrophage-tropic replication-competent HIV-1 (BaL-3) to determine if DYRK1A would have the same effect with multiple rounds of infection. We infected control or DYRK1A knockdown MDMs isolated from three different donors ( Fig. 3A) with HIV-1 BaL-3 for a total of 72 h. After 24 h of infection, we replaced the media and collected supernatants after 48 h. We then used the collected virus to infect TZM-bl indicator cells. As shown in Fig. 3B to D, knockdown of DYRK1A increased replication of HIV-1 8-to 10-fold depending on the donor. Cell lysates from the MDMs were processed for Western blotting and showed a significant increase in HIV Gag proteins in DYRK1A knockdown cells (Fig. 3E). Finally, we employed INDY, an established inhibitor of DYRK1A which binds to the protein's ATP pocket to inhibit its action (49). Treatment with INDY increased HIV replication in a dose-dependent manner in all three donors ( Fig. 3F to H). Treatment with INDY did not result in proliferation of MDMs (Fig.  3H). Taken together, these results confirm that DYRK1A has a more pronounced effect on HIV replication in macrophages.
Depletion of cyclin L2 abolishes DYRK1A-mediated HIV restriction. Since cyclin L2 promotes HIV replication in macrophages and DYRK1A has the opposite effect, we investigated whether the DYRK1A restriction of HIV replication in macrophages is dependent on intact cyclin L2. If that were the case, depletion of cyclin L2 would abolish the effect of DYRK1A illustrated in Fig. 2   Next, we used cyclin L2 CRISPR/Cas9 knockout THP-1 cells to interrogate the effect of DYRK1A inhibition in the context of cyclin L2 depletion. In control parental cells, treatment with INDY increased HIV infection 8-fold. However, in cyclin L2 knockout cells, the effect of INDY was reduced only 0.8-fold ( Fig. 4F and G), without a decrease in cell numbers (Fig. 4H). This shows that the interaction between the two proteins has functional consequences on HIV replication and that intact cyclin L2 is required for the effect of DYRK1A on HIV replication.
DYRK1A controls cyclin L2 levels in macrophages. Next we proceeded to determine a possible mechanism for how cyclin L2-DYRK1A interactions control HIV replication in macrophages. Since DYRK1A is a kinase for cyclin L2 and multiple kinases modulate the levels of their respective cyclins (50,51), we wondered if DYRK1A controls cyclin L2 levels in macrophages. For instance, if DYRK1A reduced the levels of cyclin L2 in macrophages, that could reduce the amount of HIV produced in those cells. First, we immunoblotted for levels of the two proteins in dividing and nondividing THP-1 cells. As shown in Fig. 5A and B, while the levels of DYRK1A were higher in differentiated THP-1 cells, cyclin L2 levels were 2.5-fold lower in the same cells. When the cells were treated with the proteasome inhibitor MG132, the levels of cyclin L2 in dTHP-1 cells were restored to the levels detected in dividing cells. This suggests that increased expression of DYRK1A in nondividing cells accelerates degradation of cyclin L2 through the proteasome. To determine a direct role for DYRK1A in modulating cyclin L2 levels, we immunoblotted for cyclin L2 in control versus DYRK1A knockdown differentiated THP-1 cells. Figure 5C shows that compared to those in controls, cyclin L2 levels increased 2-fold in cells with DYRK1A depletion. Next, we immunoprecipitated cyclin L2 in dTHP-1 cells treated with control or DYRK1A siRNA and immunoblotted for total cyclin L2 or phosphorylated cyclin L2. While total cyclin L2 was increased, the phosphorylated portion was reduced, as expected ( Fig. 5D and E). In addition, total SAMHD1 levels were reduced but phosphorylated levels remained the same, consistent with our previous findings that increased cyclin L2 reduces SAMHD1 levels but has no effect on the phosphorylated portion (34). Finally, we overexpressed increasing amounts of DYRK1A to determine the effect on cyclin L2 levels. Increasing levels of DYRK1A resulted in proportional degradation of cyclin L2 (Fig. 5F and G). When MG132 was added where DYRK1A had the most effect, cyclin L2 degradation was rescued. Taken together, these data show that DYRK1A increases the turnover of cyclin L2 through the proteasome, resulting in reduced HIV production.

Cyclin L2 is stabilized by dephosphorylation to increase HIV replication.
To test the possibility that DYRK1A-mediated phosphorylation increases cyclin L2 degradation, we mutated serine residues in cyclin L2 shown previously to be phosphorylated by DYRK1A (39). We made phospho mutants of cyclin L2 by replacing serine residues that are adjacent to proline with alanine in the RS domain at the C terminus. For mutant 1, named P-369, four serine residues were mutated to alanine: S330, S338, S348, and S369. For mutant 2, named P-330, seven serine residues were mutated to alanine: S330, S338, S348, S369, S394, S401, and S427.
First, we expressed the wild type (WT) and mutants together with DYRK1A in 293T cells and immunoblotted for phosphorylated and total cyclin L2. As shown in Fig. 6A, while the total levels of the mutants were increased compared to those of WT cyclin L2, the phosphorylated portions were reduced, as expected. This indicated that the dephosphorylated cyclin L2 mutants may be more stable than the WT. To determine the stability of the WT and mutants in a more methodical manner, we treated cells with cycloheximide to block new protein synthesis and performed a chase for 36 h to generate degradation curves. The results showed that the mutant with the most mutated DYRK1A phosphorylation sites (P-330) was the slowest to degrade ( Fig. 6B and C). This shows that phosphorylation of cyclin L2 by DYRK1A increased cyclin L2  degradation. Next, we tested if the mutants were functional and whether they affected HIV-1 replication. We expressed the Myc-tagged mutants in cyclin L2 CRISPR/Cas9 knockout THP-1 cells, differentiated with PMA and infected with HIV-1 Luc. As shown in Fig. 6D, wild-type cyclin L2 retained its function in THP-1 cells. We observed a 5-fold increase in HIV replication in cells expressing wild-type cyclin L2 and up to a 7.5-fold increase in cells expressing cyclin L2 (P-330), the mutant with the most dephosphorylation. This suggests that the more stable cyclin L2 in macrophages, the higher the HIV replication.

DYRK1A inhibition increases HIV-1 transcription.
To determine which step of the viral life cycle is modulated by the cyclin L2-DYRK1A interaction, we performed experiments to evaluate late reverse transcription (RT) and transcription of HIV provirus. We infected MDMs and dTHP-1 cells with HIV-1 BaL-3 and quantified the levels of HIV-1 DNA by quantitative PCR (qPCR) 48 h postinfection to determine whether DYRK1A knockdown had any effect on late reverse transcription. As shown in Fig. 7A to C, the levels of HIV-1 gag DNA were unchanged, suggesting that knockdown of DYRK1A did not alter reverse transcription in either MDMs or differentiated THP-1 cells. Next, to determine HIV mRNA expression from integrated HIV provirus, we infected differentiated THP-1 cells with full-length HIV-1 for 24 h, washed and treated with integrase inhibitor raltegravir and protease inhibitor darunavir to prevent new integration and reinfection, respectively. After 24 h of INDY treatment, we isolated RNA from the cells and measured gag mRNA as a measure of HIV transcription. We observed a 3-fold increase in gag mRNA upon DYRK1A knockdown (Fig. 7D) or inhibition with INDY (Fig.  7E). Finally, to confirm that this effect on transcription was through cyclin L2, we used cyclin L2/CRISPR/Cas9 knockout cells. As shown in Fig. 7F, knockout of cyclin L2 abrogated the increase in transcription observed with DYRK1A knockdown. Taken together, these data show the requirement for cyclin L2 for DYRK1A actions on HIV-1 replication in macrophages.

DISCUSSION
Identification of cellular factors involved in HIV-host interactions is critical to understanding mechanisms of persistence and latency, which could eventually lead to cure. Here we show that cyclin L2, previously shown to degrade SAMHD1, interacts with the kinase DYRK1A. Knockdown or pharmacological inhibition of DYRK1A had pronounced effects on HIV replication in macrophages compared to dividing cells, even with multiple rounds of infection. Overall, the data point to a mechanism whereby inhibition of DYRK1A results in increased amounts and a more stable cyclin L2 which consequently promote HIV replication in macrophages. Since cells with knockdown of both proteins produce little HIV, it is likely that cyclin L2 plays the central role in this interaction. Degradation of a cyclin by DYRK1A is not without precedent. Thompson et al. showed that DYRK1A phosphorylates cyclin D3 to promote its degradation and enhance quiescence in T and B cells (52). Degradation of cyclin L2 leads to restriction of HIV infection in macrophages but less so in dividing cells. The reason for the pronounced effect in nondividing cells is still unclear. Given that SAMHD1 phosphorylation makes it inactive in dividing cells (53,54), it is possible that differential phosphorylation in dividing versus nondividing cells could be the ultimate mechanism. Here, in the context of a clean background (cyclin L2 knockout cells), we show that expression of cyclin L2 in macrophages increases HIV replication, confirming our previous results (34). We found that terminally differentiated cells express more DYRK1A, and less cyclin L2, which correlates with the production of less HIV. It will be interesting to determine how the interplay between the two proteins affects HIV replication in resting T cells, which are considered differentiated and critical for HIV latency. Supporting the assertion that cyclin L2/DYRK1A interactions may be critical in HIV latency are the findings that inhibitors of DYRK1A increase HIV latency reversal in resting T cells (55). Although the mechanism for this finding is not known, it will not be surprising if cyclin L2 plays a role in resting T cells similar to the one it plays in macrophages.
A unique feature of the cyclin L proteins is possession of the RS domain, a hallmark of splicing factors (33,37,39,56). Cyclin L2 interacts with splicing factors and is believed to be involved in pre-mRNA splicing. Interestingly, we found that the RS domain is required for interactions with DYRK1A, making it likely that DYRK1A also localizes in splicing factor compartments. We found that knockdown of DYRK1A resulted in increased HIV mRNA production from integrated provirus. This could be at the level of transcription or splicing. Since knockdown of cyclin L2 abrogated this effect, it would suggest that cyclin L2 plays a role in HIV transcription or splicing in macrophages. Recent findings that SAMHD1 plays a role in HIV transcription in resting cells (57) make this an interesting line of investigation for future studies. Previous studies showed that DYRK1A inhibits HIV transcription in actively dividing cells through nuclear factor of activated T cells (NFAT) (48). However, how HIV transcription or splicing in macrophages is controlled by cyclin L2, DYRK1A, and NFAT interplay remains to be determined.
We found that cyclin L2 is degraded through the proteasome and that this degradation was dependent on DYRK1A. Knockdown of DYRK1A or mutations of its phosphorylation sites in cyclin L2 had the same effect of stabilizing cyclin L2 and increasing HIV replication in macrophages. In this regard, cyclin L2 resembles cyclin D1 and cyclin E, which are also stabilized upon dephosphorylation (50,51). Given that cyclin L2 has multiple potential phosphorylation sites, especially in the RS domain, it is likely that other kinases or phosphorylation sites play a role in its phosphorylation in macrophages. This could explain the moderate increases in HIV replication when the phospho mutants were expressed in macrophages compared to the wild type.
Our previous studies showed cyclin L2-mediated degradation of SAMHD1 resulted in fewer HIV reverse transcription products. In this study, DYRK1A did not affect HIV reverse transcripts but knockdown increased HIV mRNA production. Therefore, it is possible that cyclin L2 affects the HIV life cycle at two distinct points depending on the specific interacting protein partner.
From the foregoing, a model emerges whereby cyclin L2 control of HIV replication in macrophages is modulated in part by the kinase DYRK1A. Interaction between the two proteins keeps the levels of cyclin L2 to a minimum, a situation that favors HIV restriction in macrophages. We conclude that regulation of cyclin L2 levels by DYRK1A contributes to HIV restriction in macrophages.

DYRK1A Controls HIV in Macrophages
Journal of Virology

MATERIALS AND METHODS
luciferase activity normalized to total protein concentration per well was then measured to assess viral replication postinfection. Reads of Յ300 U were considered background. Immunoprecipitations and Western blotting. Twenty-four to 48 h posttransfection, cells were washed once with PBS and lysed with PBS buffer with 0.2% NP-40 and protease inhibitor cocktail (Roche). Total protein was measured with a bicinchoninic acid (BCA) protein assay kit. For immunoprecipitation assays, supernatants were incubated with immunoprecipitating antibody overnight at 4°C. Protein Aagarose beads were then used to capture the complexes for 2 h at room temperature or overnight at 4°C, followed by three washes with PBS-0.1% NP-40. Immunoprecipitants were eluted with SDS sample buffer at 100°C for 5 min. Protein samples were separated on a 12.5% SDS-PAGE gel and transferred to nitrocellulose. The membrane was blocked for 1 h at room temperature or overnight at 4°C in 5% milk or 4% BSA in PBS-0.05% Tween 20, followed by incubation with primary antibodies overnight at 4°C. The membrane was washed with PBS-0.05% Tween and probed with horseradish peroxidase (HRP)conjugated secondary antibodies at room temperature for 1 h. The membranes were stained with Immobilon Western chemiluminescent HRP (Millipore) or Femto SuperSignal, and quantitation was processed with ImageJ software.
Statistics. Student's t test was used for pairwise comparisons and one-way analysis of variance (ANOVA) with a follow-up Tukey multiple-comparison test for multiple comparisons. P values of Ͻ0.05 were considered significant. GraphPad Prism 9.00 software (GraphPad Software, Inc., La Jolla, CA) was used for calculations.

ACKNOWLEDGMENTS
G.B.K. is a recipient of the Harold Amos Medical Faculty Development grant from the Robert Woods Johnson Foundation. This work was supported by NIH K08 grant 1K08 AI120854 to G.B.K.
The funders had no role in experimental design, data analysis, or writing of the manuscript.
G.B.K. came up with the concept for the studies and wrote and edited the manuscript. J.K.K. performed experiments and wrote the initial draft. L.R. provided experimental inputs and reagents and edited the manuscript.
We declare no conflict of interest.