The Functional Redundancy of Neddylation E2s and E3s in Modulating the Fitness of Regulatory T Cells

Neddylation is necessary for activation of Cullin-RING ligases (CRLs), which degrade various immune regulatory proteins. Our recent study showed that while depletion of neddylation E2–E3 pair Ube2f-Sag in regulatory T (Treg) cells had no obvious phenotype, the same depletion of either Ube2m or Rbx1 caused inflammation disorders with different severity. Whether these E2s or E3s compensate each other in functional regulations of Treg cells is, however, previously unknown. In this report, we generated Foxp3Cre;Ube2mfl/fl;Ube2ffl/fl or Foxp3Cre;Rbx1fl/fl;Sagfl/fl double-null mice by simultaneous deletion of both neddylation E2s or E3s in Treg cells, respectively. Remarkably, Ube2m&Ube2f double-null mice developed much severe autoimmune phenotypes than did Ube2m-null mice, indicating that Ube2m markedly compensates Ube2f in Treg cells. The minor worsened autoimmune phenotypes seen at the very early stage in Rbx1&Sag double-null than Rbx1-null mice is likely due to already severe phenotypes of the later, indicating a minor compensation of Rbx1 for Sag. The RNA profiling-based analyses revealed that up- and down-regulations of few signaling pathways in Treg cells are associated with the severity of autoimmune phenotypes. Finally, severer inflammation phenotypes seen in mice with double E3-null than with double E2-null Treg cells indicate a neddylation-independent mechanism of 2 E3s, also known to serve as the RING component of CRLs in regulation of Treg cell fitness.

Our previous studies revealed that during mouse embryonic development, Rbx1 and Sag/Rbx2 are functional nonredundant, since in mice total knockout of Rbx1 or Sag leads to embryonic lethality at different stage with different mechanism [8,9]. Although whether Ube2m or Ube2f is also functionally redundant during development is currently unknown, the fact that neddylation of cullin-1 to cullin-4 versus cullin-5 by the Ube2m-Rbx1 pair versus the Ube2f-Sag pair [4] strongly suggests their functional independency. On the other hand, we recently showed a cross-talk between UBE2M and UBE2F, 2 neddylation E2s, in which UBE2M serves as a ubiquitylation E2 to promote ubiquitylation of UBE2F for proteasome degradation [10]. Whether there is a cross-talk between RBX1 and SAG remains unknown.
Regulatory T (T reg ) cells are a suppressive subpopulation of CD4 + T lymphocytes, and transcription factor Foxp3 is the master marker of T reg cells [11][12][13]. T reg cells are essential for immune homeostasis, which is illustrated by the early-onset fatal inflammation caused by depletion of T reg cells [14]. We recently investigated the role of neddylation-CRL system in T reg cells using Foxp3 Cre -LoxP system by generating 4 conditional knockout mouse models with deletion in T reg cells of 2 neddylation E2s, Ube2f or Ube2m, or 2 dual E3s, Sag or Rbx1, individually [15]. Interestingly, mice with the deletion of Ube2f or Sag in T reg cells had no visible phenotype, indicating that the Ube2f-Sag axis plays little, if any, role in regulation of T reg cells under physiological condition. Meanwhile, mice with the deletion of Ube2m or Rbx1 in T reg cells suffered from severe autoimmune inflammatory phenotypes, indicating the functional requirement of the Ube2m-Rbx1 axis in T reg cells [15]. The fact that much severer phenotypes are observed in Rbx1-null mice than in Ube2m-null mice indicates a neddylation-independent role of Rbx1 in T reg cells. However, it is still unknown whether 2 neddylation E2s or E3s are redundant in functional regulation of T reg cells.
In this study, we addressed this functionally redundant question by generating T reg double knockout mouse models of 2 E2s, Foxp3 Cre ;Ube2m fl/fl ;Ube2f fl/fl , or 2 E3s, Foxp3 Cre ;Rbx1 fl/fl ;Sag fl/fl , for simultaneous depletion of Ube2m and Ube2f or Rbx1 and Sag in T reg cells, respectively. Both T reg cell double knockout mice developed severe autoimmune disorders with an early-onset fatality. Given that Ube2f deletion in T reg cells had no phenotype, much severe autoimmune phenotypes in Ube2m&Ube2f double deletion than in Ube2m single deletion in T reg cells indicate that Ube2m compensates the function of Ube2f in T reg cells. On the other hand, minor increased severity in T reg Rbx1&Sag double deletion than Rbx1 single deletion suggests a major role of Rbx1 in regulation of T reg cell function. Furthermore, a greater severity in autoimmune phenotypes of mice with Rbx1&Sag deficiency than Ube2m&Ube2f deficiency in T reg cells suggests a neddylation-independent function of Rbx1/Sag. The comparison of RNA profiling in T reg cells with paired genotypes revealed a positive correlation of up-or down-regulation of few key signaling pathways with the severity of autoimmune phenotypes of T reg cell knockout mice.

Results
Fatal inflammation in Foxp3 Cre Ube2m fl/fl ; Ube2f fl/fl mice We have previously reported the phenotypical changes caused by individual deletion of Ube2m or Ube2f in T reg cells, and found that while the Foxp3 Cre ;Ube2f fl/fl mice had no visible phenotypic changes, the Foxp3 Cre ;Ube2m fl/fl mice developed severe inflammation disorders, where ~50% of mice died at about 4 months [15]. To study possible functional redundancy of 2 neddylation E2s in T reg cells, we intercrossed the single-null mice [15] and Foxp3 YFP-Cre (Foxp3 Cre ) mice [16] to generate conditional knockout mice with deletion of both Ube2m and Ube2f in T reg cells simultaneously (designated as "Foxp3 Cre ;Ube2m fl/fl ;Ube2f fl/fl "). Strikingly, the Foxp3 Cre ;Ube2m fl/fl ;Ube2f fl/fl mice suffered from an early-onset alterations of appearance, including reduced body size, collapsed ears, festered skin (Fig. 1, A and B), and a significantly shortened life span with 100% of death rate at p55 (Fig. 1C). In Foxp3 Cre ;Ube2m fl/fl ;Ube2f fl/fl mice at ~p20, the organ hematoxylin and eosin (H&E) staining revealed lymphocyte infiltration in liver, lung, kidney, stomach, colon, and skin (Fig. 1D). The autopsy revealed swollen peripheral immune organs, including lymph nodes and spleens, although the cellularity did not reach the statistically significant level ( Fig. 1E and Fig. S1A). A more detailed characterization revealed a robust activation of immune cells from Foxp3 Cre ;Ube2m fl/fl ;Ube2f fl/fl mice (~p20), as evidenced by a decreased ratio of CD4 + /CD8 + T cells (Fig. S1B) and increased proportion of effector/memory T cells (CD44 hi CD62L lo , T eff/mem cells) among conventional T cells (CD4 + Foxp3 − , T con cells) ( Fig.  1F and Fig. S1C). Such fatal inflammatory disorders recaptured the phenotypes observed in mice with depleted T reg cells [14] in severity, indicating a pivotal role of neddylation E2s in the T reg cells.

Impaired suppressive functions and altered transcriptome of Ube2m&Ube2f-deficient T reg cells
Given that both the T reg /CD4 + ratios (Fig. 2, A and B) and absolute numbers of T reg cells (Fig. 2C) are largely similar in peripheral lymph nodes derived from either Foxp3 Cre ;Ube2m fl/fl ;Ube2f fl/fl or Foxp3 Cre control mice around p20, we hypothesized that the severe inflammation disorders observed in Foxp3 Cre ;Ube2m fl/fl ;Ube2f fl/fl mice at the same age is most likely attributable to the loss of the suppressive function of T reg cells. To test this hypothesis, we performed an in vivo suppression assay in immunodeficient Rag1 −/− mice. Transfer of naive T (T nai ) cells (CD4 + Foxp3 − CD44 lo CD62L hi ) into Rag1 −/− recipients led to a wasting disease with colitis within 4 weeks. We found that such colitis was prevented by simultaneous transfer of T reg cells from wild-type mice, but not from Ube2m&Ube2f-deficient mice (Fig. 2D), demonstrating that double deletion of neddylation E2s impairs the suppressive function of T reg cells.
To define the underlying mechanism, we performed RNA profiling and analyzed the transcriptome alterations caused by Ube2m&Ube2f deletion in T reg cells. Considering that the inflammation in Foxp3 Cre ;Ube2m fl/fl ;Ube2f fl/fl mice would affect the transcript signature of T reg cells, we generated the inflammation-free female mice, in which one X chromosome expresses wild-type Foxp3 allele (Foxp3 wt ) and the other expresses Foxp3 Cre mutant (designated as Foxp3 Cre/wt ). The random inactivation of one X chromosome in each somatic cell caused about half of T reg cells to express Foxp3 Cre mutant, so Ube2m and Ube2f genes were deleted in these T reg cells; meanwhile, the other T reg cells expressed normal Foxp3 wt allele and were kept absolutely normal. Due to the protection of the normal T reg cells, the Foxp3 Cre/wt ;Ube2m fl/fl ;Ube2f fl/fl female mice are inflammation-free and healthy, so the CD4 + YFP + T reg cells in Foxp3 Cre/wt ;Ube2m fl/fl ;Ube2f fl/fl mice were deficient of Ube2m&Ube2f and avoided the suffering from inflammation at the same time.
We then sorted the CD4 + YFP + T reg cells from Foxp3 Cre/wt ; Ube2m fl/fl ;Ube2f fl/fl and Foxp3 Cre/wt mice, respectively, followed by transcriptome analysis. The Ube2m&Ube2f-deficient T reg cells (Fig. S2A) showed significant alterations in transcriptome, with 714 genes up-regulated and 1,289 genes down-regulated with fold change (Fc) > 1.5 and P value < 0.05 (Fig 2E and Fig.  S2, B and C). Among them, many T reg cell function-related genes were down-regulated, including Il10 [16] and Cst7 [17] (suppressive cytokines); Entpd1, Nt5e [18], and Il2r [19] (regulators of immune cell metabolism); Icos [20], Ctla4 [21], Pdcd1 [22], Lag3 [23], Nrp1 [24], and Tigit [25] (inhibitor via dendritic cells); and Gzmb [26] and Lgals3 [27] (apoptosis inducer)    ( Fig. 2F). Altered expression in migration-associated surface molecules was also observed in Ube2m&Ube2f-deficient T reg cells, such as down-regulation of Cxcr3, Cxcr5, Cxcr6, and Cd44 [28] and up-regulation of Sell [29] (Fig. 2F). On the other hand, among the top 20 genes up-regulated in Ube2m&Ube2fdeficienct T reg cells (Fig. S2D), none of them are previously known to negatively regulate T reg cells upon induction, which is certainly an interesting subject for future investigation. Gene set enrichment analysis (GSEA) of the altered genes revealed dramatic changes in multiple pathways upon Ube2m&-Ube2f deficiency (Fig. S2E), specially down-regulation of T regrelated pathways, such as cytokine-cytokine receptor interaction [30] (Fig. 2G) and T helper 17 (T H 17) cell differentiation [31] (Fig. S2F). Some up-regulated pathways were also seen in Ube2m&Ubef-deficienct T reg cells, including DNA replication and repair, cell cycle, and metabolisms of amino acids tyrosine, phenylalanine, and tryptophan, as well as cellular senescence ( Fig. S2E and Fig. 2H). How these pathways, upon double knockout of neddylation E2s, are up-regulated and how they functionally regulate T reg cells are the open and interesting questions for future investigation, particularly for possible involvement of senescence pathway.  [15] and Foxp3 Cre ;Ube2m fl/fl ;Ube2f fl/fl mice revealed a much shortened life span in double E2-null mice (Fig. 3A), indicating a significant contribution of Ube2f in the maintenance of T reg cells to the overall survival of mice. On the other hand, nondetectable phenotypic change in Ube2f T reg -depleted mice [15] indicates that the function of Ube2f in T reg cells is fully compensated by Ube2m. Thus, 2 neddylation E2s are functionally redundant in the regulation of T reg cells. While the main functions of Ube2f in T reg cells are compensated by Ube2m, Ube2f fails to functionally compensate Ube2m.

Phenotype and transcriptome comparison between
We next compared the transcriptome alterations between Ube2m&Ube2f-deficient and Ube2m-deficient T reg cells [15], along with the Foxp3 Cre control. We normalized 2 sets of data (each from 3 individual mice, run at 2 different time periods), followed by unsupervised cluster analysis. The results showed that the transcriptional profile patterns were largely comparable in Foxp3 Cre control T reg cells from 2 batches of experiments ( Fig. 3B), suggesting a reproducibility of the experiments. The profiling comparison revealed 5 clusters (groups) of genes with altered expressions among 3 groups with an overall greater change seen in Ube2m&Ube2f double-deficient T reg cells (Fig. 3B).
GSEA analysis of the genes specially altered in double-null T reg cells revealed the changes in multiple pathways (Fig. S3B), and some down-regulated pathways are known to be involved in T reg cell regulation, such as cytokine-cytokine receptor interaction [30] (Fig. 3D) and T H 17 cell differentiation [31] (Fig. S3C), whereas lysosome pathway, which is negatively correlated with function of T reg cells [42], along with cell cycle and phosphatidylinositol 3-kinase/Akt pathways, is up-regulated in Ube2m&Ube2f double-deficient T reg cells (Fig. S3, B and D). Thus, these greater alterations in gene expression are associated with and likely contributed to severer inflammatory disorders in Foxp3 Cre ;Ube2m fl/fl ;Ube2f fl/fl mice than in Foxp3 Cre ;Ube2m fl/fl mice.

Early-onset fatal inflammatory disorders in
Foxp3 Cre ;Rbx1 fl/fl ;Sag fl/fl mice We next determined potential functional redundancy in T reg cells between 2 E3s, Rbx1 and Sag/Rbx2, dual for both neddylation and ubiquitylation by CRLs. We crossed the Rbx1-flox, Sag-flox, and Foxp3 YFP-Cre (Foxp3 Cre ) mice [16] to generate the Foxp3 Cre ;Rbx1 fl/fl ;Sag fl/fl mice, with simultaneous deletion of both Rbx1 and Sag, the only 2 known catalytic subunits of CRLs, in T reg cells. The Foxp3 Cre ;Rbx1 fl/fl ;Sag fl/fl mice phenocopied the mice ablated of T reg cells in vivo [14], with smaller body size (Fig. 4, A and B), much shortened life span (Fig. 4C), inflammatory changes in multiple organs (Fig. 4D), swollen peripheral immune organs ( Fig. 4E and Fig. S4A), and robust activation of immune cells ( Fig. 4F and Fig. S7, B and C). Thus, Rbx1 and Sag are absolutely essential for the maintenance of T reg cell fitness.
We then performed the in vivo suppression assay in Rag1 −/− immunodeficient mice to confirm the role of Rbx1&Sag deficiency in the suppressive function of T reg cells and found that the Rbx1&Sag-deficient T reg cells failed to prevent the colitis induced by T nai cells (Fig. 5A), indicating impaired suppressive function. Thus, despite that the ratios and number of Rbx1&Sag-deficient T reg cells were dropped dramatically around p20 in vivo (Fig. 5, B to D), the severe inflammation disorder in Foxp3 Cre ;Rbx1 fl/fl ; Sag fl/fl mice was not the simple consequence caused by the decreased number of T reg cells.

Earlier occurrence of inflammatory disorder in Foxp3
Cre ;Rbx1 fl/fl ;Sag fl/fl mice than in Foxp3 Cre ;Rbx1 fl/fl mice We next addressed the question of functional redundancy between 2 E3s in T reg cells. Since mice with Rbx1-deficient T reg cells already had an early-onset fatal phenotype, whereas mice with Sag-deficient T reg cells had no phenotype [15], we first compared the survival of mice with 2 genotypes, Rbx1&Sag double deletion versus Rbx1 single deletion [15], in T reg cells and found no difference, both having early-onset fatal death (Fig. S6A). Moreover, no difference was found in the activation of immune response when mice were near moribund at age of ~20 days (Fig. S6B). We then focused on potential difference in inflammatory responses occurring at the very early stage of 8 days after the birth between 2 types of mice. Compared to the Foxp3 Cre control, the ratio of T eff/mem cells (CD4 + Foxp3 − CD44 hi CD62L lo ) among T con cells (CD4 + Foxp3 − ) is elevated dramatically in Rbx1null or Rbx1&Sag double-null mice (Fig. 6A), indicating a robust immune activation. A substantially higher level was seen in double-null mice (Fig.A and B), demonstrating a more severe immune overactivation in vivo. Thus, like the Ube2m-Ube2f E2 pair, the Rbx1-Sag E3 pair also showed a functional redundancy in T reg cells, although it is minor due to the dominant effect of Rbx1 to mask the Sag effect. Collectively, the Sag function is being fully compensated by Rbx1, whereas the Rbx1 function cannot be compensated by Sag in T reg cells.

Rbx1&Sag double deficiency in comparison to Rbx1 single deficiency in T reg cells
We further compared the transcriptome alterations between Rbx1&Sag-and Rbx1-deficient T reg cells, along with Foxp3 Cre control T reg cells, generated previously [15] and in this study, which showed large reproducibility (Fig. 6C, first 6

cells; and (d) genes up-regulated in both
Rbx1&Sag double-deficient and Rbx1 single-deficient T reg cells (Fig. 6C). Many group 2 genes are related with T reg cell regulation, including Cxcr3 [44], Itch [45], Il7rb [46], and Il10rb [34] (Fig. 6D). Among the top 20 up-regulated genes in group 3 (Fig.  S7A), none of them are known to negatively regulate T reg cell function upon induction, and Bach2 in group 3 (not in the list of top 20 genes) appears to be the only gene known as a negative regulator of T reg cells [43] (Fig. 6E). GSEA analysis of the genes specially altered in Rbx1&Sagdeficient T reg cells revealed changes of multiple pathways (Fig.  S7B), where cytokine-cytokine receptor interaction was downregulated (Fig. 6F), which are involved in T reg cell regulation [30] and up-regulation of multiple pathways, such as the olfactory transduction pathway (Fig. 6G). Taken together, it appears that many genes are associated with and may contribute to severer inflammatory disorder seen in Foxp3 Cre ;Rbx1 fl/fl ;Sag fl/fl mice.

Phenotypes and transcriptome comparisons of mice and T reg cells with double E2 versus double E3 deficiency
It is well established that cullin neddylation is essential for activation of CRLs, and Rbx1 and Sag serve as dual E3 for both neddylation and CRL-medicated ubiquitylation [47,48]. To verify the functional relationship between neddylation and CRLs in T reg cells, we compare the degree of autoimmune disorders and transcriptional alterations caused by Ube2m&Ube2f versus Rbx1&Sag deficiency in T reg cells.
Although both Foxp3 Cre ;Ube2m fl/fl ;Ube2f fl/fl mice and Foxp3 Cre ; Rbx1 fl/fl ;Sag fl/fl mice suffered from severe autoimmune disorders with much shortened life span, the degree of severity still differs. In fact, the survival of Foxp3 Cre ;Rbx1 fl/fl ;Sag fl/fl mice is significantly shorter than that of Foxp3 Cre ;Ube2m fl/fl ;Ube2f fl/fl mice, with P = 0.0350 (Fig. 7A). In the peripheral lymph nodes from mice at p19 to p21, the ratios of T eff/mem cells (CD4 + Foxp3 − CD44 hi CD62L lo ) among T con cells (CD4 + Foxp3 − ) are higher in Foxp3 Cre ;Rbx1 fl/fl ;Sag fl/fl mice than in Foxp3 Cre ; Ube2m fl/fl ; Ube2f fl/fl mice, although the difference is not statistically significant due to variations among mice (Fig. 7B). Thus, double E3-null mice had more dramatic immune overactivation than double E2-null mice.
The comparison of transcriptome data of CD4 + YFP + T reg cells from Foxp3 Cre/wt , Foxp3 Cre/wt ;Ube2m fl/fl ;Ube2f fl/fl , and Foxp3 Cre/wt ;Rbx1 fl/fl ;Sag fl/fl mice revealed 6 groups of altered genes (up or down), unique to either double E2-deficient or double E3-deficient or both deficient T reg cells (Fig. 7C). We mainly focused on genes uniquely altered in double E3-null or double E2-null T reg cells. Among selectively down-regulated genes in double E3-null T reg cells, many are related to function or regulation of T reg cells (Fig. 7D), whereas among up-regulated genes (Fig. S8A), none of them are known to negatively regulate T reg cells upon induction. Similarly, among selectively altered genes (down and up) in double E2-null T reg cells (Fig.  S9, A and B), few down-regulated, but not up-regulated, genes were known to regulate T reg cell functions. Nevertheless, the changes unique in double E3-null T reg cells likely represent CRL-dependent mechanism, whereas the changes unique in double E2-null T reg cells likely represent CRL-independent, but neddylation-dependent, mechanisms, consistent with our recent studies [49][50][51].
Similar GSEA pathway analysis of 2 groups of genes unique to double E3-deficient T reg cells revealed remarkable alterations in many pathways (Fig. S8B), including down-regulation of Hippo signaling pathway (Fig. 7E), which plays an important role in the maintenance of T reg cell function [52], and of pathways related to lysosome, as well as up-regulation of olfactory transduction and metabolisms of arachidonic acid and retinol (Fig. S8, B and C), which are previously unknown in regulation of the T reg cell functions.
GSEA pathway analysis of 2 groups of genes unique to double E2-deficient T reg cells also revealed alterations in a number of pathways (Fig. S9C), including down-regulation of ubiquitinmediated proteolysis and up-regulation of DNA replication pathways (Fig. S9, D and E). Notably, the number of genes unique to double E2-deficient T reg cells is relatively smaller than in double E3-deficient T reg cells, which is consistent with the observation that Foxp3 Cre ;Rbx1 fl/fl ;Sag fl/fl mice suffer more dramatic autoimmune disorders than do the Foxp3 Cre ;Ube2m fl/fl ;Ube2f fl/fl mice. Collectively, the early-onset fatal disorders seen in both double E2-and double E3-deficient mice indicate functional similarity of neddylation and CRLs in regulation of T reg cells. More severe inflammatory disorders in double E3-deficient mice suggest a neddylation-independent function of CRLs in T reg cells.

Discussion
In this study, we investigated the role of the neddylation-CRL axis in functional regulation of T reg cells using several in vivo conditional knockout mouse models. Double deletion of neddylation E2s Ube2m&Ube2f or E3s Rbx1&Sag in T reg cells leads to the impairment of suppressive function of T reg cells and results in severe autoimmune disorders, fully demonstrating the pivotal role of the neddylation-CRL axis in the maintenance of T reg cell fitness. Thus, the neddylation-CRL axis is essential for T reg cell functions as the new key regulators of T reg cells beyond Foxp3 [11][12][13].
The functional independence of the Ube2m-Rbx1 and Ube2f-Sag axes was previously demonstrated both biochemically [4,53] and biologically [8,9]. Specifically, our recent study revealed that the Ube2m-Rbx1 axis plays a pivotal role in functional regulation of T reg cells, while the Ube2f-Sag axis is dispensable in T reg cells at the steady status [15]. This T reg -based functional study also supports the notion of functional independence between the Ube2m-Rbx1 and Ube2f-Sag axes.
Here, we used double knockout of either neddylation E2s or E3s in T reg cells and showed that both E2s and E3s are functionally redundant in T reg cells, as evidenced by the following: (a) The phenotypes of autoimmune disorders in Foxp3 Cre ;Ube2m fl/fl ;Ube2f fl/fl (double E2 knockout in T reg cells) mice are much more severe than in Foxp3 Cre ;Ube2m fl/fl (single E2 knockout in T reg cells) mice with much shortened life span (Fig. 3A). (b) The same autoimmune phenotypes are severer in Foxp3 Cre ;Rbx1 fl/fl ;Sag fl/fl (double E3s knockout in T reg cells) mice than in Foxp3 Cre ;Rbx1 fl/fl (single E3 knockout in T reg cells) mice at the very early age only ( Fig. 6A and B). We further found that the autoimmune phenotypes were much severer in Foxp3 Cre ; Rbx1 fl/fl ;Sag fl/fl mice than in Foxp3 Cre ;Ube2m fl/fl ;Ube2f fl/fl mice (Fig. 7A).
We conclude from these results that (a) the Ube2f-Sag axis plays a role in T reg cells, which is, however, compensated by the Ube2m-Rbx1 axis. More specifically, the autoimmune phenotypes by Ube2f depletion were largely compensated by Ube2m, whereas the same autoimmune phenotypes by Sag depletion were minimally compensated by Rbx1, due to the dominant role of Rbx1 in T reg cells. (b) The functional redundancy in T reg cells is in one-way direction from the Ube2m-Rbx1 axis to the Ube2f-Sag axis, indicating much more significant role of CRL1 to CRL4 than of CRL5. (c) Rbx1 and Sag play neddylationdependent and neddylation-independent roles in regulating T reg cell fitness.
To pursue the possible underlying mechanism, we performed the comprehensive transcriptome profiling analyses of T reg cells derived from the wild-type control or double knockout mice, and compared between single-null versus double-null T reg cells for both E2s or E3s, respectively. Compared to the wild-type control T reg cells, T reg cell deficiency of Ube2m&Ube2f caused dramatic alterations in mRNA transcriptome. Many genes (such as Il10, Cst7, and Cxcr3 to Cxcr6) and pathways (e.g., cytokinecytokine receptor interaction and T H 17 cell differentiation), previously known to regulate T reg cells, were down-regulated ( Fig. 2 and Fig. S2), supporting the lost-of-function phenotypes. On the other hand, interestingly, many up-regulated genes (none of the top 20 genes) (Fig. S2D), with the exception of Sell (Fig.  2F) [29], or pathways (Fig. S2E) are previously unknown to negatively regulate T reg cell function upon activation. Nevertheless, some up-regulated pathways, such as cellular senescence (Fig.  2H), could contribute to the loss-of-function phenotype and deserve detailed future investigation.
Likewise, compared to the wild-type control T reg cells, T reg cell deficiency of Rbx1&Sag also caused dramatic alterations in mRNA transcriptome. Many genes (such as Il10, Cst7, Entpd1, and Nt5e) and pathways (again cytokine-cytokine receptor interaction and T H 17 cell differentiation), previously known to regulate T reg cells, were down-regulated ( Fig. 5 and Fig. S5), further supporting the lost-of-function phenotypes. On the other hand, interestingly, many up-regulated genes (again none of the top 20 genes) or pathways are previously unknown to regulate T reg cell function upon activation. One exception is Bach2, a known negative regulator of T reg cell functions [43], which is up-regulated in Rbx1&Sag-deficienct T reg cells (Fig.  5G). However, it is completely unknown how some up-regulated pathways, such as those involving olfactory transduction ( Fig.  5I and Fig. S5E), negatively regulate T reg cell functions.
Transcriptome comparison between T reg cell deficiency of Ube2m&Ube2f versus Ube2m also showed many alterations. A list of 11 genes (e.g., Ccr4, Egr2, and Il10), uniquely downregulated in double E2 null (Fig. 3C), are all known to regulate T reg cell functions. Again, the pathways associated with cytokine-cytokine receptor interaction and T H 17 cell differentiation were also down-regulated uniquely to double E2 null T reg cells ( Fig. 3D and Fig. S3C). Among top 20 up-regulated genes (Fig. S3A), Smad7 is the only one reported to inhibit T reg cell differentiation in rheumatoid arthritis upon up-regulation by the EZH2-FOXP3-RUNX1 axis [41]. The other up-regulated genes, along with up-regulated pathways (Fig. S3B), unknown to negatively regulate T reg cell functions, open a new window to study neddylation involvement in T reg cell functions, for example, how up-regulation of lysosome pathway ( Fig. S3B and D) contributes to the loss-of-function phenotype of T reg cells.
Transcriptome comparison between T reg cell deficiency of Rbx1&Sag versus Rbx1 also showed many alterations, but to a lesser extent (Fig. 6C and Fig. S7). Down-regulation of Cxcr3, Fas, Egr2, and Itch and of cytokine-cytokine receptor interaction pathway appeared to be unique to double E3-null T reg cells (Fig. 6, D and F). Among up-regulated genes and pathways (Fig. S7), Bach2 and olfactory transduction (Fig. 6, E and G) appeared to be unique. Given a minor phenotypic difference between Rbx1/Sag double-null and Rbx1 single-null mice, these alterations are unlikely to play a major role in the lossof-function phenotypes.
Finally, we compared the phenotypes of Foxp3 Cre ;Ube2m fl/fl ; Ube2f fl/fl (double E2-null) mice with Foxp3 Cre ;Rbx1 fl/fl ;Sag fl/fl (double E3-null) mice and found that double E3-null mice suffer more severe autoimmune disorders with a shorter life span (Fig. 7A). Consistently, the transcriptional alterations caused by Rbx1&Sag deficiency in T reg cells are more dramatic than those in Ube2m&Ube2f-deficient T reg cells, with a more remarkable decrease of T reg cell regulatory genes and down-regulation of Hippo signaling pathway (Fig. 7, D and E), which was known to play an important role in the maintenance of T reg cell function [52]. Among up-regulated genes and pathways, such as olfactory transduction and metabolisms of retinol and arachidonic acid, which are unique to double E3-null ( Fig. S8B and C), it is completely unknown whether and how they negatively regulate T reg cell function, nor their contribution to severe phenotypes. In addition, we also identified genes and pathways unique to double E2 knockout T reg cells (Fig. 7C and Fig. S9). These alterations are likely independent of inactivation of Rbx1/ Sag-CRLs but involve neddylation modification of non-cullin substrates via other dual E3s for both neddylation and ubiquitylation [49][50][51].
It is not surprising that many changes in gene expression occur in these T reg conditional knockout models, given the fact that neddylation E2/E3s are required for cullin neddylation to activate CRLs, which are responsible for ubiquitylation and degradation of ~20% cellular proteins doomed for proteasome degradation [54], thus regulating many biochemical and biological processes. Mechanistically, it is anticipated that inactivation of CRLs, resulting from depletion of neddylation E2/E3, will cause the accumulation of many cellular substrates, including (a) transcription factors and repressors, which would directly affect gene expression, and (b) other signal molecules and cell cycle regulators, which would indirectly affect gene expression. Unfortunately, given that T reg cells are a rare population in vivo and no in vitro cell lines were established, it is technically challenging to define the detailed underlying mechanisms by these altered genes and pathways in our T reg cell conditional knockout setting. Nevertheless, our study defined genes and pathways responsible for both neddylation-dependent and neddylation-independent roles for neddylation E2s and dual E3s.
In summary, in this study, we used mouse models of T reg cell double knockout of neddylation E2s or E3s and made 3 major findings: (a) Neddylation enzymes play an absolute essential role in the maintenance of T reg cell fitness and their dual disruption causes severer autoimmune phenotypes; (b) there is previously unknown functional redundancy between Ube2m and Ube2f, and to a lesser extent between Rbx1 and Sag; and (c) dual E3s Rbx1 and Sag have neddylation-dependent and neddylation-independent role. Finally, our study has sound physiological and pathological relevance to autoimmune diseases, given the fact that a variety of human autoimmune diseases, including systemic lupus erythematosus, inflammatory bowel disease, and rheumatoid arthritis, are subjected to fine regulation by CRLs [55], whose activation requires neddylation E2, UBE2M/UBE2F, and E3 RBX1/RBX2.
Mice were fed in specific pathogen-free (SPF) conditions. All animal experiments were approved by the Animal Ethics Committee of Zhejiang University; animal care was provided in accordance with the principles and procedures by the regulatory standards at Zhejiang University Laboratory Animal Center.

Flow cytometry
Peripheral lymph nodes and spleens from indicated mice were grinded into single cells. Cells were washed in phosphatebuffered saline containing 2% (w/v) fetal bovine serum and then stained with indicated antibodies for analysis of surface proteins. For intracellular proteins, cells were fixed and permeabilized with Pharmingen Transcription Factor Buffer Set (BD Pharmingen, 562574). Flow cytometry was performed on CytoFLEX LX (Beckman).

Cell sorting
The CD4 + T cells were isolated from the single-cell suspension of peripheral lymph nodes and spleens by Mouse CD4 T Lymphocyte Enrichment Set-DM (BD Biosciences, 558131), followed by fluorescence-activated cell sorting to purify T reg cells (CD4 + YFP + ) and/or T nai cells (CD4 + YFP − CD44 lo CD62L hi ) with purities >99%, performed on SONY Cell Sorter (SH800S).

Transcriptome profiling
CD4 + YFP + T reg cells were sorted from the peripheral lymph nodes and spleens of indicated mice (8 to 12 weeks old). To generate enough materials, 2 to 3 mice were pooled for one sample. RNA was purified from the sorted cells with the miRNeasy Mini Kit (Qiagen, 21704). The RNAs were reverse-transcribed, amplified, and labeled (Affymetrix GeneChip Pico Kit, 703308) to achieve enough cDNAs. Then, the cDNA samples were hybridized to Clariom S Arrays, mouse (902931). The Applied Biosystems Expression Console Software 1.4 was employed to analyze the microarray datasets.

Statistical analysis
The P values were calculated by Mann-Whitney test, 2-tailed unpaired Student's t test, using GraphPad Prism software. Mouse survival and respective P values were analyzed by the log-rank test. The statistical significance was evaluated by P < 0.05. All error bars represent SEM (n ≥ 3).
Natural Science Foundation of China (82172699 and 81801567 to D.W. and U22A20317 and 92253203 to Y.S.), and Zhejiang Provincial Natural Science Foundation of China (LY21H100005 to D.W. and LD22H300003 to Y.S.). Author contributions: D.W. designed and performed experiments, analyzed data, and wrote the manuscript. Y.S. conceived project, designed experiments, analyzed data, wrote and finalized the manuscript, and oversaw the project. Competing interests: The authors declare that they have no competing interests.

Data Availability
The microarray data generated in this study have been deposited in Gene Expression Omnibus under accession code GSE237499.