Abstract

   SDS22 forms an inactive complex with nascent protein phosphatase PP1
   and Inhibitor-3. SDS22:PP1:Inhibitor-3 is a substrate for the ATPase
   p97/VCP, which liberates PP1 for binding to canonical regulatory
   subunits. The exact role of SDS22 in PP1-holoenzyme assembly remains
   elusive. Here, we show that SDS22 stabilizes nascent PP1. In the
   absence of SDS22, PP1 is gradually lost, resulting in substrate
   hyperphosphorylation and a proliferation arrest. Similarly, we identify
   a female individual with a severe neurodevelopmental disorder bearing
   an unstable SDS22 mutant, associated with decreased PP1 levels. We
   furthermore find that SDS22 directly binds to Inhibitor-3 and that this
   is essential for the stable assembly of SDS22:PP1: Inhibitor-3, the
   recruitment of p97/VCP, and the extraction of SDS22 during holoenzyme
   assembly. SDS22 with a disabled Inhibitor-3 binding site co-transfers
   with PP1 to canonical regulatory subunits, thereby forming
   non-functional holoenzymes. Our data show that SDS22, through
   simultaneous interaction with PP1 and Inhibitor-3, integrates the major
   steps of PP1 holoenzyme assembly.

   Subject terms: Holoenzymes, Cell signalling, Proteomics,
   Neurodevelopmental disorders
     __________________________________________________________________

   SDS22 is a poorly characterized regulator of PP1. Here, the authors
   show that SDS22 prevents the aggregation of nascent PP1 and coordinates
   its stepwise incorporation into functional holoenzymes.

Introduction

   Protein phosphatase-1 (PP1) is a member of the PPP family of protein
   Ser/Thr phosphatases^[56]1. It is expressed in all eukaryotic cells and
   dephosphorylates a wide array of proteins. The selectivity of PP1
   largely stems from its association with one or two regulatory subunits,
   referred to as “Regulatory-Interactors-of-Protein-Phosphatase-One”
   (RIPPOs), to form specific PP1 holoenzymes^[57]2,[58]3. Vertebrates
   express >200 structurally unrelated RIPPOs, accounting for the huge
   diversity of PP1 holoenzymes, each with its own subset of substrates
   and mechanism of regulation. RIPPOs specify the function of associated
   PP1 by blocking or extending its active site or a substrate-binding
   groove, by recruiting substrates via dedicated domains, and/or by
   targeting the phosphatase to a specific subcellular location that
   contains a subset of substrates. The PP1-binding domain of RIPPOs is
   usually (largely) structurally disordered and contains one to several
   short linear motifs (SLiMs) that combine to create a dynamic,
   high-affinity binding interface for PP1. The most widespread SLiM,
   present in a large majority of RIPPOs, is the so-called RVxF motif,
   which binds to a hydrophobic groove on PP1 that is remote from the
   active site and functions as a PP1-anchoring motif.

   SDS22 (PPP1R7) and Inhibitor-3 (HCGV/IPP3/PPP1R11/TCTEX5) are the two
   first-evolved and most widespread RIPPOs, hinting at their key role in
   the regulation of PP1^[59]4,[60]5. SDS22 is one of only a few RIPPOs
   with a structured PP1-binding domain. It largely consists of 12
   leucine-rich repeats (LRRs) that adopt a banana-shaped structure to
   generate a large PP1-interaction interface. In contrast, Inhibitor-3
   (I3) has an unstructured PP1-binding domain with an RVxF motif and a
   unique PP1-binding SLiM (CCC motif) with three consecutive
   cysteines^[61]6. Since SDS22 and I3 have non-overlapping PP1-binding
   sites, they can form a ternary complex with PP1^[62]7,[63]8. SDS22 and
   I3 are “non-canonical” RIPPOs in that they are not involved in
   substrate selection or subcellular targeting of PP1, and do not appear
   to be a component of functional PP1 holoenzymes^[64]5. Instead, they
   selectively bind to newly translated PP1, before its transfer to
   “canonical” RIPPOs^[65]8. SDS22 locks nascent PP1 in an inactive
   conformation that lacks one of two metals that are essential for
   catalysis^[66]9. In addition, the CCC motif of I3 occludes the active
   site, thereby preventing the access of substrates^[67]6. The inactive
   SDS22:PP1:I3 complex is a substrate for p97/VCP, an AAA^+ ATPase that
   uses ATP hydrolysis to co-extract I3 and SDS22, thereby freeing PP1 for
   association with canonical RIPPOs to form functional PP1
   holoenzymes^[68]8,[69]10–[70]12. p97/VCP directly targets I3, which is
   unfolded in the p97/VCP-hexamer channel, and this somehow results in
   the co-extraction of SDS22. Mature PP1 holoenzymes are highly dynamic
   because of continuous competition between canonical, SLiM-based RIPPOs
   for binding to the limited pool of PP1^[71]2,[72]13. Hence, the
   diversity of functional PP1 holoenzymes that accumulate in cells is
   largely determined by the relative abundance of canonical RIPPOs and
   their (regulated) affinity for PP1.

   Several key questions concerning the biogenesis of PP1 holoenzymes
   remain unanswered. Why does newly translated PP1 selectively form a
   ternary complex with SDS22 and I3? What prevents I3 in the SDS22:PP1:I3
   complex from being dynamically exchanged for other SLiM-based RIPPOs,
   like RepoMan and MYPT1, that also have non-overlapping PP1-binding
   sites with SDS22? Why does the extraction of I3 from SDS22:PP1:I3, in
   contrast to the dissociation of other PP1:RIPPO complexes, require
   energy? What is the mechanism underlying the co-extraction of SDS22 and
   I3? What is the function of SDS22 in the ternary SDS22:PP1:I3 complex
   and in subsequent steps of PP1-holoenzyme assembly?

   Here, using specific cellular and molecular research tools, we show
   that the binding of SDS22 to nascent PP1 is required to stabilize PP1.
   We describe an individual with an unstable SDS22 mutant and decreased
   levels of PP1. We also demonstrate that SDS22 and I3 bind
   simultaneously to PP1 and to each other, making their recruitment
   selective and irreversible, and their co-extraction energy-dependent.
   Finally, we show that p97/VCP is only recruited by SDS22:PP1:I3 and
   that the co-extraction of I3 and SDS22 is essential to generate
   functional PP1 holoenzymes. Our data provide key molecular insights
   into the coordination of the major steps of PP1-holoenzyme assembly by
   SDS22.

Results

SDS22 depletion results in a cell-cycle arrest and a post-mitotic
re-attachment failure

   To explore the function of SDS22 in mammalian cells, we engineered the
   human HCT116 colorectal carcinoma-cell line, using a CRISPR/Cas9
   approach, for the inducible proteolytic degradation of endogenous
   SDS22^[73]14. The designed guide RNA targeted the Cas9 protein to a PAM
   site at the 5′-end of the stop codon of the SDS22-encoding PPP1R7
   alleles (Supplementary Fig. [74]1a). Donor DNA constructs with 5′- and
   3′-homology arms flanking a miniature Auxin-Inducible Degron (mAID),
   the fluorescent mClover protein (mClover), and a neomycin- or
   hygromycin-resistance cassette, were used as templates for homologous
   recombination-mediated repair of Cas9-induced double-strand breaks to
   generate modified PPP1R7 alleles encoding a SDS22-mAID-mClover fusion.
   The used HCT116 cell line also contains a transgene at the AAVS1 locus
   that expresses the F-box protein TIR1 from Oryza sativa in a
   doxycycline (Dox)-dependent manner. Addition of Dox to induce the
   expression of TIR1 and IAA (indol-3-acetic acid, a synthetic analog of
   auxin) to recruit the endogenous SCF-type E3 ubiquitin ligase to mAID
   leads to the ubiquitination and proteasomal degradation of
   SDS22-mAID-mClover in the SDS22-degron cell line (Fig. [75]1a). The
   SDS22 fusion was already partially degraded 2 h after Dox/IAA addition
   and became undetectable after 24–48 h, as shown by both immunoblotting
   (Fig. [76]1b) and mClover-fluorescence imaging (Supplementary
   Fig. [77]1b). However, Dox/IAA addition did not affect the level of
   SDS22 in the parental cell line (Supplementary Fig. [78]1c). A washout
   of IAA in the degron cell line resulted in the re-appearance of
   SDS22-mAID-mClover after 48 h (Fig. [79]1b and Supplementary
   Fig. [80]1b), attesting to the reversibility of SDS22 depletion.

Fig. 1. The degradation of SDS22 leads to a proliferation arrest.

   [81]Fig. 1
   [82]Open in a new tab

   a Scheme of Dox/IAA-induced degradation of SDS22-mAID-mClover. b Time
   course and reversibility of SDS22-mAID-mClover degradation. Dox was
   added 18 h before IAA (upper panel). SDS22-mAID-mClover re-accumulation
   after Dox/IAA washout (bottom panel). c The effect of SDS22 depletion
   on cell proliferation (IncuCyte). The cells were untreated or
   pretreated with Dox and/or IAA for 6 h, before the start of scanning.
   The solid lines are means of three technical replicates and the shaded
   areas show the SD range from a single data set, representative for
   three experiments. d The effect of SDS22 degradation on cell-cycle
   progression. Dox/IAA-treated cells were arrested in G[1]/S, released
   from thymidine, and analyzed by flow cytometry after propidium iodide
   (PI) staining. e Quantification of S → M duration in SDS22-degron
   cells, (non-)treated with Dox/IAA for 8 h, by time-lapse imaging of
   individual cells. S → M duration was defined as the time between
   release from a single-thymidine arrest and cell rounding. Cells were
   imaged at 10-min intervals. The P value is from two-sided unpaired
   t-test (n = 78 cells for each conditions). f Cells treated as in (e),
   but scored by time-lapse imaging for the duration of M-phase (time from
   cell rounding to either cell flattening or formation of loosely
   attached cell clumps). The P value is from two-sided unpaired t-test
   (n = 73 cells for each conditions). g Volcano plot of RNA-sequencing
   data (n = 3) showing differentially expressed genes (DEGs) in
   Dox/IAA-treated (48 h) SDS22-degron versus parental cell lines. The P
   values of the likelihood ratio test were calculated by edgeR. The
   cut-off for the DEGs was set at P value < 0.05 and logFC > 1. h Dot
   plot of the gene-ontology (GO) pathway enrichment analysis for the DEGs
   i Mitotic phenotypes in SDS22-depleted degron cells, as quantified by
   live time-lapse imaging. Details as in (d). The data are expressed as
   means ± SD (n = 5 independent experiments; >50 cells analyzed in each
   condition). P values were from two-sided unpaired t-test. j
   Morphological changes in parental and SDS22-degron cells, treated or
   not with Dox/IAA. Cells were fixed and stained for DAPI (blue) and
   α-tubulin (yellow). Scale bars, 10 μm.

   The Dox/IAA-induced depletion of SDS22 in the HCT116-degron cell line
   caused a nearly complete proliferation arrest, as detected by both
   IncuCyte live-cell analysis (Fig. [83]1c) and SRB assays (Supplementary
   Fig. [84]1d). However, Dox/IAA addition only had a minor effect on the
   proliferation of the parental cell line, and the addition of Dox or IAA
   alone did not affect the proliferation of either cell line. Genome-wide
   RNAi-screens ([85]https://depmap.org/portal/) also identified SDS22 as
   one of only a few RIPPOs that are important for the proliferation of
   hundreds of cancer-cell lines (Supplementary Fig. [86]1e). A
   flow-cytometry analysis of cells released from a G[1]/S arrest revealed
   that SDS22-depleted cells were severely delayed in the ensuing
   cell-cycle progression (Fig. [87]1d), accounting for their reduced
   proliferation (Fig. [88]1c). Time-lapse video imaging of individual
   cells released from a G[1]/S arrest showed that SDS22-degron cells that
   were still proliferating needed more time for transition from S- to
   M-phase (Fig. [89]1e) and for progression through M-phase
   (Fig. [90]1f). In the latter experiments, the total duration of the
   Dox/IAA-treatment was limited to 30 h to preclude a nearly total
   proliferation arrest, as seen with prolonged treatments (Fig. [91]1c).
   Global RNA-sequencing disclosed a similar number (≈600) of up- and
   downregulated genes in SDS22-depleted cells (Fig. [92]1g; Supplementary
   Data [93]1). However, a gene-ontology (GO) pathway analysis of the
   differentially expressed genes showed a strong enrichment for genes
   implicated in cell division and DNA replication/repair (Fig. [94]1h and
   Supplementary Fig. [95]1f), in accordance with the deficient cell-cycle
   progression of SDS22-depleted cells.

   In addition to a proliferation arrest, about 25% of SDS22-depleted
   cells were floating in the medium after a 48 h culture (Supplementary
   Fig. [96]1g). When the culture medium was refreshed, without prior cell
   splitting, a similar % of cells were floating after 48 h, indicating
   that adherent cells had lost their attachment and/or did not to
   re-attach after mitosis. Time-lapse video imaging indeed showed that
   SDS22-depleted cells often failed to re-attach upon completion of cell
   division (Fig. [97]1i) or formed loosely attached, slow-growing cell
   clumps (Fig. [98]1j). In conclusion, our data revealed that the removal
   of SDS22 from HCT116 cells culminates in severe cell-cycle progression
   defects and a post-mitotic cell re-attachment failure.

SDS22-depleted cells gradually lose PP1, resulting in substrate
hyperphosphorylation

   Since SDS22 is an established RIPPO, we subsequently investigated
   whether the severe phenotype of SDS22 depletion is due to changes in
   the phosphorylation of PP1 substrates, with a specific focus on
   substrates that are important for cell division and attachment. PP1 is
   required for spindle-assembly checkpoint (SAC) silencing during
   metaphase, once all kinetochores form correct amphitelic microtubule
   attachments^[99]15,[100]16. Moreover, in fission yeast, SDS22 is
   essential for the metaphase–anaphase transition^[101]17, consistent
   with a role in SAC silencing. To examine whether deficient SAC
   silencing accounts for the prolonged mitosis in SDS2-depleted cells, we
   measured the mitotic duration of HeLa cells before and after the
   knockdown of SDS22 and/or MAD2 (Fig. [102]2a and Supplementary
   Fig. [103]2a). MAD2 is a key component of SAC signaling and its removal
   results in a SAC override. As in SDS22-depleted HCT116 cells
   (Fig. [104]1f), the knockdown of SDS22 in HeLa cells also prolonged
   mitosis, and this phenotype was completely rescued by the simultaneous
   knockdown of MAD2 (Fig. [105]2a and Supplementary Fig. [106]2a). This
   showed that the prolonged mitosis of SDS22-depleted cells stems from a
   failure to silence the SAC. In accordance with this conclusion,
   time-lapse imaging revealed that the depletion of SDS22 in HeLa cells
   resulted in a prolonged metaphase (Supplementary Fig. [107]2b, [108]c),
   but had no effect on the prophase → metaphase duration (Supplementary
   Fig. [109]2d), arguing against rate-limiting effects of SDS22 depletion
   on achieving stable kinetochore–microtubule attachments. The extended
   mitosis of SDS22-depleted cells was rescued by expression of
   siRNA-resistant EGFP-tagged wildtype SDS22 (SDS22-WT) (Fig. [110]2b and
   Supplementary Fig. [111]2e). However, previously described PP1-binding
   mutants of SDS22 (SDS22-M)^[112]18 only partially rescued this
   phenotype, and the extent of the rescue was inversely correlated with
   the number (1→4) of mutated PP1-binding residues. These data suggested
   that the failure of SAC silencing in SDS22-depleted cells was due to a
   deficient dephosphorylation of SAC proteins by PP1. Consistent with
   this interpretation, a global phospho-proteome analysis disclosed
   hyperphosphorylation of key SAC components in SDS22-depleted HCT116
   cells, including KNL1, CDC20, BUB1, BUB3, PLK1 and Aurora B
   (Fig. [113]2c and Supplementary Fig. [114]2f; Supplementary
   Data [115]2). A GO-analysis of affected pathways in SDS22-depleted
   cells showed a clear overlap between the global phosphoproteome
   (Supplementary Fig. [116]2g) and RNA-sequencing data (Fig. [117]1g),
   with a strong enrichment for cell division (Fig. [118]2d, Supplementary
   Figs. [119]1f and [120]2g).

Fig. 2. Misregulation of PP1 in SDS22-depleted cells.

   [121]Fig. 2
   [122]Open in a new tab

   a Time-lapse imaging (10-min interval) of individual HeLa cells
   released from a single-thymidine arrest after treatment with siRNAs
   (siCTR, siMAD2 and/or siMAD2). P values are from two-sided Mann–Whitney
   U test (n = 15 cells for each condition). b Rescue of prolonged M-phase
   in SDS22-depleted HeLa cells by expression of EGFP-tagged SDS22-WT or a
   PP1-binding mutant of SDS22 (1M: E192A; 2M: E192A + E300A; 3M:
   F170A + E192A + E300A; 4M: F170A + F214A + E192A + E300A). P values are
   from two-sided Mann–Whitney U test (n = 25 cells/conditions). c
   Phosphopeptides from SAC components in SDS22-degron cells as compared
   to that in parental cells, treated with Dox/IAA. d Venn diagram of the
   affected biological processes (BP) after SDS22 degradation in the
   phosphoproteomics and RNA-sequencing data sets, as determined by
   GO-pathway analysis (left panel). The right panel shows the
   significance (−log[10] of the P values) of the 26 overlapping pathways.
   e Time-dependent degradation of SDS22 and ERM phosphorylation (pERM).
   IAA was added for the indicated times, in the presence of Dox, to
   induce the degradation of SDS22 (left). Also shown is the effect of a
   subsequent Dox/IAA (D/I) washout for 0 → 72 h on the recovery of SDS22
   protein and ERM dephosphorylation (right panel). f Immunofluorescence
   staining of SDS22-degron cells, either untreated or treated with
   Dox/IAA for 72 h. The fixed cells were stained for DNA (DAPI),
   phalloidin and pERM. The scale bars are 25 μm. g Cell-based assay of
   ERM dephosphorylation. Parental and SDS22-degron cells were incubated
   with Dox/IAA and thymidine. Calyculin A (25 nM, 30 min) served to
   maximize ERM phosphorylation. Subsequently, the cells were released in
   medium with 50 nM staurosporine to enable ERM dephosphorylation. h Time
   course of SDS22 depletion in the degron cell line and its corresponding
   effect on PP1 protein in cell lysates. The cell lysates were prepared
   from the combined attached plus floating cells. i Quantification of the
   data for PP1 shown in (h). The bars represent means ± SD (n = 3
   independent experiments). The P value is from two-sided unpaired
   t-test.

   ERM (Ezrin, Radixin and Moesin) proteins link the plasma membrane to
   the cortical actin cytokeleton^[123]19. Their phosphorylation at the
   mitotic entry stiffens the cell cortex, resulting in cell rounding,
   while their dephosphorylation by PP1 at the mitotic exit is required
   for cell flattening and re-attachment. In the HCT116-degron cell line,
   ERM proteins became strongly hyperphosphorylated upon depletion of
   SDS22, but their phosphorylation level normalized after the
   re-accumulation of SDS22 in the absence of Dox/IAA (Fig. [124]2e). We
   also noted a strong staining for phosphorylated ERM proteins (pERM) in
   SDS22-depleted cells (Fig. [125]2f and Supplementary Fig. [126]2h). ERM
   hyperphosphorylation was also detected in SDS22-depleted cells that
   were synchronized in prometaphase (nocodazole addition) and
   subsequently released by nocodazole washout (Supplementary
   Fig. [127]2i). To explore whether ERM hyperphosphorylation was caused
   by deficient ERM dephosphorylation, we performed a cell-based ERM
   phosphatase assay (Fig. [128]2g). Dox/IAA-treated parental and degron
   cells were first treated with the cell-permeable PPP-type phosphatase
   inhibitor calyculin A to obtain maximally phosphorylated ERM.
   Subsequently, the cells were resuspended in fresh medium with 50 nM of
   the kinase inhibitor staurosporine. This resulted in a rapid and
   complete dephosphorylation of ERM proteins in the parental cells.
   However, a portion of the ERM proteins in the degron cells remained
   phosphorylated, hinting at a dephosphorylation deficit. This difference
   between parental and degron cell lines was not seen, however, in cells
   that were not treated with Dox/IAA (Supplementary Fig. [129]2j).

   The above data demonstrated that PP1-mediated SAC silencing and ERM
   dephosphorylation is hampered in SDS22-depleted cells and in cells
   expressing a PP1-binding mutant of SDS22. As SDS22-associated PP1 is
   catalytically inactive^[130]7,[131]9, the contribution of SDS22 to SAC
   silencing and ERM dephosphorylation must be indirect. In view of the
   role of SDS22 in the maturation of newly translated PP1 (see
   Introduction), we examined whether SDS22 depletion affects the PP1
   protein level. The Dox/IAA-induced depletion of SDS22 indeed resulted
   in a gradual loss of PP1, as detected by immunoblotting with an
   antibody that recognizes all PP1 isoforms (Fig. [132]2h, [133]i).
   Eventually, 96 h after the induction of SDS22 degradation, the PP1
   level was decreased by about 80%. We verified that the loss of PP1 in
   SDS22-depleted cells was not caused by decreased transcription, as the
   transcript levels of the PP1α, β and γ isoforms were not affected by
   SDS22 depletion (Supplementary Fig. [134]2k). Collectively, our data
   strongly suggest that the severe phenotype associated with the
   depletion of SDS22 stems from a loss of PP1 protein, resulting in
   hyperphosphorylation of PP1 substrates and misregulation of cellular
   processes that are critically dependent on PP1, including SAC silencing
   and cell re-attachment after cell division.

Loss of SDS22 and PP1 in a patient with severe neurodevelopmental delay

   A 10-year-old female patient, further referred to as P1, exhibited a
   psychomotor developmental delay from the early months of life.
   Additionally, she experienced febrile seizures from 14 months to 3
   years of age. Throughout her life, she never achieved the ability to
   sit, walk independently or speak. The child presents with severe
   intellectual disability, pronounced hypotonia with present
   osteotendinous reflexes, hyperkinesia, choreodystonic movements, ptosis
   and swallowing difficulties. Comprehensive metabolic studies of blood,
   urine and cerebrospinal fluid did not yield aberrant results
   (Supplementary Note [135]1; Supplementary Tables [136]1 and [137]2). At
   the age of 8, brain magnetic resonance imaging (MRI) revealed a
   reduction in the volume of white matter, nerves, the optic chiasm,
   thalami and the medulla oblongata, with no signal abnormalities.
   Electromyography and nerve conduction velocities appeared within the
   normal range (Supplementary Fig. [138]3a). However, a muscle biopsy
   performed at the age of 4 showed minimal, non-specific changes
   (Supplementary Note [139]1).

   Whole exome sequencing (WES) trio analyses revealed two homozygous
   variants (in SACS and NID1 genes), with only the mother being
   heterozygous (Supplementary Data [140]3). MLPA and microsatellite
   analyses were consistent with maternal isodisomy of chromosome 13. This
   was further confirmed by whole-genome SNP array (750 K Cytoscan,
   Applied Biosystems), showing a female hybridization pattern with no
   evidence of clinically significant copy number alterations according to
   current knowledge, and revealing loss of heterozygosity (LOH) in two
   segments of chromosome 13 spanning 61.6 Mb (arr[GRCh37]
   13q11q12.3(19450957_31639909) x2 hmz,
   13q14.13q32.1(46699801_96106269)x2 hmz). Detailed analysis of the WES
   data and subsequent mRNASeq analysis did not disclose any putative
   pathogenic variants within this region. Therefore, the partial
   heteroisodisomy is unlikely to be the cause of the severe phenotype of
   P1.

   Importantly, further analysis of the WES data also uncovered a de novo
   heterozygous mutation (2:242109282G > A; c.906G > A:p.Trp302*) in the
   SDS22-encoding PPP1R7 gene, absent in gnomAD v2 (Supplementary
   Fig. [141]3b). This mutation was only observed in 24% of the
   transcripts (out of a total of 124 reads), which is less than expected
   for a heterozygous mutation (50%), suggesting that the mutated SDS22
   transcript is about twofold less stable than the wildtype transcript.
   Moreover, the mutation in patient P1 introduces a stop codon in exon 10
   of PPP1R7^[142]20, putatively resulting in the expression of a
   C-terminally nicked SDS22 variant, lacking the last 58 residues
   comprising LRR11, LRR12 and the LRR-Cap structure (Fig. [143]3a). To
   verify the consequences of the heterozygous PPP1R7 c.906G > A mutation
   on the expression levels of SDS22 and PP1, we isolated fibroblasts from
   a skin biopsy of P1 and four healthy controls, and found that the level
   of SDS22 in the cells of patient P1, as compared to that in controls,
   was decreased by about 80% (Fig. [144]3b). This decrease was larger
   than expected from a heterozygous mutation, suggesting that
   SDS22-(1-301) in P1 may function as a dominant-negative mutant.
   Importantly, the level of PP1 in the patient’s fibroblasts was also
   decreased by about 65%. Collectively, these data indicated that the
   mutated SDS22 transcript and protein in patient P1 are unstable and
   that the resulting decreased expression levels of SDS22 are associated
   with a co-depletion of PP1. Multiple efforts to identify additional
   patients on matchmaking platforms with Trp302* mutations in SDS22 were
   unsuccessful thus far.

Fig. 3. Loss of SDS22 and PP1 in a human patient with neurodevelopmental
disease.

   [145]Fig. 3
   [146]Open in a new tab

   a Domain structure of SDS22 (upper panel) and mapping of the
   heterozygous SDS22-W302* mutation in patient P1 (lower panel),
   predicted to result in the expression of an SDS22 variant that lacks
   LRR11, LRR12 and the C-terminal LRR-Cap (C-Cap). The blue-colored
   fragment is missing in SDS22-W302*. b Levels of PP1, SDS22 and
   α-tubulin in fibroblasts from patient P1 and four healthy controls, as
   detected by immunoblotting using antibodies that bind to the N-terminus
   of SDS22 or the C-terminus of all PP1 isoforms (left panel). The right
   panel shows the quantification of the immunoblotting data. The results
   are shown as means ± SD (n = 3 independent experiments). The control
   value in each experiment was the average for four controls and the
   values for one control were set at 100%. P values are from two-sided
   unpaired t-test. c Co-immunoprecipitation (EGFP-traps) of endogenous
   BCLAF1, PP1 and I3 with transiently expressed EGFP, EGFP-SDS22-WT (WT)
   or EGFP-SDS22-W302* (W302*) in HEK293T cells. d Immunofluorescence
   staining of HEK293T cells transiently transfected with expression
   vectors for EGFP-SDS22 or EGFP-SDS22-W302^*. The cells were treated
   with or without 10 μM MG132 for 8 h before fixation. The fixed cells
   were stained for DNA (DAPI), EGFP and α-tubulin. The scale bars are 20
   μm. e The effect of MG132 on the levels of EGFP-SDS22 and
   EGFP-SDS22-W302*, as measured in 1 h intervals by IncuCyte live-cell
   analysis for 24 h. HEK293T cells were transiently transfected with
   EGFP-SDS22-WT or EGFP-SDS22-W302^*. 10 μM MG132 was added 1 h before
   the first measurement. Green fluorescence (Green Calibrated Unit, GCU)
   was measured to show EGFP intensity. f EGFP-SDS22-W302^* accumulates in
   granules. Detail of cells shown in (d).

   We have subsequently compared the fate of EGFP-tagged SDS22-WT and
   SDS22-W302* in transiently transfected HEK293T cells. Traps of
   EGFP-SDS22-WT contained PP1, I3 and BCLAF1, but these ligands were not
   detected in traps of EGFP-SDS22-W302* (Fig. [147]3c), showing that the
   SDS22 mutant of patient P1 is dysfunctional. Moreover, SDS22-W302* was
   expressed at much lower levels than EGFP-SDS22-WT (Fig. [148]3c–e), but
   this phenotype was largely rescued by the addition of the proteasome
   inhibitor MG132 (Fig. [149]3d,[150]e). This suggested that SDS22-W302*
   was less abundant because of an increased degradation rate. Finally, we
   noted that SDS22-W302*, before and after treatment with MG132, largely
   appeared in granules, hinting at an increased tendency to aggregate
   (Fig. [151]3f). Collectively, these data validated and extended our
   observations on SDS22-W302* in fibroblasts of P1, and demonstrated that
   this mutant is deficient in ligand binding, less soluble and readily
   targeted for proteasomal degradation, consistent with previous data on
   the essential role of the C-terminus of SDS22^[152]21,[153]22.

SDS22 also interacts directly with I3

   While the previous data disclosed a key role for SDS22 in stabilizing
   newly translated PP1, they did not rule out additional functions for
   SDS22 in the SDS22:PP1:I3 complex, for example, related to I3
   recruitment or the later steps of PP1-holoenzyme assembly. Modeling of
   the ternary complex using the AlphaFold-Multimer tool correctly
   predicted with high confidence the experimentally determined
   interaction sites of SDS22 and I3 for PP1 (Fig. [154]4a, [155]b and
   Supplementary Fig. [156]4a–d)^[157]23. Thus, SDS22 mainly bound to PP1
   via its LRR repeats, consistent with crystallographic
   data^[158]9,[159]18. I3 has a degenerate RVxF motif that properly
   docked to its well-characterized hydrophobic binding groove on
   PP1^[160]6,[161]24. The model also correctly predicted that the CCC
   motif of I3 occludes the active site of PP1, as recently
   demonstrated^[162]6. Unexpectedly, the AlphaFold-Multimer model of
   SDS22:PP1:I3 also predicted a hitherto unknown direct interaction
   between SDS22 and I3. Co-precipitation experiments using purified
   His-I3 and SDS22 confirmed their direct interaction, independent of PP1
   (Fig. [163]4c). To further validate the SDS22:I3 interaction in a
   cellular context, we transiently transfected HEK293T cells with
   EGFP-tagged I3-fragments and examined the interaction of the trapped
   fusions with PP1 and SDS22 (Fig. [164]4d, [165]e). This analysis
   disclosed a key role for I3 residues 71–84 in the interaction with
   SDS22 which, however, were not required for PP1 binding. I3-(71–84)
   comprises four acidic residues that were predicted to interact with
   seven basic residues at the concave side of LRR3-8 of SDS22
   (Fig. [166]4b, [167]f, [168]g). All of these charged residues of I3 and
   SDS22 are phylogenetically conserved (Fig. [169]4g and Supplementary
   Fig. [170]4e, [171]f). Moreover, alanine mutation of the four
   implicated acidic residues in EGFP-I3 (I3-DE4A) abolished its
   interaction with SDS22 in EGFP-pulldown experiments, but only
   moderately decreased (35 ± 10%; means ± SEM, n = 6) its binding to PP1
   (Fig. [172]4h). Conversely, alanine mutation of EGFP-SDS22 at the seven
   implicated basic residues (SDS22-KR7A) abolished its binding to I3 in
   EGFP-trapping assays and, as further discussed below, even considerably
   increased its interaction with PP1 (Fig. [173]4i). While I3-DE4A and
   SDS22-KR7A showed a reduced interaction with SDS22-WT and I3-WT,
   respectively (Fig. [174]4h, i), their interaction was re-established by
   reciprocal charge-reversal mutations of both I3 (I3-DE4R) and SDS22
   (SDS22-KR7E) (Fig. [175]4j). Finally, we performed ITC binding assays
   with purified proteins, which confirmed a direct interaction between
   SDS22 and His-I3 (K[d] = 7.6 ± 5.5 μM). However, their binding affinity
   was much reduced (K[d] = 127.6 ± 28.1 μM) by mutation of the
   I3-interaction site of SDS22 (Fig. [176]4k and Supplementary
   Fig. [177]4g). Together, these data firmly established a direct, ionic
   interaction site between SDS22 and I3.

Fig. 4. Mapping of an SDS22:I3 interaction interface.

   [178]Fig. 4
   [179]Open in a new tab

   a Domain structure of SDS22 and I3. Shown are the LRR and C-Cap
   structures of SDS22, the PP1-binding RVxF- and CCC-motifs of I3 as well
   as the predicted SDS22-binding region (SBR) of I3. b The rank-1
   AlphaFold-Multimer model (out of 25 predicted models) of SDS22:PP1:I3,
   using the amber-relaxation setup option. Residues with pLDDT scores <30
   are not displayed. SDS22 (Hs), blue; PP1α (Hs), gray; I3 (Hs), pink.
   The PP1-binding motifs of I3 are highlighted with dashed boxes and are
   also shown as zoom-in views. c Immunoblotting of trapped His-I3 for the
   presence of PP1α and SDS22. His-I3 (2 μM) or His were incubated with
   buffer, PP1 (0.3 μM), SDS22 (0.3 μM) or PP1 + SDS22, and trapped by
   Ni-NTA magnetic beads. d Co-immunoprecipitation (EGFP-traps) of
   endogenous PP1 and SDS22 with ectopically expressed EGFP or EGFP-I3
   variants, i.e., EGFP-tagged I3-(1–126; WT), I3-(1–84), I3-(1–70) or
   I3-(85–126). NT no transfection. e Schematic representation of the used
   EGFP-I3 truncation mutants and their interaction with PP1 and SDS22, as
   derived from data in (d). f Electrostatic surface of the SDS22:PP1:I3
   AlphaFold model. The zoom-in view shows the charged residues of SDS22
   and I3 that are predicted to interact. g Sequence alignment of the
   interacting fragments of SDS22 and I3, with the same color code for the
   involved charged residues. The alignment was generated using the
   EMBL-EBI MAFFT tool with the following UniProt identifiers: Homo
   sapiens, [180]Q15435; Drosophila melanogaster, [181]Q9VEK8;
   Caenorhabditis elegans, [182]P45969; Saccharomyces cerevisiae,
   [183]P36047; Schizosaccharomyces pombe, [184]P22194; Arabidopsis
   thaliana, [185]Q84WJ9. h Association of endogenous PP1 and SDS22 with
   transiently expressed, EGFP-trapped EGFP-I3-WT or EGFP-I3-DE4A. i
   Binding of endogenous PP1 and I3 to transiently expressed, EGFP-trapped
   EGFP-SDS2-WT or EGFP-SDS22-KR7A. j Co-immunoprecipitation (Strep-traps)
   of transiently expressed variants of Strep-I3 (Strep-I3-WT;
   Strep-I3-DE4R) and EGFP-SDS22 (EGFP-SDS22-WT; EGFP-SDS22-KR7E). The
   lower part of the EGFP blot (middle panel) was used for Strep-I3
   detection (upper panel). k Isothermal titration calorimetry (ITC)
   experiments of purified SDS22-WT or SDS22-KR7A and His-I3. 5 μM
   SDS22-WT or SDS22-KR7A was titrated with 50 μM His-I3. The inset shows
   average K[d] values ± SEM (n = 3).

   The seven basic residues of SDS22 that interact with I3 have also been
   implicated in the binding of the splicing factor BCLAF1^[186]18. We
   confirmed that ectopically expressed SDS22-WT, but not SDS22-KR7A,
   binds to both I3 and BCLAF1 (Supplementary Fig. [187]4h). However, an
   ectopically expressed PP1–SDS22 fusion was associated with I3, but not
   with BCLAF1, in pulldown experiments. This indicates that there is no
   competition between I3 and BCLAF1 for binding to PP1–SDS22, probably
   because I3 is additionally anchored through PP1 binding. Hence, our
   data suggest that BCLAF1 only interacts with the cellular pool of SDS22
   that is not associated with I3 and/or PP1.

The SDS22:I3 interaction makes the recruitment of I3 to nascent PP1
irreversible

   Next, we investigated the importance of the SDS22:I3 interaction for
   the recruitment of I3 to PP1. An SDS22-binding mutant of ectopically
   expressed mClover-tagged PP1γ (K147A, K150A)^[188]18, referred to as
   PP1-2KA, was deficient in the binding of both SDS22 and I3 in
   mClover-trapping experiments (Fig. [189]5a), indicating that the
   efficient recruitment of I3 depends on SDS22. We also made use of a
   HeLa cell line that inducibly expressed Strep-tagged PP1γ^[190]8 and
   found that I3 was no longer recruited to newly translated Strep-PP1γ
   (induction for 1 h) after the knockdown of SDS22 (Fig. [191]5b).
   Conversely, the recruitment of SDS22 to Strep-PP1γ was not affected by
   the knockdown of I3 (Fig. [192]5c). Finally, we used purified
   components to study how the binding of His-I3 to GST-PP1α is affected
   by SDS22 addition. The data confirmed that the PP1:I3 interaction was
   enhanced by SDS22 addition (Supplementary Fig. [193]5a) and that SDS22
   addition rendered the interaction of I3 with GST-PP1, but not the
   SDS22-binding mutant GST-PP1-2KA, resistant to competitive disruption
   with NIPP1-(143–224) (Supplementary Fig. [194]5b). Together, these data
   demonstrated that SDS22 is recruited to nascent PP1 before I3 and that
   the efficient recruitment of I3, in the presence of RIPPOs that compete
   for binding to PP1, depends on its interaction with SDS22.

Fig. 5. The recruitment of I3 to nascent PP1 depends on SDS22.

   [195]Fig. 5
   [196]Open in a new tab

   a Association of mClover-trapped mClover, mClover-PP1-WT or
   mClover-PP1-2KA with endogenous SDS22 and I3. The expression of
   mClover-PP1 fusions in HeLa FlpIn T-REx cell lines was induced for 24 h
   with Dox. b Reduced recruitment of I3 to nascent Strep-PP1
   (Strep-trapping) following the knockdown of SDS22. Strep-PP1γ
   expression in a HeLa FlpIn T-REx cell line was induced for 1 h with
   Dox. c The recruitment of SDS22 to nascent Strep-PP1 (Strep-trapping)
   was not affected by the knockdown of I3. The cells were treated as
   detailed for (b). d Overview of the lysated-based split-luciferase
   protocol. The interacting proteins (a and b) are shown in blue and the
   fused luciferase fragments are shown in pink. e Scheme of the used
   fusions with LgBiT or SmBiT. f Transient expression of SmBiT/LgBiT
   fusions in HEK293T cells. g Luciferase complementation following mixing
   of HEK293T lysates containing LgBiT-PP1 and a SmBiT-tagged RIPPO (WT or
   RATA mutant). The signals were normalized for the same expression
   level. The results were plotted as a percentage of the signal with
   LgBiT-PP1α + SmBIT-RIPPO^WT (means ± SD; n = 3 independent
   experiments). P values were from two-sided unpaired t-test. h
   Comparison of luciferase complementation with the indicated
   SmBiT-RIPPOs (WT) and either LgBiT-PP1 or LgBiT-PP1–SDS22. The results
   were plotted as a percentage of the signal with
   LgBiT-PP1 + SmBIT-RIPPO. i Kinetic-trace experiments showing the
   time-dependent association of LgBiT-PP1 with I3-SmBiT or RM-SmBiT after
   addition of the RIPPO-SmBiT fusions (SmBiT). Also shown is the
   competitive disruption of the assembled complexes with 5 µM
   NIPP1-(143–224). The data are plotted as a percentage of the signal
   just before addition of competitor at 12 min. j Same experiment as in
   (i) but with LgBiT-PP1-SDS22. k Same experiment as in (j) but with
   I3-DE4A-SmBiT. l Kinetic-trace experiment showing the time-dependent
   association of LgBiT-PP1α and SDS22-SmBiT. Purified SDS22-WT and
   SDS22-E192A (10 μM) were added as competitors. The data are plotted as
   a percentage of the signal just before addition of competitor
   (t = 22 min). The experiments shown in (i–l) are representative for
   three independent experiments.

   To further define the role of SDS22 in the recruitment of I3 to PP1, we
   made use of lysate-based split-luciferase assays^[197]13. These assays
   are based on the complementation between small (SmBiT) and large
   (LgBiT) fragments of Nanoluc luciferase (Promega). The
   SmBiT/LgBiT-fragments are catalytically inactive, but complement into
   active luciferase when brought in close proximity through interaction
   of their fused partners (Fig. [198]5d). To investigate how the I3:PP1
   interaction is affected by SDS22, we generated lysates from HEK293T
   cells transiently transfected with constructs encoding either wildtype
   I3-SmBiT (I3^WT-SmBiT), I3-SmBiT with a mutated PP1-binding RVxF motif
   (I3^RATA-SmBiT), LgBiT-PP1α or LgBiT linked to a fusion of PP1α and
   SDS22 (LgBiT-PP1α-SDS22) (Fig. [199]5e, f). Luciferase complementation
   was obtained upon mixing of lysates with I3^WT-SmBiT and LgBiT-PP1α,
   but not with a mixture of lysates containing I3^RATA-SmBiT and
   LgBiT-PP1α (Fig. [200]5g), demonstrating that luciferase
   complementation depended on the binding of I3 to PP1. Interestingly,
   complementation with I3-SmBiT^WT was nearly threefold higher with
   lysates expressing LgBiT-PP1α-SDS22 rather than LgBiT-PP1α
   (Fig. [201]5h), at similar expression levels (Fig. [202]5f). These data
   suggested that I3 binds with a higher affinity to a PP1–SDS22 fusion
   than to PP1, confirming a key contribution of the I3:SDS22 interaction.
   Importantly, the increased complementation with the SDS22-PP1 fusion
   was specific for I3, as it was much less pronounced with RepoMan (RM),
   a structurally unrelated RIPPO with a PP1-binding RVxF motif
   (Fig. [203]5e–h). The latter data are in accordance with the ability of
   RM to form a ternary complex with SDS22:PP1 without, however, making
   direct contacts with SDS22^[204]18. The interaction between I2-SmBiT
   and LgBiT-PP1α was even reduced when PP1α was fused with SDS22
   (Fig. [205]5e–h), consistent with the overlapping binding sites of
   SDS22 and I2 on PP1^[206]9,[207]18.

   To delineate the effect of SDS22 on the dynamics of the PP1:I3 and
   PP1:RepoMan interactions, we performed kinetic experiments using the
   split-luciferase system. Following the time-dependent assembly of
   PP1:I3 or PP1:RepoMan upon mixing of lysates containing the
   appropriately tagged subunits, both complexes were competitively
   disrupted by the addition of an excess of a PP1-binding RVxF-peptide
   (Fig. [208]5i and Supplementary Fig. [209]5c, [210]d). However, when I3
   and RepoMan complexes were made with a PP1–SDS22 fusion, only the
   RepoMan complex could still be disrupted by addition of an RVxF-peptide
   (Fig. [211]5j and Supplementary Fig. [212]5c), indicating that the
   binding of I3 had become irreversible by its simultaneous binding to
   PP1 and SDS22. In accordance with this interpretation, an SDS22-binding
   mutant of I3 (I3-DE4A) could be dissociated from the PP1–SDS22 fusion
   (Fig. [213]5k and Supplementary Fig. [214]5c) with an RVxF-peptide.
   Importantly, the interaction between LgBiT-PP1 and SDS22-SmBiT was
   barely affected by the addition of a large excess of untagged SDS22
   (Fig. [215]5l), but untagged SDS22 did prevent the time-dependent
   formation of new complexes between the tagged fusions. In contrast, the
   addition of a PP1-binding mutant of SDS22 (SDS22-E192) did not prevent
   the assembly of new PP1:SDS22 complexes. These data demonstrated that
   SDS22 also binds irreversibly to PP1, consistent with affinity
   measurements^[216]9. Hence, both SDS22 and I3 bind irreversibly to PP1,
   explaining why their extraction from SDS22:PP1:I3 requires energy, as
   provided by p97/VCP-mediated ATP hydrolysis.

The SDS22:I3 interaction is essential for PP1-holoenzyme assembly

   Finally, we examined the importance of the direct interaction between
   SDS22 and I3 in the SDS22:PP1:I3 complex for the subsequent steps of
   PP1-holoenzyme assembly. The binding of EGFP-I3 to p97/VCP in
   EGFP-trapping experiments was lost after mutation of either its
   PP1-binding site (I3-KAEA) or SDS22-binding site (I3-DE4A)
   (Fig. [217]6a). This indicated that p97/VCP is recruited by
   SDS22:PP1:I3, but not by PP1:I3 or SDS22:I3, consistent with recent
   data showing that p97/VCP binds to both SDS22 and I3^[218]12. We also
   found that traps of an EGFP-PP1-SDS22 fusion contained I3, but
   relatively little of other SLiM-based RIPPOs, such as RepoMan or MYPT1
   (Fig. [219]6b). This suggested that the irreversible recruitment of I3,
   resulting from its simultaneous binding to SDS22 and PP1, precludes the
   competitive binding of other RIPPOs. In accordance with this
   interpretation, an EGFP-PP1-SDS22 fusion with the SDS22 moiety mutated
   in its I3-binding site (EGFP-PP1-SDS22-KR7A) did co-precipitate
   canonical RIPPOs such as MYPT1 and RepoMan. We furthermore found that
   SDS22-WT, ectopically expressed in HeLa cells, only interacted weakly
   and transiently with newly translated Strep-PP1γ (Fig. [220]6c),
   consistent with the gradual p97/VCP-mediated transfer of Strep-PP1γ to
   canonical RIPPOs^[221]8. In contrast, SDS22-KR7A interacted more
   strongly and for a prolonged time with nascent Strep-PP1
   (Fig. [222]6c), accounting for its much increased complexation with the
   global pool of PP1 (Fig. [223]4i).

Fig. 6. The SDS22:I3 interaction is essential for the assembly of functional
PP1 holoenzymes.

   [224]Fig. 6
   [225]Open in a new tab

   a Association of endogenous p97/VCP, SDS22 and PP1 with transiently
   expressed and trapped EGFP or EGFP-trapped I3 variants from HEK293T
   cell lysates. NT non-transfected. b Co-immunoprecipitation of
   endogenous I3, MYPT1 and RepoMan (RM) with transiently expressed and
   trapped EGFP-tagged PP1-SDS22-WT/KR7A or EGFP-β-galactosidase (negative
   control) from HEK293T cell lysates. c Co-immunoprecipitation of
   transiently expressed EGFP-tagged SDS22-WT or SDS22-KR7A with
   Strep-PP1γ, induced in HeLa FlpIn T-REx cells with Dox for 0→24 h. d
   Expression of EGFP-tagged SDS22-WT and SDS22-KR7A in HeLa FlpIn T-Rex
   cell lines after Dox addition for 48 h. Also shown is endogenous SDS22
   and PP1. e Cell proliferation of HeLa FlpIn T-Rex cells, treated or not
   with Dox to induce the expression of EGFP-tagged SDS22-WT or
   SDS22-KR7A. Cell proliferation was measured in 3 h intervals using
   IncuCyte. Cell confluency is shown as a percentage of the surface. The
   solid line represents the means of three technical replicates and the
   shaded area represents the SD range from a single data set, which is
   representative for three independent experiments. f The effect of
   EGFP-tagged SDS22-WT or SDS22-KR7A expression (induction for 48 h) on
   cell-cycle progression. The cells were first arrested at the G[1]/S
   transition. After thymidine release, the cells were blocked at the
   G[2]/M transition with RO3306. After RO3306 washout for the indicated
   times, the cells were fixed, stained with propidium iodide and analyzed
   by flow cytometry. g Images (left panel) and quantification (right
   panel) of M-phase duration, as derived from live-cell imaging of cells
   induced (48 h) to express EGFP-tagged SDS22-WT or SDS22-KR7A. The cells
   were analyzed after a double-thymidine arrest and a release for 6 h. P
   value is from two-sided unpaired t-test (n = 100 cells). The scale bars
   are 50 μm. h Confocal images of cells induced to express EGFP-tagged
   SDS22-WT or SDS22-KR7A for 48 h. Cells were treated with SDS22 siRNA.
   The cells were fixed following a release (60 min) from an RO3306 block.
   Cells in metaphase and anaphase were stained for DAPI, EGFP, α-tubulin,
   histone H3 phosphorylated at Thr3 (pH3T3) and pERM. Scale bars, 20 μm.

   The above data indicated that the SDS22:I3 interaction is required for
   their co-extraction by p97/VCP during PP1-holoenzyme assembly. Since
   SDS22-associated PP1 is inactive^[226]9, the erroneous transfer of
   PP1:SDS22-KR7A to canonical RIPPOs is expected to generate
   non-functional holoenzymes. To test this hypothesis, we compared the
   phenotype of stable HeLa Flp-In cell lines that inducibly express
   either SDS22-WT or SDS22-KR7A (Fig. [227]6d). While the overexpression
   of SDS22-WT had no effect on proliferation, the overexpression of
   SDS22-KR7A resulted in a reduced proliferation (Fig. [228]6e). This
   proliferation-deficit was not caused by a slower progression from
   G[1]/S to G[2]/M, as shown by flow-cytometry analysis of cells that
   were released from a G[1]/S arrest (Supplementary Fig. [229]6a).
   However, cells expressing SDS22-KR7A needed more time for the
   completion of mitosis, as illustrated by flow-cytometry analysis of
   cells released from a G[2]/M arrest with RO3306 (Fig. [230]6f and
   Supplementary Fig. [231]6b) as well as by time-lapse video imaging
   (Fig. [232]6g). The mitotic-arrest phenotype of SDS22-KR7A expression
   correlated with an increased phosphorylation of established mitotic PP1
   substrates, including ERM proteins and histone H3 at Thr3
   (Fig. [233]6h). Together, these data demonstrate that a failure to
   co-extract SDS22 and I3 from SDS22:PP1:I3 results in the formation of
   PP1 holoenzymes that contain SDS22 as a third, inhibitory subunit.

Discussion

   We have generated cellular and molecular research tools to address
   remaining key questions on the biogenesis of PP1 holoenzymes (see
   “Introduction”), with a particular focus on the role of SDS22. The
   available data allow us to propose a more detailed model on how nascent
   PP1 is transferred, in a stepwise manner, to canonical RIPPOs, thereby
   forming functional holoenzymes (Fig. [234]7). Newly translated PP1
   first recruits SDS22 (Fig. [235]5b,[236]c). The irreversible nature of
   SDS22 recruitment probably stems from the large number of contacts
   between PP1 and the highly structured PP1-binding domain of
   SDS22^[237]9,[238]18. Moreover, SDS22 locks PP1 in inactive
   conformation that precludes metal-1 loading. SDS22 recruitment serves
   to stabilize nascent PP1 (Fig. [239]2h, [240]i), which is prone to
   aggregation in its unbound state^[241]9,[242]25. In the absence of
   SDS22, newly translated PP1 is gradually lost (Figs. [243]2h, [244]i
   and [245]7), resulting in the hyperphosphorylation of PP1 substrates
   and a nearly complete proliferation arrest (Figs. [246]1c and [247]2c,
   [248]e, [249]f). The next step in the biogenesis of PP1 involves the
   recruitment of I3 to SDS22:PP1 (Figs. [250]5b, [251]c and [252]7),
   which prevents substrate binding due to occlusion of the active site by
   the CCC motif of I3^[253]6. The binding of I3 is also irreversible
   because of its simultaneous interaction with SDS22 and PP1
   (Fig. [254]5i–k). This irreversibility explains the selectivity of I3
   recruitment, as it precludes competition with other SLiM-based RIPPOs
   for binding to SDS22:PP1 (Fig. [255]6b), as well as the
   energy-dependency of its extraction during holoenzyme assembly. A first
   glance, our data on the stabilization of PP1:I3 by SDS22 (Fig. [256]5i,
   j and Supplementary Fig. [257]5c) are at variance with a recent study
   indicating that SDS22 weakens the PP1:I3 interaction^[258]26. However,
   the latter study was performed with N-terminally truncated PP1 and
   SDS22, which may affect their interaction with I3.

Fig. 7. Model of the stepwise biogenesis of PP1 holoenzymes.

   [259]Fig. 7
   [260]Open in a new tab

   SDS22 binds irreversibly to newly translated PP1 to prevent its
   aggregation. SDS22:PP1 recruits I3 in an irreversible manner, due to
   the simultaneous interaction of I3 with PP1 and SDS22. Subsequently,
   SDS22 and I3 are co-extracted from the ternary SDS22:PP1:I3 complex by
   p97/VCP, enabling the association of released PP1 with canonical RIPPOs
   to form functional holoenzymes. In the absence of I3 or after mutation
   of the SDS22:I3 interaction site, SDS22 and associated PP1 are
   co-transferred to canonical RIPPOs, resulting in the assembly of
   inactive SDS:PP1:RIPPO complexes.

   It is only when the ternary SDS22:PP1:I3 complex is formed that p97/VCP
   is efficiently recruited (Figs. [261]6a and [262]7), consistent with
   the recent report that p97/VCP interacts with both SDS22 and
   I3^[263]12. The p97/VCP-catalyzed unfolding of I3 results in the
   co-extraction of SDS22^[264]8, which is critically dependent on its
   interaction with I3 (Fig. [265]6b). SDS22 that is mutated in its
   I3-interaction site erroneously co-transfers with PP1 to canonical
   RIPPOs, resulting in the formation of inactive PP1 holoenzymes that
   contain SDS22 as an inhibitory third subunit (Figs. [266]6b, [267]h and
   [268]7). SDS22 may be co-extracted with I3 because of the pulling
   forces exerted by I3 unfolding in the p97/VCP-hexamer channel^[269]8.
   In addition, the co-extraction of SDS22 may be facilitated by the
   (hypothetical) coupled incorporation of metal 1 (Zn^2+) in the active
   site, which is known to reduce the binding affinity of SDS22 for
   PP1^[270]9. The incorporation of Zn^2+ possibly involves I3, as
   suggested by its recent identification as a Zn^2+-binding
   protein^[271]6. However, the mechanistic details of metal loading in
   the active site of PP1 are still unresolved. A final step in the
   biogenesis of PP1 holoenzymes, associated with the co-extraction of I3
   and SDS22 from SDS22:PP1:I3, is the (spontaneous) association of
   liberated PP1 with canonical RIPPOs (Fig. [272]7). As RIPPOs are
   expressed in a large molar excess over PP1^[273]2, all p97/VCP-freed
   PP1 will be rapidly titrated, which precludes uncontrolled
   dephosphorylation by the free catalytic subunit^[274]27. The diversity
   of PP1:RIPPO complexes that accumulate in the cell is largely
   determined by the relative concentration of RIPPOs and their affinity
   for PP1, which are both tightly regulated. For example, the affinity of
   many RIPPOs for PP1 is reduced during the first half of mitosis by
   phosphorylation of residues within or close to the RVxF motif^[275]28.
   As SLiM-based PP1:RIPPO interactions are highly dynamic, changes in the
   expression of RIPPOs or in their affinity for PP1 result in a
   spontaneous re-equilibration between PP1 holoenzymes.

   The proposed model of the biogenesis of PP1 holoenzymes (Fig. [276]7)
   reconciles a large body of data in the literature. Thus, the model
   explains why SDS22 is an inhibitor of PP1 in biochemical assays but an
   activator of PP1 in a cellular context (for references see ref.