The Ki-67 and RepoMan mitotic phosphatases assemble via an identical, yet novel mechanism

Ki-67 and RepoMan have key roles during mitotic exit. Previously, we showed that Ki-67 organizes the mitotic chromosome periphery and recruits protein phosphatase 1 (PP1) to chromatin at anaphase onset, in a similar manner as RepoMan (Booth et al., 2014). Here we show how Ki-67 and RepoMan form mitotic exit phosphatases by recruiting PP1, how they distinguish between distinct PP1 isoforms and how the assembly of these two holoenzymes are dynamically regulated by Aurora B kinase during mitosis. Unexpectedly, our data also reveal that Ki-67 and RepoMan bind PP1 using an identical, yet novel mechanism, interacting with a PP1 pocket that is engaged only by these two PP1 regulators. These findings not only show how two distinct mitotic exit phosphatases are recruited to their substrates, but also provide immediate opportunities for the design of novel cancer therapeutics that selectively target the Ki-67:PP1 and RepoMan:PP1 holoenzymes. DOI: http://dx.doi.org/10.7554/eLife.16539.001


Introduction
Mitotic exit comprises a complex series of events that includes sister chromatid segregation, mitotic spindle disassembly, nuclear-envelope re-assembly and chromosome decondensation (Wurzenberger and Gerlich, 2011). How these events are coordinated during mitotic exit is still an open question. The emerging picture is that exiting mitosis requires the specific engagement and activation of protein phosphatases, including ser/thr phosphatase protein phosphatase 1 (PP1) (Funabiki and Wynne, 2013;Rosenberg et al., 2011). While PP1 exhibits broad specificity, it acts in a highly specific manner by forming stable complexes, known as holoenzymes, with a host of regulatory proteins that direct PP1 activity towards specific substrates and localize PP1 to specific regions of the cell (Hendrickx et al., 2009;Bollen et al., 2010;Peti et al., 2013;Peti and Page, 2015;Choy et al., 2014;O'Connell et al., 2012). A detailed understanding of how key PP1 regulators bind and direct PP1 activity during distinct stages of the cell cycle is still largely missing.
Ki-67 is widely used as a prognostic marker for many cancers (Dowsett et al., 2011;Lin et al., 2016;Sobecki et al., 2016), yet for decades, its molecular function remained largely unknown. Recently, it was shown that one role of Ki-67 is to function as a 'DNA surfactant', preventing individual chromosomes from collapsing into a single chromatin mass upon nuclear envelope disassembly by binding directly to the surface of chromatin (Cuylen et al., 2016). A second recently discovered function is that it binds and regulates the activity of PP1 during mitosis using the canonical PP1 RVxF small linear motif (SLiM) (Booth et al., 2014). The only other protein that exhibits any sequence similarity with Ki-67 near its RVxF motif is RepoMan (recruits PP1g onto mitotic chromatin at anaphase, also known as cell division cycle associated 2, CDCA2), a nuclear-specific protein that was discovered for its ability to specifically recruit PP1g to chromosomes at anaphase onset and, like Ki-67, is upregulated in many cancers Vagnarelli, 2014). During mitotic exit, PP1, in particular PP1g (PP1 isoforms include PP1a, PP1b, PP1g and PP1g2; ! 85% identity between isoforms), is essential for histone dephosphorylation, nuclear-envelope reassembly and chromatin remodeling (Qian et al., 2011;Vagnarelli et al., 2006;Wurzenberger et al., 2012;Qian et al., 2013;Vagnarelli et al., 2011). These PP1g-specific processes are mediated largely by the Ki-67: PP1g and RepoMan:PP1g holoenzymes. Two key unresolved questions are how do these regulators assemble mitotic phosphatases and how are these interactions dynamically regulated in both space and time? Answers to these questions will provide novel opportunities for the development of Ki-67: PP1 and RepoMan:PP1 specific therapeutics for cancer. In this report, we demonstrate how Ki-67 and RepoMan form mitotic exit phosphatases, how they distinguish between distinct PP1 isoforms and how the assembly of these two holoenzymes is dynamically regulated by Aurora B kinase during mitosis.

Results
The Ki-67 and RepoMan PP1 interaction domain  and RepoMan (1023 aa) bind PP1g to form isoform-specific holoenzymes (Booth et al., 2014;Trinkle-Mulcahy et al., 2006). Previously, we showed that these proteins exhibit sequence similarity in only a very short region, of about 40 residues ( Figure 1A) (Booth et al., 2014). This region includes the canonical RVxF SLiM that is critical for PP1 binding ( 505 RVSF 508 , Ki-67; 392 RVTF 395 , RepoMan; more than 70% of PP1 regulators contain the RVxF SLiM) Vagnarelli et al., 2011). Because recent efforts to understand how PP1 activity is directed by its >200 distinct regulators has revealed that residues outside the RVxF motif are also essential for PP1 holoenzyme formation and function O'Connell et al., 2012;Terrak et al., 2004), we reasoned that this entire region was critical for PP1 binding. Using isothermal titration calorimetry (ITC), we showed that Ki-67 496-536 and RepoMan 383-423 bind PP1g 7-323 with a 1:1 stoichiometry and with nearly equivalent affinities ( Figure 1B Table 1). We also showed that extending this domain does not enhance binding (Table 1). Finally, NMR spectroscopy experiments with RepoMan confirmed that all residues in this conserved region interact with PP1 (Figure 1-figure supplements 2,3). Together, these data suggest that Ki-67 496-536 and RepoMan 383-423 bind PP1 using identical mechanisms and that this conserved region, which extends beyond the RVxF SLiM, constitutes the full PP1 interaction domain.

The discovery of a novel PP1 interaction SLiM, the KiR-SLiM
The Ki-67/RepoMan residues C-terminal to the canonical RVxF motif are not present in any other PP1 regulator whose holoenzyme structure is known, suggesting these regulators bind PP1 using a novel mechanism. To identify this mechanism, we determined the crystal structures of the Ki-67 496-536 :PP1g 7-308 and RepoMan 383-423 :PP1g 7-308 holoenzymes complexes to 2.0 Å and 1.3 Å , respectively (Figure 1C,D; Figure 1-source data 1). As predicted by the ITC and NMR studies, Ki-67 and Repo-Man bind PP1 using an identical mechanism. Namely, Ki-67 and RepoMan form a classical b-hairpin on the top of PP1 that extends from the PP1 RVxF binding pocket towards the PP1 N-terminus and then back again ( Figure 1D;~2600 Å of buried surface area). The structures also show that two established PP1-specific SLiMs-the RVxF-SLiM (Egloff et al., 1997) Figure 1D). However, as predicted, the Ki-67/RepoMan residues C-terminal to these SLiMs (aa 517-535/404-422) also bind PP1 and do so in a manner never observed for any other PP1 regulator ( Figure 1D,E). Residues F517 Ki-67 /F404 RM and P523 Ki-67 / A410 RM are anchored to PP1g via two hydrophobic pockets while multiple residues in Ki-67/Repo-Man bind the side chains of R74 PP1 , Y78 PP1 and Q294 PP1 via polar and salt bridge interactions. This The only region of homology between the two proteins is indicated in blue. The sequences corresponding to the homologous regions are shown below, with conserved residues highlighted in grey. Ki-67 residues that interact directly with PP1 are underlined. The sequences corresponding to the RVxF and FF SLiMs (blue highlight) and the newly discovered KiR-SLiM (orange) are shown. (B) left, binding isotherm of Ki-67 496-536 with PP1g 7-323 (K D , 193 ± 16 nM; the K D of the corresponding domain of RepoMan 383-423 with PP1g 7-323 is 133 ± 16 nM); right, binding isotherm of the DKiR-SLIM, RepoMan 383-404 , with PP1g 7-323 (K D , 661 ± 160 nM). (C) Crystal structure of the Ki-67:PP1g holoenzyme. PP1g is in grey and Ki-67 496-536 is in pink with the 2F o -F c electron density map contoured at 1s (2.0 Å ); no electron density was observed for Ki-67 residues 496-503 (pink dotted line) and 536. PP1 residues in green correspond to PP1 secondary structure elements helix A', loop 1 and helix B. (D) Close-up of the Ki-67 (pink) and RepoMan (blue) interaction with PP1. Ki-67 residues 503-516 and RepoMan residues 390-403 bind the PP1 RVxF and FF binding pockets (cyan surface; the Ki-67 and RepoMan RVxF and FF SLiM residues are labeled and in italics). Ki-67 residues 517-535 and RepoMan residues 404-422 bind the newly defined KiR-SLiM binding pocket (beige surface). The black dotted line highlights the area shown in E. (E) The hydrophobic and polar interactions between Ki-67 (pink sticks) and PP1g (grey sticks; surface). Hydrogen bonds and salt bridge interactions are indicated by dotted lines with the interacting residues labeled. (F) HMMER-derived sequence logo of the Ki-67/ RepoMan PP1 binding domain, with the KiR-SLIM highlighted in beige (hydrophobic residues, black; acidic residues, red; basic residues blue; glycine/ serine/threonine, green; asparagine/glutamine, pink).  Figure 1 continued on next page orders PP1g L1 ( Figure 1C), a loop that is generally more dynamic in most free PP1 and PP1 holoenzyme structures as evidenced by its higher B-factors and less well-defined electron density. Although F517 Ki-67 /F404 RM are the most buried residues in both complexes, mutating this residue to an alanine does not negatively affect RepoMan function (Figure 1-figure supplement 4).
Because these residues are critical for binding (removing them decreases the affinity~five-fold; Figure 1B, right), we termed this novel PP1 interaction SLiM the KiR-SLiM (Ki-67-RepoMan SLiM). The general KiR-SLiM motif, FDxxLP(P/A)N(T/S)PL(R/K)(R/K)Gx(T/S)P was determined using HMMER ( Figure 1F) (Finn et al., 2015). Critically, a subsequent search of the UniProtK/Swiss-Prot database (de Castro et al., 2006) using the most degenerate version of the KiR-SLiM identified only Ki-67 and RepoMan proteins. This demonstrates that, unexpectedly, only these two PP1 regulators likely use the KiR-SLiM interaction surface on PP1.  The specific recruitment of PP1g by Ki-67 and RepoMan How PP1 regulators selectively recruit specific PP1 isoforms to distinct substrates remains an important, open question. PP1 has three isoforms-a, b and g-that differ primarily in the last~30 residues of their C-termini; because of this, it has been generally assumed that regulators that bind preferentially to one isoform interact directly with the C-terminus . In vivo, Ki-67 and RepoMan bind specifically to the b-and g-isoforms, but not the a-isoform, of PP1 (Booth et al., 2014;Trinkle-Mulcahy et al., 2006;Vagnarelli et al., 2011). As PP1g is recruited more efficiently than PP1b (Booth et al., 2014), we focused our study on the PP1a and PP1g isoforms. We confirmed this in vitro using ITC, which showed that the affinity of both Ki-67 496-536 and RepoMan 383-423 for PP1a 7-330 is~four-six-fold weaker than for PP1g 7-323 (Table 1, Figure 1-figure supplement 1). To elucidate the molecular basis of isoform selectivity, we first tested the role of the PP1g C-terminal residues. Deleting the last 15 residues of PP1g had no impact on binding (Table 1, Figure 1-figure supplement 1). We then tested if RepoMan binds PP1a differently by determining the structure of the RepoMan 383-441 :PP1a 7-300 holoenzyme (2.6 Å ; Figure 1-source data 1). The structures are identical (backbone RMSD = 0.22 Å ), demonstrating that the isoform selectivity is due to one or more of the six amino acid differences in the PP1 catalytic domain (residues 7-300). Only two of these differing residues are located near the Ki-67/RepoMan:PP1 interface, Arg20/Gln20 (g/a) and Lys23/Arg23 (g/a; Figure 2A). We generated variants of both isoforms in which one or both of these residues were mutated to that of the other (PP1g R20Q , PP1g R20Q/K23R and PP1a Q20R ) and determined their affinities for Ki-67 496-536 and RepoMan 383-423 using ITC ( Figure 2B; Table 1; Figure 1-figure supplement 1). The change of only a single amino acid, R20/Q20, switches PP1g into a 'PP1a'-like isoform (PP1g R20Q ; K D = 2239 ± 124 nM; ten fold weaker binding) and PP1a into a 'PP1g'-like isoform (PP1a Q20R ; K D = 199 ± 46 nM; similar binding to that of PP1g). Although R20 does not interact with Ki-67 or RepoMan directly, it confers selectively through its interactions with PP1 which order the L1 loop. Namely, the Arg20 sidechain forms a salt bridge with PP1 residue Glu77 and makes a planar stacking interaction (cation/p interaction) with Phe81. Neither interaction is possible with Q20, as the side chain is both uncharged and too short. Thus, only in PP1g is this pocket ordered and readily available for binding, which allows for the isoform specific interaction of Ki-67 and RepoMan.
Similarly, tethering experiments showed that a GFP:LacI fusion of Ki-67 301-700 wt or RepoMan wt results in the recruitment of co-expressed RFP-PP1g to a LacO array that is integrated at a single locus in DT40 chicken cells. In contrast, RFP-PP1g R20Q fails to localize, demonstrating it no longer binds Ki-67 and RepoMan ( Figure 2F,G, Figure 2-figure supplement 2). Remarkably, the R20Q mutation abolishes PP1g localization as effectively as mutating the canonical RVxF motif in RepoMan to RATA. Together, these data reveal that the isoform specificity of Ki-67 and RepoMan is not defined by residues in the PP1 C-terminus, but instead by a single residue in the PP1 catalytic domain, R20/Q20 (g/a). This is a fundamental result, as it demonstrates that regulator-specific isoform selectively is not only achieved through interactions with the C-terminus, but also through interactions with the structured catalytic domain.

Dynamic regulation of PP1g recruitment to chromosomes by Aurora B Kinase
Unlike PP1a, the localization of PP1g changes dramatically during mitosis, with the bulk of PP1g relocalizing to chromosomes at anaphase onset. This relocalization is dependent on its ability to  Figure 3A) (Vagnarelli et al., 2006;Qian et al., 2013;Vagnarelli et al., 2011;Qian et al., 2015). For example, RepoMan S400, T412 and T419 are phosphorylated by CDK1-cyclin B (Vagnarelli et al., 2011;Qian et al., 2015). Mutating these residues to phosphomimetics inhibits PP1g binding, as evidenced by the inability of EGFP-RepoMan 3D to pull-down PP1g from non-synchronized HEK293T cells [see Figure 3F in Qian et al. (2015)] and an inability of RFP-PP1g to localize to the LacO locus when co-expressed with GFP:LacI-RepoMan 3D ( Figure 3B). Notably, the identity of the kinase that phosphorylates the S507 Ki-67 and T394 RM sites, both of which have been identified in proteomic screens (Dephoure et al., 2008;Nousiainen et al., 2006), has remained elusive. Because Aurora B kinase is redistributed from centromeres to the spindle midzone at anaphase onset and because the S507 Ki-67 and T394 RM sequences match canonical Aurora B kinase phosphorylation motifs (Kettenbach et al., 2011), we reasoned that these residues are phosphorylated by Aurora B kinase. Using in vitro phosphorylation assays coupled with NMR spectroscopy and mass spectrometry, we showed that both S507 Ki-67 and T394 RM ( Figure 3C Table 1). Finally, the expression of GFP-LacI-Repo-Man T394D results in a significant reduction of the recruitment of RFP-PP1g to the LacO locus ( Figure 3B), i.e., to a level that is nearly identical to that observed for GFP-LacI-RepoMan RATA and GFP-LacI-RepoMan 3D . These data demonstrate that phosphorylation by Aurora B kinase in (pro) metaphase inhibits PP1g binding to Ki-67 and RepoMan and, as a consequence, significantly contributes to prevent their premature recruitment to chromosomes.

Conclusion
Our study builds on previous work to reveal how two mitotic exit phosphatases (Ki-67:PP1 and Repo-Man:PP1) are assembled and recruited to their cellular targets, how they selectively bind the g-isoform of PP1 and how the assembly of these holoenzymes is controlled by Aurora B kinase phosphorylation. These are key advances as, until this work, there was essentially no molecular data on mitotic phosphatase assembly and function. Our data now explains why the phosphorylation of RepoMan at three distinct sites (S400, S412 and S419) by Cdk1 inhibits PP1 binding (Qian et al., 2015). Namely, it is similar to the mechanism by which Aurora B kinase inhibits Ki-67:PP1 and Repo-Man:PP1 holoenzyme formation. All three residues, like the Aurora B targets S507 Ki-67 and T394 RM , are part of the PP1 interaction motif and their phosphorylation blocks the interaction with PP1, which, in turn, prevents the premature targeting of PP1 to chromatin. This regulation of holoenzyme assembly is critical as the premature targeting of PP1 to chromosomes leads to an increase in chromosome misalignment and a weakened spindle assembly checkpoint (Qian et al., 2015). In contrast, the expression of a PP1-binding mutant of RepoMan in HeLa cells results in extensive cell death, demonstrating the importance of the RepoMan-PP1 interaction for cell viability . This work also expands the diversity of SLiMs used by PP1 regulators to bind PP1 and demonstrates a novel mechanism by which isoform selectivity is achieved in PP1 holoenzymes, a question that has been under investigation for the last 20 years. That is, Ki-67 and RepoMan selectively assemble with PP1g because the KiR-SLiM distinguishes between a single residue in PP1: Arg20 PP1g and Gln20 PP1a . Finally, since the KiR-SLiM is present in just two regulators, the newly discovered KiR-SLiM binding pocket defines a novel surface on PP1 that is targetable for the development of drugs that inhibit only 1% of the distinct PP1 holoenzymes in the cell (2 of~200) (Hendrickx et al., 2009). This is of profound interest because both Ki-67 and RepoMan are highly upregulated in multiple cancers (Dowsett et al., 2011;Vagnarelli, 2014). Targeting unique regulator binding pockets is a powerful approach as we recently discovered that the newly discovered LxVP SLiM substrate interaction surface is a well-known drug target site in the ser/thr phosphatase Calcineurin (Grigoriu et al., 2013). Given the diversity of unique interactions that are now being revealed between PSPs and their plethora of regulators, it is now clear that these novel, unique protein:protein interaction sites, especially that defined by the KiR-SLiM, provide immediate opportunities for the design of novel, highly specific therapeutics.

Crystallization and structure determination
The Ki-67 496-536 :PP1g 7-308 holoenzyme crystallized in 1.9 M Sodium Malonate pH 4.0 (hanging drop vapor diffusion at 4˚C). Crystals were cryo-protected by a 60 s soak in mother liquor supplemented with 40% glycerol and immediately flash frozen. Data for the Ki-67 496-536 :PP1g 7-308 holoenzyme crystal structure were collected to 2.0 Å at beamline 12.2 at the Stanford Synchrotron Radiation Lightsource (SSRL) at 100 K and a wavelength of 0.98 Å using a Pilatus 6M PAD detector. The Ki-67 496-536 :PP1g 7-308 crystal data was processed using XDS (Kabsch, 2010), Aimless (Evans and Murshudov, 2013) and Truncate (French and Wilson, 1978). The data was analyzed using phenix.xtriage and the intensity statistics suggested merohedral twinning (twin fraction of 0.48) with the space group of P6 1 and twin law h, -h-k, -l. The structure was solved by molecular replacement using Phaser (McCoy et al., 2007) as implemented in PHENIX (Zwart et al., 2008) (PDBID 5INB was used as the search model). The model was completed using iterative rounds of refinement in PHENIX (McCoy et al., 2007) and manual building using Coot (Emsley et al., 2010) (Ramachandran statistics: 95.05% favored, 4.95% allowed). The final structure was refined in PHENIX with the twin law. The RepoMan 383-423 :PP1g 7-308 complex crystallized in 1 M Sodium Malonate pH 4.3 (sitting drop vapor diffusion method at 4˚C). Crystals were cryo-protected by a 30 s soak in 1 M Sodium Malonate pH 4 supplemented with 40% glycerol and immediately flash frozen. Data for the RepoMan 383-423 : PP1g 7-308 holoenzyme crystal structure were collected to 1.3 Å at the beamline 12.2 at Stanford Synchrotron Radiation Lightsource (SSRL) at 100 K and a wavelength of 0.98 Å using a Pilatus 6M PAD detector. The RepoMan 383-423 :PP1g 7-308 crystal data were processed using XDS (Kabsch, 2010), Aimless (Evans and Murshudov, 2013) and Truncate (French and Wilson, 1978). The structure was solved by molecular replacement using Phaser (McCoy et al., 2007) as implemented in PHENIX (Zwart et al., 2008) (PDB ID 1JK7 (Maynes et al., 2001) was used as the search model). A solution was obtained in space group P6 1 22. The model was completed using iterative rounds of refinement in PHENIX and manual building using Coot (Emsley et al., 2010) (Ramachandran statistics: 96.3% favored, 3.7% allowed). The RepoMan 383-441 :PP1a 7-300 holoenzyme crystallized in 100 mM Sodium Malonate pH 4.0, 12% PEG 3350. Crystals were cryo-protected by a 30 s soak in mother liquor supplemented with 30% glycerol and immediately flash frozen. Data were collected to 2.6 Å at the National Synchrotron Light Source (BNL) Beamline X25 at 100 K and a wavelength of 1.1 Å using Dectris pilatus 6M detector. Data were indexed, scaled and merged using HKL2000 0.98.692i (Otwinowski and Minor, 1997). The structure was solved by molecular replacement using Phaser as implemented in PHENIX (PDB ID 4MOV was used as the search model (Choy et al., 2014;McCoy et al., 2007;Zwart et al., 2008). A solution was obtained in space group P2 1 2 1 2 1 . The model was completed using iterative rounds of refinement in PHENIX (McCoy et al., 2007) and manual building using Coot (Emsley et al., 2010) (Ramachandran statistics: 95% favored, 5% allowed).

In vitro phosphorylation
Aurora Kinase B (AuKB) was expressed and purified as a GST-tagged protein as previously described (Qian et al., 2013). In-vitro phosphorylation of Ki-67 496-536 or RepoMan 348-450 was achieved by incubation with AuKB at a molar ratio 20:1 in phosphorylation buffer (20 mM HEPES pH 7.5, 0.5 mM EDTA, 2 mM DTT, 20 mM MgCl 2 and 10 mM ATP). The reaction was allowed to proceed for 16 hr at 30˚C. Following incubation, the protein was concentrated and purified using SEC. The phosphorylation of both Ki-67 496-536 and RepoMan 348-450 was confirmed using ESI-MS.

RepoSLiM identification
HMMER (Finn et al., 2015) (using the rp75 UniProt database) identified proteins with sequences similar to the RepoMan PP1-binding domain. 107 sequences were identified, spanning a diversity of species, from human to Xenopus. WebLogo was used to generate the resulting logo. The Uni-ProtKB/Swiss-Prot database was then scanned using ScanProsite (de Castro et al., 2006)

Cell culture and RNA interference
HeLa Kyoto cells were maintained in DMEM supplemented with 10% FBS. DT40 cells carrying a single integration of the LacO array (Vagnarelli et al., 2006) were cultured in RPMI1640 supplemented with 10% FBS and 1% chicken serum. For RNAi treatments, HeLa cells in exponential growth were seeded in 6 well plates with polylysine-coated glass coverslips and grown overnight. Transfections were performed using Polyplus jetPRIME (PEQLAB; Germany) with the indicated siRNA oligos and analyzed 48 hr later as previously described after 3 hr nocodaole arrest (Vagnarelli et al., 2006). For the rescue experiments HeLa cells at 50% confluence were transfected with 400 ng of plasmid DNA and 50 nM of siRNA oligonucleotides and analyzed 48 hr post-transfection after 3 hr nocodaole arrest. Transient transfections for DT40 were conducted as previously described (Qian et al., 2013). For quantification of the enrichment at the Laci locus, cells were fixed with paraformaldehyde 24 hr after transfection.

Live imaging
HeLa cells were seeded in a 4-Chamber 35 mm Glass Bottom Dish with 20 mm microwell, #1 cover glass (Cellvis) and transfected with X-tremeGENE 9 DNA Transfection Reagent (Roche) according to manufacturer's protocol. The DNA was stained with Hoechst 33,342 (Tocris; United Kingdom) for live imaging. Confocal images were acquired with a Leica TCS SPE laser-scanning confocal system mounted on a Leica DMI 4000B microscope, equipped with a Leica ACS APO 63X 1.30NA oil DIC objective and a live-imaging chamber ensuring 37˚C and 5% CO 2.

Indirect immunofluorescence and microscopy analyses
Cells were fixed in 4% PFA and processed as previously described (Rosenberg et al., 2011). 3D data sets were acquired using a cooled CCD camera (CoolSNAP HQ2 firewire) on a wide-field microscope (Eclipse Ti, NIKON) with a NA 1.45 Plan Apochromat lens. The data sets were deconvolved with NIS-Element AR (NIKON). Three-dimensional data sets were converted to MIP in NIS-Element AR, exported as TIFF files, and imported into Adobe Photoshop for final presentation.