topological control of cytokine receptor signaling induces ... · clone, n3c-e2, was expressed in...

RESEARCH ARTICLE SUMMARY◥

PROTEIN ENGINEERING

Topological control of cytokinereceptor signaling induces differentialeffects in hematopoiesisKritika Mohan*, George Ueda*, Ah Ram Kim†, Kevin M. Jude†, Jorge A. Fallas†,Yu Guo†, Maximillian Hafer, Yi Miao, Robert A. Saxton, Jacob Piehler,Vijay G. Sankaran, David Baker‡, K. Christopher Garcia‡

INTRODUCTION: Receptor dimerization is afundamental mechanism by which most cyto-kines and growth factors activate Type-I trans-membrane receptors. Althoughprevious studieshave shown that ligand-induced topologicalchanges in the extracellular domains(ECDs) of dimeric receptors can affectsignaling output, the physiologicalrelevance is not well understood. Thisis a difficult problem to study becausean engineered ligand systemdoes notexist that would enable a systematicexplorationof the relationship betweenligand-receptor dimer geometry andsignaling. This question, as well as thecapability of exploiting these structure-activity relationships for drugdiscovery,are important because most cytokinesexert pleiotropic effects that limit theirtherapeutic utility. New approaches areneeded to modulate cytokine signalingand identify clinically efficacious var-iants. By contrast, small-molecule ligandsfor G protein–coupled receptors havebeen successfully discovered throughmedicinal chemistry and used to in-duce conformational changes that leadto physiologically relevant biasedsignaling outputs. A comparableapproach for dimeric receptors couldopen a path to new pharmacologicalparameters for cytokines and growthfactors.

RATIONALE: In order to better un-derstand how the extracellular structureof cytokine-receptor complexes affects down-stream signaling events, we developed anengineered ligand system to precisely con-trol the orientation and proximity of dimericreceptor complexes that would enable themeasurement of structure-activity relation-ships between receptor dimer geometry,signaling, and function. We applied thisapproach to design geometrically controlledligands to the erythropoietin receptor (EpoR)

system, a well-characterized dimeric cytokinereceptor system.

RESULTS: We used the DARPin (designedankyrin repeat protein) scaffold because of

its modular nature. We isolated a high-affinityDARPin toEpoRusing yeast display and in vitroevolution and determined the crystal structureof the DARPin/EpoR complex. We then con-verted these monomeric DARPin binding mod-ules into C2 symmetric homodimeric agonistsby incorporating in silico designed dimerizationinterfaces. This rigidly connected dimeric DARPinscaffold then enabled us to design a series ofextended ligands through sequential insertionof ankyrin repeat “spacers” to systematically

control the relative orientation of the ECDs inthedimeric complex. The “angle” series varied thescissor angle between the two ECDs, whereasthe “distance” series varied their relative proxim-ity. The designedDARPin ligandswere validatedbymeans of x-ray crystallography for representa-tive complexes.The systematic variationofangularand distance parameters generated a range of full,biased, and partial agonism of EpoR signaling

in thehumanerythroid cellline UT7/EPO, as shownwith flow-cytometry andimmunoblotting for phos-phorylated downstreameffectors. In general, in-creasing the angle or dis-

tance between the receptor ECDs resulted ina progressive partial agonism, asmeasuredwithchanges in maximum response achieved (Emax)andmedianeffective concentration (EC50).Biasedsignal transducer and activator of transcription(STAT) activation was elicited by some of the sur-rogate DARPin ligands.We also evaluated the

effects of these DARPin agonists on dif-ferentiation and proliferation of hemato-poietic stem and progenitor cells (HSPCs)that were maturing into the erythroidlineage. The partial agonists displayedstage-selective effects onHSPCs,whereasthe biased agonists more selectively pro-moted signaling at either the early or latestages of differentiation.

CONCLUSION: We have designed aseries of surrogate ligands to system-atically alter receptor dimer topologyand modulate signaling outputs. Animportant feature of these ligands isthat the dimerized modules are rigidlyconnected, thus allowing us to deter-mine structures of the ligand-receptorcomplexes, or model them accurately,and correlate precise geometries withsignaling output. Some variants inducestage-selective differential signaling inprimary cells. This “topological tuning”maybe attributed to altered intracellularorientations of Janus kinase 2 (JAK2)relative to its substrates or mechanicaldistortion that leads to changes in com-plex stability and receptor internalization.Althoughour experiments donot reveal apredictive framework to relate topology to

signaling, this approach canbeused to empiricallyidentify specific dimer topologies that correlatewith therapeutically desirable signaling outputsfor any dimeric receptor system.▪

RESEARCH

Mohan et al., Science 364, 750 (2019) 24 May 2019 1 of 1

The list of author affiliations is available in the full article online.*These authors contributed equally to this work.†These authors contributed equally to this work.‡Corresponding author. Email: [email protected] (D.B.);[email protected] (K.C.G.)Cite this article as K. Mohan et al., Science 364, eaav7532(2019). DOI: 10.1126/science.aav7532

Three different EpoR/DARPin complex topologies are shownthat elicit full, partial, and null agonism. Full agonism, red;partial, green; and null, blue.The EpoR/DARPin structure shownon the green cell is from a crystal structure, whereas thecomplexes shown on the red and blue cells are models.

IMAGE:K.C

.GARCIA

AND

E.S

MITH

ON OUR WEBSITE◥

Read the full articleat http://dx.doi.org/10.1126/science.aav7532..................................................

on January 20, 2021

http://science.sciencemag.org/

Dow

nloaded from


RESEARCH ARTICLE◥

PROTEIN ENGINEERING

Topological control of cytokinereceptor signaling induces differentialeffects in hematopoiesisKritika Mohan1*, George Ueda2,3*, Ah Ram Kim4,5,6†, Kevin M. Jude7†,Jorge A. Fallas2,3†, Yu Guo8†, Maximillian Hafer9, Yi Miao1, Robert A. Saxton1,7,Jacob Piehler9,10, Vijay G. Sankaran4,5,6, David Baker2,3,11‡, K. Christopher Garcia1,7,12‡

Although tunable signaling by G protein–coupled receptors can be exploited throughmedicinalchemistry, a comparable pharmacological approach has been lacking for the modulation ofsignaling through dimeric receptors, such as those for cytokines.We present a strategy tomodulate cytokine receptor signaling output by use of a series of designed C2-symmetriccytokine mimetics, based on the designed ankyrin repeat protein (DARPin) scaffold, that cansystematically control erythropoietin receptor (EpoR) dimerization orientation and distancebetween monomers.We sampled a range of EpoR geometries by varying intermonomer angleanddistance, corroboratedbyseveral ligand-EpoRcomplexcrystal structures. Across the range,we observed full, partial, and biased agonism as well as stage-selective effects onhematopoiesis.This surrogate ligand strategy opens access to pharmacological modulation oftherapeutically important cytokine and growth factor receptor systems.

Cytokines are secreted proteins that medi-ate awide range of functions, principally inthe immune system, but they also exertmany other homeostatic functions thataffect mammalian physiology (1–3). As one

example, hematopoiesis, the process by whichblood cells are formed, is influenced at both earlyand late stages by amilieu of different cytokinesthat engage and activate specific cell-surface re-ceptors expressed on different lineages of cells(4). When cytokines bind to their receptors, re-ceptor dimerization triggers transphosphorylationof Janus (JAK)/TYK kinases, and subsequent

propagation of a phosphorylation cascade lead-ing to signal transducer and activator of tran-scription (STAT) activation and induction of geneexpression programs (3, 5). Structural and func-tional studies of cytokine-receptor complexestogether with similar studies on growth factordimeric receptor systems, including receptortyrosine and serine-threonine kinases, showthat a wide range of receptor dimer geometriesare compatible with signaling (3, 5–9).Given their powerful actions, many natural

cytokines have been investigated as therapeuticagents. Whereas select cytokines are clinicallysafe and effective, such as erythropoietin (EPO)(10), most natural cytokines, especially immunecytokines, have pleiotropic effects on many dif-ferent cells and stimulate multiple downstreamsignaling responses, resulting in toxicity or lackof efficacy (11, 12). Nevertheless, cytokines havepotential as therapeutics if their actions can beharnessed, as exemplified by recent advancesfor cytokines such as interleukin-2 (IL-2) andIL-10 (13, 14). Recent studies, using natural andengineered ligands to both cytokine and recep-tor tyrosine kinase (RTK) class receptors, showthat signal strength can be modulated down-stream of dimeric receptors (15–18). Further-more, select natural and synthetic mutants ofcytokines such as IL-2, human growth hormone(hGH), stem cell factor (SCF), and EPO that ex-hibit partial and biased signaling properties havebeen identified (17, 19, 20). In G protein–coupledreceptors (GPCRs), biased signaling is physiolog-ically relevant and generally depends on bindingof small-molecule ligands,making this exploitableas a drug discovery strategy by using medicinal

chemistry (21). Finding small-molecule drugs isnot currently feasible for Type-I transmembranereceptors that bind to protein ligands, such ascytokines, through large extracellular domains(ECDs). Nevertheless, ligand engineering strat-egies that take advantage of the potential forbiased signaling behavior by dimeric cytokinereceptors could be used as probes to better un-derstand cytokine receptor signaling biology butalso may have therapeutic potential.The overall geometry of the cytokine receptor

dimer can influence the strength of the down-stream signal (6, 22–24); however, approaches donot exist to precisely control the orientation andproximity of ligand-receptor complexes. The abilityto topologically control receptor dimerizationcould open access to a newpharmacologicmetricfor assessing structure-activity relationshipsof cytokine and growth factor receptor signal-ing. Previous studies have shown that intactantibodies (25, 26) and diabodies can act as sur-rogate cytokine agonists capable of modulatingsignaling outputs through receptor dimerization(24, 27), but their segmental flexibility preventscontrol of the precise geometry or distance be-tween monomers.We integrated a previous protein-binding

scaffold technology, termed DARPin (designedankyrin repeat protein) (28), with protein designand engineering approaches to create a self-assembling, rigidly connected dimeric scaffoldthat enables precise control of the relative orien-tations of the receptor ECDs in a dimeric complex(Fig. 1A). We focused on the erythropoietin recep-tor (EpoR), which plays a critical role in erythroidlineage commitment and differentiation to pro-ducemature red blood cells (RBCs) (20). We foundthat systematic variation of angular and distanceparameters of the DARPin-based agonists leads topartial and biased agonism of the EpoR signal. Fur-thermore, this variation in geometry has allowedus to selectively activate EpoR signaling at specificstages of hematopoiesis. Collectively, these studiesoffer a general strategy for the regulation of cyto-kine receptor signaling through topological con-trol of the receptor dimer that can be applied toother dimeric receptor systems.

ResultsSelection of monomeric DARPinsselective for EpoR

Our overall strategy was to engineer a proteinmodule that binds with high affinity to a cyto-kine receptor ECD and subsequently design themodule to form a range of dimeric geometriesthat would in turn constrain the dimerizationgeometry of the bound receptor (Fig. 1A). Wefocused on a well-characterized system, humanEpoR, as our cytokine receptor target for topo-logical tuning. We chose DARPins as the bind-ing scaffold because they are modular, whichenables size variation (28), and have been used todesign self-assembling oligomers (29). A DARPinmolecule typically consists of “X” number ofankyrin repeats flanked by N- and C-terminalcapping regions, resulting in “NXC” DARPins.Each ankyrin repeat is formed by 33 amino

RESEARCH

Mohan et al., Science 364, eaav7532 (2019) 24 May 2019 1 of 15

1Department of Molecular and Cellular Physiology, StanfordUniversity School of Medicine, Stanford, CA 94305, USA.2Department of Biochemistry, University of Washington,Seattle, WA 98195, USA. 3Institute for Protein Design,University of Washington, Seattle, WA 98195, USA. 4Divisionof Hematology/Oncology, The Manton Center for OrphanDisease Research, Boston Children’s Hospital, HarvardMedical School, Boston, MA 02115, USA. 5Department ofPediatric Oncology, Dana-Farber Cancer Institute, HarvardMedical School, Boston, MA 02115, USA. 6Broad Institute ofMIT and Harvard, Cambridge, MA 02142, USA. 7HowardHughes Medical Institute, Stanford University School ofMedicine, Stanford, CA 94305, USA. 8State Key Laboratoryof Medicinal Chemical Biology and College of Pharmacy,Nankai University, Tianjin, People’s Republic of China.9Division of Biophysics, Department of Biology, University ofOsnabrück, 49076 Osnabrück, Germany. 10Center for CellularNanoanalytics, University of Osnabrück, 49076 Osnabrück,Germany. 11Howard Hughes Medical Institute, University ofWashington, Seattle, WA 98195, USA. 12Department ofStructural Biology, Stanford University School of Medicine,Stanford, CA 94305, USA.*These authors contributed equally to this work. †These authorscontributed equally to this work.‡Corresponding author. Email: [email protected] (D.B.);[email protected] (K.C.G.)

on January 20, 2021


Dow

nloaded from


acids and consists of a b turn followed by twoantiparallel a helices. Twenty-seven of these po-sitions are highly conserved in naturally occur-ring ankyrin repeats and crucial to maintainingthe repeat protein framework (30, 31), whereasthe remaining six positions can support random-ization for library generation (Fig. 1B, blue resi-dues). We developed two yeast-displayed DARPinlibraries, N2C and N3C (Fig. 1B and fig. S1), whichwere screened against EpoR ECD, resulting in aset of low-affinity ligands with dissociation con-stants (Kd) in the micromolar range (Fig. 1C andfig. S2). A combination of random mutagenesisby means of error-prone polymerase chain re-action (PCR) and DNA-shuffling techniquesresulted in a >104-fold increase in affinity of indi-vidual clones (Fig. 1, C and D). The highest-affinityclone, N3C-E2, was expressed in Escherichia coli,and its affinity was measured with surface plas-mon resonance (SPR). E2 binds to EpoR ECDwitha 1.89 nM Kd, providing a >1000-fold increase inbinding affinity relative to 8A9 (Fig. 1D). We de-note this DARPin as NEEEC, comprising threeEpoR binding ankyrin repeats, EEE, and theN- andC-terminal capping regions.The crystal structure of NEEEC (Fig. 1, E to G,

and table S1) shows that it binds to the EpoRmembrane-distal D1 domain with a 1:1 stoichi-

ometry, at a site different but overlapping withthe EPO binding site, using three ankyrin re-peat domains. The b turn loop regions of theankyrin domains engage the top of the D1 do-main, whereas the helical regions pack againstthe side of the D1 domain with high shapecomplementarity and extensive buried sur-face area.

Generation of agonist DARPin dimers

In order to endow the monomeric DARPinNEEEC with the ability to dimerize EpoR andact as an agonist, it was converted into twodifferent homodimeric geometries by incorporat-ing designed dimer interfaces at the nonbindingface of the DARPin (29). The interface mutationsprovide hydrophobic side chain interactions atnonsolvent accessible positions and salt bridgeinteractions at solvent-exposed positions, whichresults in the formation of noncovalent self-assembling homodimers, A_R3 and C_R3, whereRx indicates the number of ankyrin repeats (Fig. 2,A and B, and figs. S3 and S4). The A and C dimersdiffer by a shift in the relative orientation of themonomers across the C2 axis. Dimer C_R3 has anoffset of one repeat at the dimerizing interfacecompared with dimer A_R3, providing a twist of30° in the relative orientation of the two di-

merized EpoR ECDs (Fig. 2B and fig. S5). Theformation of the C_R3 DARPin dimer and a2:2 C_R3/EpoR complex was confirmed withsize exclusion chromatography, and crystal struc-tures were determined to 1.2 and 3.2 Å resolu-tion, respectively. We assessed the ability of theDARPin dimers to dimerize EpoR on the surfaceof live cells using single-molecule total internalfluorescence (TIRF) microscopy (Fig. 2C and fig.S6). For this purpose, EpoR fused to anN-terminalmonomeric XFP (mXFP)–tag was expressed inHeLa cells at physiologically relevant densities(~0.5 to 1 copies/mm2; ~1000 to 2000 copies/cell)and labeled by means of specific nanobodies.EpoR dimerization was quantified through dual-color cotracking of individual receptor dimers(fig. S6 and movie S1). Like EPO, both ligandsshowed ligand-dependent dimerization of EpoRat the cell surface, although C_R3 was less ef-ficient than A_R3 (Fig. 2C, fig. S6, andmovie S2).Dimerization of EpoR initiates a sequential

cascade of phosphorylation events, the initialmembrane proximal step being phosphorylationof JAK2, followed by tyrosines on the EpoR intra-cellular domain (ICD), and thenprincipally activa-tion of STAT5 and extracellular signal–regulatedkinase (ERK), and to a lesser extent STAT1and STAT3. Other pathways not surveyed here,


Fig. 1. Engineering and characterizationof high-affinity DARPin binding toEpoR. (A) An overview of our topologicalcontrol strategy, using EpoR as a modelsystem to investigate modulation of signalingresponses. (B) Schematic representationof the yeast-displayed N3C DARPin constructused for library generation. The randomizedpositions are shown in blue. (C) Histogramsof yeast-displayed naïve N3C library, poolof DARPin clones after in vitro evolution,and pool of shuffled library clones afterin vitro evolution binding to EpoR. MFI, meanfluorescence intensity. (D) Representativesurface plasmon resonance sensogramsfor pre- and post-shuffled libraryN3C DARPin clones binding to EpoR. Kd,dissociation constant. (E) Structure ofN3C E2 binding to EpoR. (F) The structureof E2/EpoR (PDB 6MOE) is aligned withEPO/EpoR to indicate relative binding sitesof the two ligands. EPO is depicted asgreen cylinders. (G) The intimate shapecomplementarity of E2 binding to EpoR isdepicted with the side chains shown forkey interacting E2 residues.

RESEARCH | RESEARCH ARTICLEon January 20, 2021


Dow

nloaded from


including mitogen-activated protein kinase(MAPK) and phosphatidylinositol 3-kinase (PI3K),are also activated by EPO. We analyzed thephosphorylation levels of various downstreameffectors on the human erythroid cell line, UT7/EPO (32), using phospho-flow cytometry (Fig. 2D).Monomeric DARPins do not dimerize EpoR, gen-erating no pSTAT5 signal. Both DARPin dimersinduce comparable efficacy (Emax; the maximumresponseachieved) comparedwithEPOforpSTAT5and pERK. Dimer A_R3 exhibits a higher Emaxfor pSTAT1 andpSTAT3 thanEPO,whereas dimerC_R3 exhibits a lower Emax (Fig. 2D). Thus, therelatively small orientational differences betweendimers A_R3 and C_R3 result in apparent sig-naling differences.

Topological control of EpoRdimerization geometry

Tomove toward topological control, we designeda series of NEEEC homodimers, systematically

varying the scissor angle (rotation) and proximitybetween EpoR ECD monomers (distance). Wedesigned three different DARPin dimer extensionseries to interrogate signaling as a function ofthese parameters in a systematicmanner (Fig. 3).The “angle” series is intended to pivot the EpoRdimer in a scissorlike manner (Fig. 3, top). The“distance” series is intended to separate theEpoR subunits in the dimer while not substan-tially changing the dimer angle (Fig. 3, middle).The “mono-extended” series differs from the angleanddistance series in that bothEpoR subunits arebound to one extendedDARPin (monomeric) thatcontains two EpoR binding sites (Fig. 3, bottom).Systematic variation in each of these parame-

ters can be achieved by inserting nonbinding “P”repeats derived from the consensus ankyrin repeatsequence, which changes the relative postions ofthe EpoR binding repeats across the designeddimerization interface in predictable ways thatcan bemodeled. Because of the nonplanar, helical

nature of the DARPin molecule (33), this resultsin dimers in which the two EpoR binding inter-faces assume different separations and orienta-tions (Fig. 3). To engineer the “angle” series, oneor more P repeats are inserted at the C-terminalend to provideDARPins such asNEEEP..C, (Fig. 3,top, and fig. S7). The dimerizing residues (in-dicated with underline) are engineered on thesame terminus as that of the P repeat insertions.As the added repeats grow along a helical path,this results in the generation of a series of dimerswith variation in the angle between the two re-ceptor ECD monomers. The resultant extendeddimer is termed A_angle_Rx, where A indicatesthe designed dimer interface, and “x” indicatesthe total number of ankyrin repeats. For exam-ple, NEEEPPC is termed A_angle_R5.Toengineer the “distance” series, thenonbinding

P repeats are inserted at the N terminus—forexample, NPP…EEEC (Fig. 3, middle, and figs.S8 and S9). The dimerizing residues are also


Fig. 2. Dimerizationscaffolds for E2 resulting inagonism. (A and B) Pointmutations (shown in green)inserted on the rear face of theDARPin lead to the formationof noncovalent homodimerswith two different geometries.A schematic representationis shown for formation of the2:2 DARPin:EpoR complex.An N3C DARPin containingthree ankyrin repeats isrepresented by three spheres.The EpoR binding face of theankyrin repeat is shown as“E” (pink), and the dimerizingface is shown as “D” (green),whereas the two EpoRdomains D1 and D2 are shownin teal. The dimer interfacefor C_R3 (PDB 6MOG) is offsetby one repeat providingadditional twist in the EpoRgeometry in C_R3/EpoR (PDB6MOH) (B). (C) Ligand-induced dimerization of EpoRon the surface of cells, asobserved with dual-colorsingle-molecule TIRF micros-copy. (Inset) The method ofsingle-molecule cotrackinganalyses. (D) Dose-responsecurves for STAT1/3/5 and ERKactivation performed in UT7/EPO cells stimulated with EPO,A_R3, or C_R3 for 15 min. Dataare mean ± SD for twoindependent replicates. A sig-moidal dose-response analysiswas preferred over Gaussiandistribution because onlysome ligands demonstrate adecrease in MFI values at high concentrations.



Dow

nloaded from


placed at the N terminus so that this results intwo helical DARPins growing in opposite di-rections with a twofold symmetry. With the ad-dition ofmore P repeats, the binding repeatsmovefurther apart, leading to a distance-based series,termed A_dist_Rx.Last, we designed a monomeric, extended

DARPin with two binding sites included on thesame DARPin chain (Fig. 3, bottom, and fig. S9),

butwith the boundEpoRs in an inverted rotation-al orientation relative to the dimerized DARPins(“face-to-back” versus “face-to-face”). The twosets of EpoR binding repeats are positionedat the two termini, separated by a series of P re-peat insertions that can be varied in number tocontrol the distance between the two receptors—for example, NEEEPPPP…EEEC. These mono-meric extended DARPins are termed M_Rx,

where “x” indicates the total number of ankyrinrepeats.Each of the dimers was expressed in E. coli,

purified through size-exclusion chromatography,and bound to EpoR with similar affinities, asmeasured with SPR (figs. S7 to S9). To validatethe designs and also to determine whether anyclashes were apparent between the EpoR recep-tors, we determined six crystal structures of


Fig. 3. Topological control of EpoR geometry.Three different series weredesigned to systematically modify the different components of receptordimerization—specifically, angle (A_angle_Rx), distance (A_dist_Rx),and inversion of geometry (M_Rx). The insertion (indicated by blue arrows)of nonbinding “P” ankyrin repeats (blue) at different termini of nonplanarDARPins leads to different receptor dimer topologies. The relative movement

of the two receptors with the insertion of P repeats is indicated by pink arrows.The series are generated through stepwise insertion of one P repeat, andthe nomenclature “Rx” indicates the total number of ankyrin repeats. Thecartoon representations are shown along with the crystal structures for onerepresentative member per series: A_angle_R5 (PDB 6MOJ), A_dist_R7(PDB 6MOK), and M_R12 (PDB 6MOL).



Dow

nloaded from


representative members of the series: the apoC_R3 dimer (1.2 Å resolution) as well as EpoRcomplexes with C_R3 (3.46 by 3.16 Å resolu-tion), A_angle_R5 (3.39 by 2.45 Å resolution),C_angle_R5 (3.0 by 2.0 Å resolution), A_dist_R7(5.1 Å resolution), and M_R12 (4.58 by 3.14 Åresolution) (Fig. 3, fig. S10, and tables S1 and S2).The structures validate the DARPin dimer de-signs, albeit with minor rigid body variationsbecause of lever arm effects centered at theDARPin dimer interface (fig. S11), allowing usto model the full geometry series based on theknown effects of P repeat insertion. In all of thestructures determined, whereas the D1 domainof EpoR maintains a rigid contact with theDARPin, the D2 domain exhibits some seg-mental flexibility and adapts variable positionsrelative to D1 that are likely influenced by crystalpacking.Wemeasured a range of 28° variation inD2 position (fig. S12), which is consistent withthat observed in previously published structuresof EpoR.

EpoR angle series

The basic building block of the angle series isA_R3 with zero P repeats (NEEEC), and sub-sequent members with one or more P repeatsadded in, A_angle_R4 (NEEEPC) throughA_angle_R7 (NEEEPPPPC), progressively tilteach EpoR monomer toward the membrane(Fig. 4A). As measured with phospho-flow cy-tometry in UT7/EPO cells, the dimer A_angleseries exhibits a progressive reduction in Emaxlevels of STAT5 phosphorylation as the EpoRdimer scissor half-angle increases from ~38°(A_R3) to 72° (A_angle_R9) (Fig. 4, A and B).Dimers A_angle_R4 and A_angle_R5 act aspartial agonists, whereas A_angle_R7 does notelicit a signal detectable with flow cytometry.Similarly, the C_angle series shows by meansof flow cytometry a stepwise decline in activa-tion that is confirmed with immunoblotting (fig.S13). In addition to the reduction in Emax, theeffects of increase in angle are also reflected bychanges in median effective concentration (EC50),which is defined as the ligand concentration thatprovides 50% of maximal response (Emax). Aprogressive right shift was observed for the EC50values, implying the need for a higher ligand con-centration to generate the same levels of signal,despite all of the dimers having the same affinityfor EpoR, as measured with SPR by using solubleEpoR ECD (fig. S7). We speculate that as thescissor angle between EpoRs becomes more ex-treme, strain is induced as a result of cell mem-brane constraints, and this is reflected in lessefficient bivalent complex formation.We also analyzed a broad panel of down-

stream phosphorylated proteins using phospho-rylation barcoding, which provides a broad“footprint” of relative signal strengths. This datashowed a similar trend of reduced Emax withincreasing scissor angle (Fig. 4C, fig. S13B, andtables S3 and S4). At higher concentrations, abias was observed for STAT1 and STAT3 by A_R3compared with EPO by means of phospho-flow.Because the phosphorylation barcoding data are

intended to convey qualitative relative signalstrengths, we more precisely confirmed withimmunoblotting the appearance of signal bias(Fig. 4D). A similar STAT1 bias was observedfor A_angle_R5 and C_angle_R4, with extremelyhigh pSTAT1 signal at highest concentrations(fig. S13). Immunoblot analysis of cell lysatesfurther verified the STAT1 bias observed for A_R3,A_angle_R5, andC_angle_R4 comparedwith EPO(Fig. 4D and figs. S13C and S14).

EpoR distance series

The basic building block of the distance seriesis A_R3 with zero P repeats (NEEEC), and sub-sequent members incorporate one or more Prepeats: A_dist_R4 (NPEEEC) through A_dist_R9(NPPPPPPEEEC) (Fig. 5A). Contrary to the angleseries in which a reduction in pSTAT5 and pERKEmax was observed, the first few members ofA_dist exhibit the sameEmax despite the increaseddistance between the receptors (Fig. 5B). In-stead, the effects of the distance are reflected ina right shift in the EC50 values. Furthermore,a bias toward higher pSTAT1 and pSTAT3 Emaxthan EPO or other distances was observed forA_dist_R5 (figs. S14 to S16). Analysis of a panel ofdownstreamphosphorylated proteins bymeans ofphospho-barcoding reflected the above trends, inwhich the strong biaswas observed for STAT1 forthe A_dist_R5 ligand (Fig. 5, C to E; figs. S14 andS15; and table S5). The biased and prolongedpSTAT1 response for A_dist_R5 was furtherverified through immunoblotting with 15- and60-min ligand treatments (Fig. 5D) and may beattributable in part to the differential intracel-lular processing rate for the internalized recep-tors (Fig. S17). In general, ligands with fasterinternalization of cell-surface EpoR upon ligandtreatment showed a more potent response. Fur-thermore, with an increase in angle or distance, adecrease in activity was observed, which corre-lates with the slower rate of internalization of thereceptor. Recycling of EpoR back to the surface isslower for A_dist_R5 than other DARPin ligands.For the distance series, the right shift of the EC50was surprising because we would not imaginestrain on the dimer through pure translation ofthe receptors to wider intermonomer distancesunless there was receptor preassociation thatcould present an energy barrier to separation.Although we did not see EpoR predimerizationwith single-molecule TIRFmicroscopy, increasingthe distance may interfere with weak synergisticinteractions between the receptor subunits, pos-sibly mediated by the transmembrane domains(34, 35).

Inverted EpoR geometry

We assessed the impact on signaling of relativeEpoR ECD rotation with respect to one anotherwhile maintaining an approximately similar dis-tance. In the native EPO/EpoR dimer structure,the receptors are face-to-face. Having validatedthe design of a face-to-back orientation with thecrystal structure of the DARPin M_R12/EpoRcomplex (Fig. 3), we assessed face-to-back orien-tation while maintaining a close intermonomer

distance (Fig. 5F) by comparing the signalingresponse for A_dist_R6 (face-to-face dimer, dis-tance 49Å) and M_R11 (face-to-back dimer, dis-tance 53 Å), where M_R11 is a mono-extendedDARPin containing two EpoR binding sitesseparated by five ankyrin repeats, NEEEPPPP-PEEEC. M_R11 delivers an extremely weakpSTAT5 signal (Fig. 5G), suggesting that theface-to-face orientation appears to be favorablefor the intracellular JAK2 to transphosphorylatewithin an EpoR dimer, thus more effectivelyinitiating signaling. These experiments showthat rotational reorientation of the receptorECDs can profoundly affect signaling output,which is consistent with previous studies inother systems using chimeric receptors (23)but now provides a ligand-mediated approach toexploit this parameter on natural, unmodifiedreceptors.

Effects of designed ligandson erythropoiesis

Erythropoiesis begins with lineage commit-ment of hematopoietic stem and progenitor cells(HSPCs) into the erythroid lineage, and cellsprogressively differentiate andmature to becomeRBCs (Fig. 6A). This process of RBC developmentcan be faithfully recapitulated in vitro by cultur-ing human primary HSPCs with various cyto-kines over 2 to 3 weeks (Fig. 6A) (36, 37). As amain driver of RBC development, EPO plays arole in all aspects of this process, from lineagecommitment to terminal maturation (Fig. 6A)(38, 39). Therefore, distinct stages of erythroidmaturationmay have differences in EPO signaling.Using DARPins that either increased the dis-

tance or altered interaction angles between EpoRdimers, we found a graded reduction in over-all erythroid differentiation and cellular expan-sion (Fig. 6B and fig. S18). Looking at overallerythroid differentiation by using flow cytom-etry allowed us to assess the kinetics of HSPCcommitment into the erythroid lineage andthe continuation of erythroid terminal matu-ration throughout the culture. In addition, pro-liferation is one of the important hallmarks inerythropoiesis that is coupled with the overall dif-ferentiation efficacy. Therefore, these analysestogether in a time-dependent manner can giveus further insights into stage-selective depend-encies of specific EpoR activities in erythropoiesis.This range of erythroid differentiation efficacy

stimulated by various DARPin series was as-sessed by using twomarkers: CD71, a marker ofearly erythroid commitment, and CD235a orglycophorin A, a marker of terminal erythroidmaturation (Fig. 6C). The majority of cells cul-tured with EPO, A_R3, A_angle_R4, C_R3, andA_dist_R5 differentiated into CD235a+ erythroidcells, whereas the later members of the series(wider angle or farther distance of the EpoRdimer) impaired or failed to effectively promoteerythroid commitment and differentiation (Fig.6C). Consistent with the pattern we observedwith flow cytometry for cell surface markers,treatment of cells with agonists that resultedin delayed or impaired erythroid differentiation




Dow

nloaded from


also resulted in reduced cellular proliferation(Fig. 6D). The cells cultured with EPO, A_R3, orA_dist_R5 showed at maximally potent concen-trations very similar erythroid differentiationefficacy on differentiation day 5 (early phase oferythropoiesis), with CD235a+CD71+ of 31.7 ±1.0%, 28.0 ± 1.6%, and 30.9 ± 0.6%, respectively(Fig. 6E). However, on differentiation day 12,cells cultured with A_dist_R5 showed lowerCD235a+CD71––positive cells compared withEPO and A_R3. Cells cultured with A_dist_R5

exhibited 1.7 ± 0.1% CD235a+CD71– cells, whereascellswith the recombinant EPOandA_R3 showed28.7 ± 2.0% and 26.8 ± 1.6% in this population,respectively, indicating the impairment of ter-minal erythropoiesis in the late stage withA_dist_R5 (Fig. 6E). Furthermore, cells culturedwith A_dist_R5 demonstrated reduced prolifer-ation, relative to EPO or A_R3, in the later stagesof the erythroid differentiation culture (Fig. 6D),demonstrating that A_dist_R5 specifically pro-motes early erythropoiesis with impairments at

the later stages. In addition, C_R3 promotedearly erythropoiesis poorly but showed increaseddifferentiation efficacy in the later stages (fig. S18),suggesting that this ligand selectively promotesthe later stages of erythropoiesis while having re-duced stimulation earlier. Collectively, these resultsdemonstrate how topological alterations in EpoRdimerization can selectively promote effectivesignaling at specific stages of human erythropoiesisbut fail to effectively enable differentiation andproliferation at other stages.


Fig. 4. Signaling responses induced through variation in EpoR dimerangle. (A) Addition of P repeats at the C terminus leads to increase in theangle between the EpoR ECDs, as shown. (B) Signaling bias observed forthe dose-response curves for the angle series. Shown are percentage ofpSTAT1/3/5 and pERK induced by the various DARPin ligands relative toEPO in UT7/EPO cells. Data are mean ± SD for two independent replicates.A sigmoidal dose-response analysis was preferred over Gaussian distribution because only some ligands demonstrate a decrease in MFI values at highconcentrations. (C) Bubble plot representation of the various downstream pathways activated by EPO or the DARPin ligands in UT7/EPO cells with15, 60, and 120 min of stimulation. The size of the bubble correlates to the signal strength determined as the log2 value of the ratio of the signalgenerated by the ligand to unstimulated cells. (D) The levels of phosphorylated downstream effectors were analyzed in the cell lysates by immunoblottingafter 15 and 60 min of ligand treatment. EpoR-immunoprecipitates were analyzed by immunoblotting for phosphorylation of specific Tyr residues inthe EpoR ICD. The dashed lines indicate the sites where the immunoblots were cropped for clarity. The uncropped immunoblots are included in fig. S14.



Dow

nloaded from



EpoRD1

EpoRD1

D2 D2

Fig. 5. Signaling responses induced by variationin EpoR ECD proximity. (A) Addition of P repeats atthe N terminus leads to increase in the distance betweenthe EpoR ECDs, as shown. (B) Signaling bias observedfor the dose-response curves for the distance series.Percentage of pSTAT5 and pERK induced by the variousDARPin ligands relative to EPO in UT7/EPO cells. Dataare mean ± SD for two independent replicates. A sigmoidaldose-response analysis was preferred over Gaussiandistribution because only some ligands demonstrate a decrease in MFI values at high concentrations. (C) Bubble plot representation of the variousdownstream pathways activated by EPO or the DARPin ligands in UT7/EPO cells with 15, 60, and 120 min of stimulation. The size of the bubble correlatesto the signal strength. (D) Immunoblot analysis of phosphorylated downstream effectors in cell lysates after 15 and 60 min of ligand treatment, andanalysis of EpoR-immunoprecipitates for phosphorylation of specific Tyr residues in the EpoR ICD. The dashed lines indicate the sites where theimmunoblots were cropped for clarity. The uncropped immunoblots are included in fig. S14. (E) The log2 value of the ratio of the signal generated by theligand to unstimulated cells at 60 min. (F) The model for M_R11 binding to EpoR. (G) Inversion of the relative geometry of the EpoR ECDs for themono-extended M_R11, which possesses a similar inter-ECD distance to A_dist_R6, leads to a complete loss of pSTAT5 activity.



Dow

nloaded from



Fig. 6. Effects on hematopoiesis of topologically controllable EpoRligands. (A) Overview of erythropoiesis and in vitro differentiation ofhuman primary hematopoietic cells. (B) Erythroid differentiation fromCD34+ HSPC with surrogate EPO ligands. Differentiation has beenassessed with flow cytometry with the markers CD71 and CD235a at day12 of differentiation. Means ± SEM for three independent experiments areshown. (C) Total CD235a+ cells and (D) overall cell proliferation duringerythroid differentiation. Data are means ± SEM for three independentexperiments (n = 2 independent experiments for A_dist_R5). (E) Flow

cytometry analysis of erythroid differentiation by EPO, A_R3, or A_dist_R5at different time points are shown with flow cytometry analysis. Means ±SEM for three independent experiments are shown. (F) Experimentalscheme of analyzing ligand-stimulated downstream signaling stimulatedby EPO, A_R3, A_dist_R5, and C_R3 during erythropoiesis. (G) Total MFIof pSTAT1/3/5 stimulated with EPO, A_R3, A_dist_R5, or C_R3 at early(day 5 of differentiation) and late (day 12 of differentiation) erythropoiesis.Data are (mean ± SEM) for three independent experiments [(B), (E),and (G)]. **P

These phenotypes led us to ask whether theremay be biased signaling responses downstreamof EpoR that correlate with such effects (Fig. 6F).Erythroid cells stimulatedwith A_dist_R5 showedhigher activation of STAT1/3/5 than with therecombinant EPO or A_R3 in the early stages(day 5). However, erythroid cells stimulated withA_dist_R5 showed higher activation of STAT1and STAT3, but lowered pSTAT5 was observedin the late stage of erythropoiesis (day 12) relativeto EPO or A_R3 in the late stage of erythropoiesis(Fig. 6G). Furthermore, cells stimulated withC_R3 showed low activation of STAT5 on dif-ferentiation day 5 (Fig. 6G). However, althoughthe level of activation of STAT5 was still lowerwith C_R3 relative to EPO or A_R3, the level ofSTAT5 activation between A_dist_R5 and C_R3was not significantly different (P = 0.40) ondifferentiation day 12, a time point when C_R3began to perform better than A_dist_R5 (Fig. 6,B and G). This data suggests that biased signal-ing is affected by changes of EpoR dimer geom-etry in primary human erythroid cells. However,we have only surveyed a limited range of down-stream effectors and thus may be missing specificsignals that mediate the observed phenotypiceffects. Our data suggests that this biased sig-naling is also dependent on the differentiationstage–specific context of the cells on which theyact. The same agonists can result in differentdownstream signaling depending on cell state.The ability to modulate signaling through asingle cytokine pathway at various stages ofhematopoietic differentiation has broad appli-cations in tuning desired responses for specificligands.

Discussion

Although tunable signaling is a foundationof GPCR pharmacology that can be exploitedthrough small-molecule medicinal chemistry,similar pharmacological principles have not beendeeply explored for dimeric receptors that engagecytokines and growth factors because of the dis-tinct structural challenges presented by the dif-ferent systems. For GPCRs, signal modulation isachieved through small-molecule ligands thatbind within pockets and induce different con-formations of the GPCR helices within the mem-brane, resulting in differential activation of Gproteins and arrestins (40). Structure-activityrelationships for GPCR agonists can be obtainedthrough medicinal chemistry. However, a com-parable approach does not exist to explore signaltuning by the large class of Type-I transmem-brane receptors, which require protein ligands tobind to large ECDs and induce large confor-mational changes that lead to dimerization.Although there have been hints that conforma-tional changes that affect dimer topology maybias signaling (6, 15, 18, 24), the functional rel-evance has not been well characterized and isnot well understood. We describe a generalstrategy to probe the effects of topologicalvariation of receptor dimerization on signalingand function for cytokines and, in principle, anyType-I class receptor that signals as a dimer. This

new strategy allows tunable manipulation ofgeometric parameters, but there currently isno predictive framework to relate these geometricchanges to signaling outcome. As such, the topo-logical variation this system affords throughprotein ligand engineering can be analogized toexploration of small-molecule pharmacology byusing medicinal chemistry.Although the strategy itself is mechanistically

agnostic, the signaling effects we observed inresponse to the topologically varied ligands arelikely due to two factors. First, the orientationaleffects of the ECD could propagate to the ICDsthat associate with the JAK2 kinases, affectingtheir relative proximities and orientations, whichalters the efficiency by which JAK2 acts on itssubsequent substrates on the EpoR ICD andSTAT1, -3, and -5 (Figs. 4D and 5D and fig. S14).In different receptor topologies, the JAK2/ICDcomplexes are likely repositioned into orienta-tions that are more or less conducive for trans-phosphorylation of the opposing JAK2 in thedimer, as well as tyrosine sites on the EpoR ICD.This could subsequently result in altered recruit-ment and/or phosphorylation of STATs andother adaptors that directly engage the JAK2and ICD phospho-tyrosines, and these mem-brane-proximal phosphorylation differenceswill subsequently propagate to gene expressionprograms in the nucleus (41, 42). In general,the more extreme angles and distances resultin less efficient downstream phosphorylationof all downstream adaptors to similar extents.However, uncoupling in the efficiency of JAK2and STAT phosphorylation for some of theligands—for example, A_R3, A_angle_R5, andA_dist_R5—results in the emergence of biasedSTAT activation.A second explanation of the observed effects

likely relates to mechanical distortion. As thetopologies are forced into more extreme ranges,strain is induced in the system, resulting in fewernumbers of signaling complexes formed, withsubsequent reduced receptor dimerization effi-ciency as observed with single-molecule assays,leading to lower Emax and right-shifted EC50sand disproportionately affecting some secondmessengers more than others. Supporting thisinterpretation, there is a systematic pattern ofreduced Emax that correlates with increasingdimer angle. For the distance series, there is nochange in Emax until a sharp threshold, at whichpoint it decreases abruptly. This could reflect a“rupture” distance at which point dimerizedJAK2 molecules can no longer transphospho-rylate efficiently. This rupture happens earlierwith the angle series than with the distanceseries because the contortions needed to recoverare greater. Possibly, both the angle and distanceseries interfere with weak, cooperative interac-tions between the transmembrane domains ofEpoR that have been reported (34, 35).As we have demonstrated for RBC biogenesis,

this approach is suitable for examining howspecific properties of cytokine signaling path-ways may have differentiation-stage specificityin hematopoiesis or other aspects of cell dif-

ferentiation. This approach could also be usedto survey cytokine receptor dimer geometriesin order to identify signaling outputs of clinicalinterest in other cytokine receptor system, suchas in the immune system. The scaffold we de-scribe here could be used as a preclinical phar-macological tool to determine whether differentdegrees of agonism are optimal for a givencytokine in a particular therapeutic indication.

Materials and methodsEpoR protein expression and purification

The DNA encoding for EpoR ECD was clonedinto pAcGP67-A, an insect cell protein expressionvector. The vector includes a C-terminal 3C pro-tease site, followed by a biotin-acceptor peptidetag (BAP tag, GLNDIFEAQKIEW), and a hexa-Histidine tag for affinity purification. The DNAencoding for the glycomutant of EpoRECD-N46Q,N157Q (25) was also cloned into pAcGP67-A witha 6-His tag for affinity purification. The proteinswere expressed using the baculovirus expressionsystem. The baculovirus stocks were prepared bycotransfection of the BaculoSapphire DNA (Orbi-gen) and the pAcGP67-A DNA into Spodopterafrugiperda (Sf9). Next, the viruses were used toinfect the Trichoplusia ni (High Five) cells. After72 hours, the proteins were harvested from thesupernatant and purified using nickel affinitychromatography (Nickel-NTA, Qiagen), followedby size-exclusion chromatography with a Super-dex S200 column (GE Healthcare). The proteinswere maintained in HEPES buffered saline (HBS,20mMHEPES pH 7.4, 150mM sodium chloride).EpoR ECD was site-specifically biotinylated atthe C-terminal BAP tag using BirA ligase andre-purified by size exclusion chromatography.Protein concentrations were determined by UVabsorbance at 280 nm using a Nanodrop2000spectrometer (Thermo Scientific), and then storedat –80°C.

DARPin expression and purification

The DNA encoding for the N2C and N3C DARPinproteinswas cloned into pET-28 vector, a bacterialexpression vector with a Kanamycin antibioticresistancemarker. The vector includes aC-terminalhexa-Histidine tag for affinity purification. Theplasmids containing the gene of interest wereused to transform Rosetta DE3 competent cells.The cells were grown at 37°C in 2YT media sup-plementedwith Kanamycin (40 mg/mL) until theculture reached log-phase growth. Next, IPTGwas added to the culture to induce protein ex-pression at a final concentration of 1 mM. Theculture was shaken at 37°C for 3 hours. Next, theproteins were harvested from the cells by sonica-tion, and purified using nickel affinity chromatog-raphy, followed by size-exclusion chromatographywith a Superdex S75 column (GE Healthcare). Theproteins were maintained in phosphate-bufferedsaline (PBS, 135 mM NaCl, 2.5 mM KCl, 8.0 mMNa2HPO4, 30 mM KH2 PO4, pH 7.2). Proteinconcentrations were determined by UV absorb-ance at 280 nm using a Nanodrop2000 spec-trometer (Thermo Scientific), and then storedat –80°C.




Dow

nloaded from


The constructs for DARPin dimers based onthe dimer C_R3 scaffold include an N-terminalhexa-Histidine tag for affinity purification. All ofthe extended DARPin proteins were maintainedin Tris-buffered saline (50mMTris pH 8, 150mMNaCl). The DARPin dimers A_dist_R5 throughA_dist_R10, all of the mono-extended M_RxDARPins, A_angle_R7, A_angle_R9, C_angle_R6,and C_angle_R8 were induced at 16°C for 16 to18 hours.

Cell culture

The human UT7/EPO cells (32) were culturedin DMEM (Dulbecco’s Modified Eagle Medium)supplemented with 10% v/v fetal bovine serum,penicillin-streptomycin, 1 mML-glutamine,HEPESbuffer, and 2 U/mL recombinant EPO (Amgen)(20). The cells were maintained at 37°C with5% CO2. The cultures were passaged every 2to 3 days.

Yeast display of DARPins

DARPins were displayed on the surface of yeastS. cerevisiae strain EBY100 by fusion to theC terminus of the Aga2 protein. The DARPininsert, along with an N-terminal 3C proteasecleavage site and a C-terminal cMyc epitopewere cloned into the PCT302 vector. Briefly,competent yeast cells were electroporated withplasmids and recovered in SDCAA selectionmedia at 30°C (43). Next, grown cultures werepassaged in SDCAA media before inducing ex-pression in SGCAA media. The cultures wereinduced at 30°C for 24 hours, followed by 20°C.The display of full-length DARPins on the sur-face of yeast was confirmed by staining the cellswith an Alexa Fluor 488-labeled anti-cMyc anti-body (1:100 dilution; Cell Signaling) and moni-toring the fluorescence by flow cytometry.

Yeast displayed DARPin libraries

Two site-directed DARPin libraries (N2C andN3C) were created by assembly PCR using thefollowing degenerate codons. The gel-purifiedassembled product was combined with linearizedPCT302 vector and electroporated into EBY100yeast cells (43). The electroporated yeast wererecovered with SDCAA selection media and in-duced with SGCAA media, as described above.N2C Library:TCCGACCTGGGTAAGAAACTGCTGGAAG-

CAGCTCGTGCTGGTCAGGACGACGAAGTTCGT-ATCCTGATGGCTAATGGAGCAGATGTCAACGC-GASWGACNWTDBGGGTDYGACTCCGCTG-CACCTGGCTGCTMDSRHMGGTCACCTGGA-AATCGTTGAAGTTCTACTGAAGAACGGTGCTG-ACGTTAACGCTASWGATNWTWMCGGA-DYGACACCACTTCATCTAGCAGCGAVW-MDSGGACATCTCGAGATTGTCGAGGTCT-TACTCAAACACGGCGCAGACGTAAATGCA-CAGGACAAATTCGGTAAGACCGCTTTCGATA-TCTCCATCGACAACGGTAACGAGGACCTGGC-TGAAATCCTGCAATheoretical diversity: 2.1 × 1010

Diversity of library generated through elec-troporation: 1.8 × 109

N3C Library:

TCCGACCTGGGTAAGAAACTGCTGGAAG-CAGCTCGTGCTGGTCAGGACGACGAAGTTC-GTATCCTGATGGCTAACGGAGCAGATGTCA-ACGCGASWGACSWCWMCGGTKYCACA-CCTCTGCACTTAGCTGCAMWARHCGGTC-ATTTGGAAATCGTCGAAGTTCTACTGAAG-CACGGTGCTGACGTTAATGCTASWGAT-MWSWMCGGADYGACTCCGTTACATC-TAGCGGCTANTADSGGACATCTCGAGA-TTGTAGAGGTGTTACTCAAATACGGCGC-AGATGTAAACGCATMCGATSWCWMC-GGCRYGACGCCATTGCACCTCGCAGCT-AVWAWSGGCCACTTAGAGATAGTTGAAGT-GCTCTTAAAGTATGGAGCCGACGTGAATGC-CCAGGACAAATTCGGTAAGACCGCTTTCGA-CATCTCCATCGACAACGGTAACGAGGACCTG-GCTGAAATCCTGCAATheoretical diversity: 5.2 × 1011

Diversity of library generated through elec-troporation: 1.0 × 109

Evolution of EpoR binding DARPins

The two yeast-displayed DARPin libraries (N2Cand N3C) were panned against the EpoR ECD toselect for high affinity clones using a combina-tion of magnetic-activated cell sorting (MACS)and fluorescence-activated cell sorting (FACS).Biotinylated EpoR was used for all the selectionrounds unless otherwise noted. The first roundof selection was performed with 5.4 × 109 N2Cand 5 × 109 N3C yeast cells to ensure coverageof library diversity. First, EpoR was mixed with500 ml of magnetic streptavidin microbeads(Militenyi) to a final concentration of 400 nM.The mixture was then incubated with DARPin-displaying yeast cells at a final volume of 20 mLin PBE (PBS, 0.5% bovine serum albumin, 2 mMEDTA). The cells were incubated with shakingfor 2 hours at 4°C. Next, the yeast cells werewashed with PBE and flowed over MACS LSseparation columns (Miltenyi) to isolate theEpoR binders. The enriched library pool wasregrown in SDCAA media and then induced inSGCAA. In subsequent rounds, the total numberof yeast cells used was ten times the number ofyeast cells eluted in the previous round or 1 ×108 cells, whichever is larger.For round 2, the on-bead selection with high

avidity effects was repeated to select for moreEpoR binders without increasing the stringency.EpoR was mixed with 250 ml of magnetic strep-tavidinmicrobeads (Miltenyi) to a final concentra-tion of 400 nM and incubated with yeast-displayedDARPin cells in 10mL final volume. In round 3, thelibrary pool (1 × 108 yeast cells) was incubatedwith Alexa Fluor 647-labeled anti-cMyc antibody,to enrich for transformants displaying full lengthDARPin clones. The mixture was incubated withshaking for 2 hours at 4°C, followed by PBEwashes. The yeast cells were then incubatedwith anti-Alexa Fluor 647 microbeads (Miltenyi)and flowed overMACS LS separation columns toisolate the full-length clones.In subsequent rounds, FACS method was em-

ployed to selectively enrich for EpoR bindingclones. For round 4, the propagated and inducedyeast cells from round 3 were incubated with

EpoR at a final concentration of 1 mMmonomer,in 1 mLPBE, with shaking for 2 hours at 4°C. Theyeast cells were washed with PBE and then in-cubated with Alexa Fluor 488-labeled anti-cMycantibody (Cell Signaling) and streptavidin-conjugated Alexa Fluor 647. The cells were thensorted using FACS to select for clones that ex-hibited high levels of EpoR binding (MFI of AlexaFluor 647) relative to DARPin display levels (MFIof Alexa Fluor 488) on yeast.Similarly, for the N2C library, the subsequent

rounds of selections were performed at 333 nM,40 nM and 4 nMEpoR concentrations for rounds5, 6, and 7, respectively. Whereas, for the N3Clibrary, the subsequent rounds of selectionswere performed at 333 nM, 10 nM and 4 nMEpoR concentrations for rounds 5, 6, and 7,respectively.

Second-generation library: Evolution ofhigh-affinity DARPins

The two second generation DARPin libraries(N2C and N3C) containing DNA-shuffled cloneswere panned against biotinylated-EpoR ECD toselect for high affinity clones, as described above.Diversities of the second-generation librariesgenerated through electroporation:N2C library: 3.2 × 109

N3C library: 4.6 × 109

The first round of selections was performedwith 1 × 1010 yeast, to ensure coverage of librarydiversity. For round 1, EpoR was mixed with1000 ml of magnetic streptavidin microbeads(Militenyi) to a final concentration of 400 nM,and EpoR binding clones were isolated usingMACS LS columns. For round 2, the library poolswere incubated with 50 nM and 10 nM EpoR-tetramers for N2C andN3C libraries, respectively.The EpoR tetramers were generated by mixingbiotinylated EpoR and streptavidin-conjugatedAlexa Fluor 647, at amolar ratio of 4.5:1. Next, theyeast cells were incubatedwith EpoR-tetramers for2 hours at 4°C. The yeast cells were then washedwith PBE and incubated with anti-Alexa Fluor 647microbeads. The EpoR binders were isolated usingMACS LS separation columns.To further enrich the library pool, FACS

method was employed in subsequent rounds.For both the libraries, the subsequent roundsof selections were performed at 10 nM, 1 nM,and 0.1 nM EpoR concentrations for rounds 3,4, and 5, respectively. For round 6, the librarypool was subjected to a kinetic sort; a methodto select for clones with slower off-rates. First, thelibrary pools were incubated with 5 and 2.5 nMEpoR for N2C and N3C libraries, respectively.The yeast cells werewashed and thendual-stainedwith Alexa Fluor 488-labeled anti-cMyc andstreptavidin-conjugated Alexa Fluor 647. Next, a100-fold excess of unbiotinylated EpoR, 500 nMfor N2C and 250 nM for N3C in 1 mL was addedto the cells. The mixture was shaken at roomtemperature, and the levels of bound proteinwere measured at regular intervals using flowcytometry. The levels did not change signifi-cantly over 21 hours. The cells were washed,restained, and then subjected to FACS.




Dow

nloaded from


Analysis of individual clonesTo isolate individual clones, yeast cultures fromthe last selection rounds were plated on SDCAAagarose plates. Individual colonies were trans-ferred to SDCAAmedia in culture tubes, passagedand then induced. Next, 93 clones were screenedagainst a fixed EpoR concentration. For eachclone, ~5 × 105 cells were transferred to a V-bottom plate and incubated overnight with bio-tinylated EpoR at 4°C. The cells were washedand stained with streptavidin-conjugated AlexaFluor 647. Next, the clones were analyzed forbinding using flow cytometry. The highest af-finity clones were identified, and their plasmidswere isolated using the Zymoprep Yeast PlasmidMiniprep II Kit (Zymo Research), and sequenced.Yeast-displayedDARPin ligandswere also titratedagainst EpoR at different concentrations to obtainEC50 values. As described above, ~5 × 10

5 cellswere transferred to a V-bottom plate and in-cubated with 100 mL of serially diluted EpoRsolution. Next, the cells were washed, stained,and then analyzed for EpoR binding by flowcytometry.

Surface plasmon resonance

Dissociation constants (Kd) for EpoR-bindingDARPin ligands were determined by surfaceplasmon resonance (SPR) using the BIAcoreT100 instrument (GE Healthcare). First, bio-tinylated EpoR was captured on a streptavidin-coated (SA) sensor chip (GE Healthcare) withimmobilization density in the range of 100 reso-nance units (RU). Similarly, a control flow cellwas also preparedwith an off-target protein suchas biotinylated-IFNAR1 or biotinylated-Fzd8 forreference subtraction. The binding kinetics wereperformed at 25°C with a flow rate of 50 ml/min.EpoR binding DARPin ligands were serially di-luted in HBS buffer supplemented with 0.005%P20 surfactant (GEHealthcare) and injected overthe SA chip. Dissociation kinetics weremonitoredfor 200-500 s, as needed, based on the ligandaffinity. For high affinity ligands, the chip wasregenerated with 4 M MgCl2 after each ligandinjection. The binding kinetics and dissociationconstants were determined using the BIAcoreT100 evaluation software. The sensograms ob-tained were either fit to the 1:1 kinetic bindingmodel or to the steady-state affinity model. Thekinetic binding curves for each ligand were gen-erated by plotting the time-dependent responseunits in Prism 7 (GraphPad).

Error-prone PCR

The EpoR binding individual clones obtainedfrom the first set of selections were subjectedto error prone PCR to introduce random muta-tions using the GeneMorph II Random Muta-genesis kit (Agilent). Various amounts of DNAtemplate were tested to quantify the rate ofmutagenesis. The GeneMorph kit was used ata concentration of 0.1 ng DNA/reaction, toachieve an average mutation rate of 2.5 to 3mutations/gene. The PCR product was furtheramplified and purified by gel electrophoresis forDNA shuffling.

DNA shufflingThe EpoR binding clones with random muta-tions inserted through error-prone PCR weresubjected to DNA shuffling (44) to enhance bind-ing affinity. The amplified and purified DNAproduct from various clones was combined, andthen used as the template for DNA shuffling.Briefly, 2 mg of combined DNAwas subjected toDNase treatment (DNase I, Invitrogen) by in-cubating at 15°C for threemins, followed by 80°Cfor 10min to denature the enzyme. The DNA frag-ments were subjected to PCR clean-up (Qiagen)to remove extremely small fragments. Next, thefragments were allowed to recombine at regionsof moderate homology. The PCR was performedusing Taq (New England Biolabs) and PfuUltra(Agilent) polymerases with progressive hybrid-ization. The reaction product was further ampli-fied using PfuUltra polymerase. The product waspurified using gel electrophoresis and the bandscorresponding to N2C and N3C sizes were spe-cifically extracted and purified. The gel-purifiedproduct was combined with linearized PCT302vector and electroporated into EBY100 yeast cellsto create the second-generation libraries. Diver-sity of library generated through electroporation:N2C library: 3.2 × 109

N3C library: 4.6 × 109

Engineering the extended DARPins

The modular nature of the DARPin was utilizedtoward the generation of the extended DARPins.The DARPin molecules are non-planar, and eachrepeat provides a 2-3° counter-clockwise rotationrelative to the preceding repeat (33). Thus, it iscrucial to maintain the overall curvature of theDARPin to avoid a loss in binding affinity whileinserting more ankyrin repeats. The sequence ofthe P repeat was carefully designed to maintainthe curvature observed in the N3C E2 monomer,and the A_R3 and C_R3 dimers. The frameworkresidues in the P repeat were based on the se-quence of the consensus repeat. The positionsthat were randomized in the library design (de-noted as X) (fig. S1) were assigned residues froma Frizzled8-binding DARPin. As a negative con-trol, a DARPin dimer comprising of only non-binding P repeats was designed (NPPPPC) toeliminate the possibility of off-target (Frizzled8)effects. The control dimer shows no affinity to-ward EpoR or Frizzled8 (fig. S6).As described above, the E2 DARPin is com-

prised of three EpoR binding repeats, denoted as“EEE,” and the N- and C-terminal capping re-gions. To engineer the “angle” series, one ormore P repeats was inserted at the C terminusto provide DARPins such as NEEEP..C. Further-more, the dimerizing repeats (indicated by un-derline) are placed on the same terminus as theP repeat insertions for maximum impact. Indimer A_R3/C_R3 (NEEEC), the dimerizing res-idues span over the three ankyrin repeats andthe two capping regions. To generate A_angle_R4,one P repeat was inserted at the C-terminal end(NEEEPC) and the set of dimerizing residues werealso shifted by one repeat. Similarly, the design forA_angle_R5 can summarized as NEEEPPC.

P repeat sequence: DAAGMTPLHLAAANGH-LEIVEVLLKYGADVNAKTo engineer the “distance” series, the non-

binding P repeats were inserted at theN terminus,for example NPP…EEEC. Insertion of repeats atthe N terminus proved challenging, as illustratedin fig. S9. A straightforward design based on theangle series such as NPEEEC displayed a sig-nificant loss in binding affinity toward EpoR.Three possible reasons were identified. First,an analysis of the models revealed a clash be-tween residues of the inner helices from the firstbinding repeat and the preceding P repeat (fig.S7A). No such clashes were observed in the E2DARPin where the N-terminal capping regionhelix precedes the first binding repeat (fig. S7B).Second, an analysis of the sequence of the E2N-terminal capping region revealed a Glu to Lysmutation in the inner helix, which resulted fromthe error-pronemutagenesis performed duringaffinity maturation. The Lys residue is well posi-tioned to interact with EpoR. Third, insertion ofthe P repeat implies existence of b-turn loopspreceding the loop region from the first bindingrepeat. In E2, A_R3 (NEEEC) and the angleseries, no b-turn loops precede the loop regionfor the first ankyrin repeat. The newly insertedloops could introduce destabilizing interactions.To address these issues, a modified repeat (P′)was used as the repeat preceding the first Erepeat such as NP..P′EEEC. In the P′ repeat, thesequence of the inner helix was modified fromHLAAANG to HKAARAG. The “KAARA” se-quence is derived from the inner helix of theN-terminal capping region and installed in the P′repeat based on structural symmetry of the twohelices. This eliminates the clashes between thehelices and also re-positions the Lys residue tointeract with EpoR. Furthermore, the loop regionwith the sequence DAAGMwas shortened to Gly,to facilitate the turn while eliminating any de-stabilizing interactions, (fig. S7C). Increase inthe loop length for DARPins has been previouslypublished (45). Here, we show that the loops canalso be shortened without affecting the structurefor A_dist_R7 (fig. S7D). The dimerizing residuesare now maintained at the N-terminal repeats.To generate A_dist_R4, P′ repeat and the dim-erizing residues were inserted at the N-terminalend (NP′EEEC).Whereas, to generate A_dist_R5,one P followed by one P′ repeat was inserted,NPP′EEEC. Similarly, the design for A_dist_R6can summarized as NPPP′EEEC. Also, the se-quence of the P repeat used for generating thedistance series is different from the P repeat usedfor the angle series (C-terminal insertions). Butboth the sequences are referred to as “P repeats”for the sake of simplicity.P repeat sequence: DAAGGTPLHEAARAGH-

LEIVEVLLKYGADVNAVP′ repeat sequence: DGTPLHKAARAGHLEI-

VEVLLKYGADVNAVTo engineer the “mono-extended” DARPin

series, two sets of EpoR binding repeats wereincluded on the same DARPin molecule. Thisseries is based on monomeric extended DARPins,and thus named “mono-extended.” To generate




Dow

nloaded from


a series, the two sets of binding repeats werepositioned at either termini, and the P repeatsare inserted in the middle, for example NEEEPP…EEEC. The second set of EpoR binding repeatsare positioned similar to the distance series.Thus, the P’ approach was adopted for the gen-eration of the mono-extended series, as well. Theinsertion of one to four repeats in the middleshowed severe clashes in the models gene-rated. As a result, the first member of the seriesconsisted of five non-binding repeats, M_R11,NEEEPPPPP’EEEC. Similarly, the design forM_R12 can be summarized asNEEEPPPPPP’EEECwith a total of six non-binding repeats inserted.P repeat sequence: DAAGGTPLHEAARAGH-

LEIVEVLLKYGADVNAVP’ repeat sequence: DGTPLHKAARAGHLE-

IVEVLLKYGADVNAV

Phospho-flow signaling assays

UT7/EPO cells were starved overnight for~18 hours in base media without EPO. The cellswere stimulated with EPO or the DARPin lig-ands for 15 min at 37°C, followed by fixationwith paraformaldehyde (Electron MicroscopySciences) for 10 min at room temperature. Next,the cells were permeabilized for intracellularstaining by treatment with methanol (Fisher) for30 min at -20°C. The cells were then incubatedwith the desired antibodies at a 1:50 dilution for1 hour at room temperature. The levels of thevarious phosphorylated proteins correspond tothe measured values of mean fluorescence in-tensity (MFI). The background fluorescence ofthe unstimulated samples was subtracted fromall the readouts.Datawas acquiredusingCytoFlex,flow cytometer instrument (Beckman Coulter).The MFI values were normalized to the MFIof EPO at 37.04 nM and plotted in Prism 7(GraphPad). The dose-response curves were gen-erated using the “sigmoidal dose-response” analy-sis. A Gaussian distribution analysis was excludedas only some ligands demonstrate a decrease inMFI values at high concentrations. All the anti-bodies usedwere purchased fromBDBiosciences:Alexa Fluor 488 Mouse Anti-Stat1 (pY701), AlexaFluor 647 Mouse Anti-Stat3 (pY705), Alexa Fluor488 Mouse Anti-Stat3 (pY705), Alexa Fluor 647Mouse Anti-Stat5 (pY694), and Alexa Fluor 488Mouse Anti-ERK1/2 (pT202/pY204).

Phosphorylated protein analysis

UT7/EPO cells were starved overnight and stim-ulated with saturating concentrations of EPOand DARPin ligands for 15, 60, and 120 min at37°C, followed by fixation with paraformaldehydefor 10 min at room temperature. Next, the cellswere washed and resuspended in 0.5% bovineserum albumin in PBS, and stored at –80°C. Thecells were then prepared for staining accordingto standard protocols. The cells were stainedwitha panel of antibodies (Primity Bio PathwayPheno-typing service) and analyzed on LSRII Flowcytometer (Becton Dickinson). The levels of phos-phorylated protein were calculated as the Log(base 2) ratio of theMFI of stimulated samples tounstimulated samples.

Cell lysis and immunoprecipitationUT7/EPO cells were starved overnight and stim-ulated with saturating concentrations of EPO(111 nM) and DARPin ligands (1 mM) for 5 and15 min at 37°C. Next, the cells were washedwith ice-cold PBS and resuspended in 100 ml/5E6 cells lysis buffer and incubated with rotationfor 30 min at 4°C. The Pierce IP lysis buffer(Thermo Scientific) was supplemented withphosphatase inhibitor cocktail (Abcam, 1/100 v/v),Halt phosphatase inhibitor cocktail (ThermoScientific, 1/100 v/v), Sodium orthovanadate(NEB, 0.5 mM final concentration), and EDTA-free protease inhibitor cocktail (Roche, 1 tablet in10 mL of buffer). The cell lysates were cleared bycentrifugation at 10 krpm for 15 min at 4°C. Thesamples were denatured by the addition of 30 mlsample buffer (0.28 M Tris, pH 6.8; 30% v/vglycerol; 1%w/v SDS; 0.55M b-mercaptoethanol;0.0024% w/v bromophenol blue) and boiling for5min, resolved by 12.5% SDS-PAGE, and analyzedby immunoblotting.For anti-EpoR immunoprecipitations, anti-

EpoR antibody VP2E8 (Abcam, 4.6 mg/18E6 cells)was added to the clarified cell lysates and in-cubated with rotation for 1 hour at 4°C. Next,30 ml of Pierce protein Gmagnetic beads (ThermoScientific) were added, and incubated with rota-tion for 1 hour at 4°C. The beads were thenwashed three time with the supplemented lysisbuffer. Next, the proteins were eluted by theaddition of 160 ml sample buffer and boiling for5 min. The beads were removed by magnetic sep-aration. The samples were resolved by 12.5%SDS-PAGE and analyzed by immunoblotting.The antibodies used are as follows: PurifiedMouse Anti-Stat1 (pY701, BD Biosciences), Rabbitanti-STAT1 42H3 (Cell Signaling Technology),PurifiedMouseAnti-Stat5 (pY694,BDBiosciences),Rabbit anti-STAT5 3H7 (Cell Signaling Tech-nology), anti-EpoR (VP2E8, Abcam), EpoR (D-5)HRP (sc365662, Santa Cruz Biotechnology), anti-JAK2 (EPR108[2], Abcam), anti-phospho(Y1007 +Y1008) JAK2 (E132, Abcam), phosphor-EpoRY426(PA5-40305, Thermo Fisher Scientific), phospho-EpoR Y368 (PA5-38483, Thermo Fisher Scientific),phospho-EpoR Y456 (PA5-64795, Thermo FisherScientific), swine anti-rabbit immunoglobulins/HRP (P0399, Dako), and rabbit anti-mouse im-munoglobulins/HRP (P0161, Dako).

Receptor internalization

UT7/EPO cells were starved overnight and thentreated with 15 mg/mL cycloheximide (Sigma) for4 hours. Next, the cells were stimulated withEPO (111 nM) or the DARPin ligands (1 mM) for15, 60, or 120 min at 37°C. The ligand solutionswere also prepared in media containing cyclo-heximide. The cells were then washed with ice-cold PBS and incubatedwith anti-EpoR antibodyfor 1 hour at room temperature. Next, the cellswere incubated with Alexa Fluor 647 goat anti-mouse antibody at a 1:100 dilution for 1 hour atroom temperature. The levels of cell-surfaceEpoR correspond to the measured values ofmean fluorescence intensity (MFI). The MFIvalues were normalized to the MFI of untreated

cells. The antibodies used are as follows: anti-EpoR (VP2E8, Abcam), Alexa Fluor 647 Goatanti-mouse (405322, Biolegend), and Alexa Fluor647 Goat anti-mouse (A21235, Invitrogen).

Crystallization, data collection,and refinement

Protein complexes were formed by mixing 1:2molar ratios of DARPin and EpoR ECD (or 1:1DARPin/EpoR in the case ofmonomeric E2) alongwith Carboxypeptidases A and B at 1:100 (w/w),and phenylmethane sulfonyl fluoride at 0.5 mMfinal concentration. After overnight incubationat 4°C, complexes were purified by FPLC on anS200 column, eluted in HBS buffer, and con-centrated. Crystallizationwas performed by sittingdrop vapor diffusion using aMosquito nanoliterpipetting robot. Monomeric E2-EpoR complex crys-tals were grown at 9mg/mL from 1MNaH2PO4/K2HPO4 pH 5.6. Crystals were cryoprotected byaddition of ethylene glycol to 25% before har-vesting for data collection at APS beamline 23ID-B. C_R3 crystals were grown at 17 mg/mLfrom 0.1 M Tris pH 8.5, 25% PEG 3350, 1 mMreduced glutathione, and 1 mM oxidized gluta-thione. Crystals were cryoprotected by the addi-tion of ethylene glycol to 25% before harvestingfor data collection at SSRL beamline 12-2.The C_R3/EpoR complex crystals were grown

at 6 mg/mL from 0.1 M HEPES pH 7.3 and 1.4 MNaKHPO4. Crystals were cryoprotected by theaddition of glycerol to 24% before harvestingfor data collection at ALS beamline 8.2.2. TheA_angle_R5/EpoR complex was crystallized at4.5 mg/mL from 0.2 M K/Na tartrate, 0.1 M BisTris propane pH 8.5, and 20% PEG 3350 andcryoprotected by the addition of glycerol to 30%before harvesting for data collection at ALSbeamline 5.0.1. Crystals of the C_angle_R5/EpoRcomplex were grown at 6 mg/mL from 0.3 MMgSO4, 0.1 M Bis-Tris Propane pH 7.0, and 22.5%PurePEGS cocktail (Anatrace) and cryoprotectedby the addition of ethylene glycol to 30% beforeharvesting for data collection at ALS beamline5.0.1. The A_dist_R7/EpoR complex was crystal-lized at 8mg/mL from 0.2MNaCl, 0.1 M Tris pH8.5, and 25% PEG 3350 and harvested withoutfurther cryoprotection for data collection at ALSbeamline 8.2.1. Initial crystals of the M_R12/EpoR complex were grown at 10 mg/mL from0.2 M NaCl, 0.1 M CAPS pH 10.5, and 20%PEG 8000. These crystals were pulverized toform a seed stock for subsequent seeding into0.2 M ammonium tartrate and 20% PEG 3350.The resulting crystalsweredehydrated for 1monthby increasing the reservoir concentration to25%PEG 3350. The crystals were cryoprotectedby addition of 10% glycerol and 10% ethyleneglycol before harvesting for data collection atAPS beamline 23ID-B. All datasets were integ-rated and scaled using XDS (46) before mer-ging symmetry-related reflections with aimless(47, 48). Due to anisotropy of the C_R3/EpoR,A_angle_R5/EpoR, C_angle_R5/EpoR, andM_R12/EpoR datasets, final ellipsoidal truncation wasperformed using STARaniso before mergingreflections (49).




Dow

nloaded from


Structures were solved by molecular replace-ment with Phaser (50) using the crystal structureof EpoR [Protein Data Bank (PDB) ID 1ERN],split into D1 and D2 domains, and the predictedDARPin structures as search models, or in thecase of the C_R3/EpoR complex using the re-fined C_R3 crystal structure.With the exceptionsnoted below, crystallographic models were pre-pared by iterative rounds of manual rebuildingand reciprocal space refinement with COOT(51) and Phenix (52–54). For the C_R3/EpoR,C_angle_R5/EpoR, and M_R12/EpoR complexes,additional refinement with Buster (55, 56) wasperformed to aid convergence of the structurebefore final refinement in Phenix. TLS refine-ment was performed for all structures exceptfor the C_R3 dimer, for which anisotropic atomicdisplacement parameters (ADPs) were refined,and A_dist_R7/EpoR. Individual ADPs were re-fined for all structures except that grouped ADPswere refined for M_R12/EpoR and ADPs werenot refined for A_dist_R7/EpoR. For the low-resolution A_dist_R7/EpoR structure, aftermolec-ular replacement the structure was manuallyedited in COOT to correct the amino acid se-quence and to delete loop 4 of A_dist_R7, whichwas present in the search model but absent inthe purified protein. Model refinement was thencompleted by rigid body refinement in Phenixand jelly body refinement in REFMAC (57–59).Software used in this project was installed andconfigured by SBGrid (60). Crystallographic datacollection and refinement statistics along withPDB deposition codes are reported in tables S1and S2.

Generation of models

The synthetic EpoR agonists described in thiswork have a unique property that differentiatethem from previous efforts at generating syn-thetic type I cytokine agonists, namely they havefairly well-defined and rigid conformations thatprovide a structural basis for each observedphysiological output. This advantage was lever-aged by generating a crystallographic structure-informed set of signaling complexmodels for eachof the A_dist_Rx, A_angle_Rx, and C_angle_Rxseries. To best recapitulate both the backbonetopology of the P repeat and the N-/C-terminalmode of EpoR binding, the backbone coordinatesfrom theM_R12/EpoR crystal structure were usedto construct each experimentally tested memberof the extension series using the Rosetta macro-molecular modeling suite. While most presentednon-clashing complex models upon superposi-tion of the E2-EpoR structure with the extendedbackbone, several cases showed significant clashesbetween EpoR subunits that could not be physi-cally feasible.We used a constrainedminimizationprotocol on each of the clashing complex modelsto generate a complete set of plausible models.

Backbone and symmetric complex generation

Backbone coordinates for each DARPin extensionlength were derived from the M_R12/EpoR struc-ture while keeping capping modules intact. ForA/C_angle, the N-terminal EpoR binding mode

was preserved from M_R12. For A_dist, theC-terminal EpoR binding mode was preservedwhich included the P′ shortened loop modifica-tion which was impactful for binding. EachDARPin/EpoR complex was aligned to chainA of dimers A (pdb id 5HRY) andC (pdb id 5KBA).The synthetic signaling complex was generatedin the Rosetta macromolecular modeling suiteusing a symmetry definition file to generate theC2 symmetry starting from each alignedDARPin/EpoR complex. The experimentally tested se-quences were threaded to each backbone usingthe Rosetta packing algorithm with the defaulttalaris2014 score function followed by side-chainenergy minimization.

Relax of EpoR clashing models

Models that presented clashes between themembrane-proximal regions of the EpoR ECDwere identified by visual inspection. An analysisof the EpoR conformation in all the crystal struc-tures obtained singled residues 113-120 as a hingeregion which allows for conformational flexibilitybetween the binding domain and the membraneproximal domain. A Rosetta relax trajectory insearch of a state with an energy minimum al-lowing backbone torsion angles to be modeledflexibly in this region while keeping the rest ofthe construct fixed using coordinate constraintswas carried out for each clashing model. In thecase of dimer A_R3, the first attempt with Rosettarelax to relieve the severe clashes between EpoRsubunits was unsuccessful without having tomake significant moves to the head region orrearranging the designed dimer interface orienta-tion, so the EpoR conformation fromA_angle_R5was used instead for the relax trajectory. Therelaxed output EpoR conformation from dimerA_R3 was used to rescue a non-clashing modelfor the A_angle_R4 complex. Although the gen-erated non-clashing models do not representunique solutions, they demonstrate the stericplausibility of a given complex.

Structural analysis of complex models

Although the synthetic agonists themselves arerather rigid, the EpoR ECD has some intrinsicflexibility between theD1 andD2 domains givingrise to uncertainty inmeasurements derived frommodels. Indeed, multiple EpoR conformers wereobserved in the crystal structures obtained for thiswork (fig. S10). Because the DARPin-EpoR D1 in-teraction is rigid, we made all distance and anglemeasurements using Val119 of EpoR, at the hingebetween D1 and D2. Distance measurements be-tween EpoR ECDs were made between the Casof the residue Val119. Scissor half angle measure-ments were taken using Val119 of EpoR as thevertex with rays extending through Asp2 at theN-terminal cap of the DARPins and throughan alanine at the start of the C-terminal cap ofthe DARPin (due to varying insertion lengthsin DARPins with X repeats, this was residue num-ber 39 + 33X). Although the measurementspresented (Figs. 3 and 4) represent the generaltrend of complex geometries within the ex-tension series, these measurements do not di-

rectly correspond to the relationship betweenthe transmembrane or intracellular portions ofEpoR in the cellular context.

Single molecule fluorescence imagingSample preparation

HeLa cells were transiently transfected by cal-cium chloride precipitation with 10 mg of pSemsneo leadermXFPEpoR. In principle, a fluorescent-dead Green Fluorescent Protein (meGFP-Y66F/mXFP) was N-terminally fused to EpoR. Aftertransfection, cells were seeded on a PLL-PEG-RGDcoated glass cover slide (61). Imaging experimentswere typically carried out 24h post transfection.Labeling, ligand incubation and subsequent imag-ing were performed in a custom-made incubationchamber with a volume of 500 ml, which wasmounted on the microscope. The experimentspresented here were performed at room temper-ature in the presence of media without phenolred. Additionally, each sample was supplementedwith an oxygen scavenger and a redox-activephoto protectant to minimize photo bleaching(62). For cell surface labeling of mXFP-taggedEpoR DY647- and Rho11-labeled Nanobodies(NBs) were added in equal concentrations (2 nMeach) and incubated for at least 5 min. High equi-librium binding of the labeled NBs was ensuredby keeping these in bulk solution. Co-locomotion ofEpoR was probed before and after incubating lig-ands at indicated concentrations for at least 10min.

Single molecule microscopy

Single molecule imaging was performed by TIRFmicroscopy. As described previously (20, 24) aninverted microscope (Olympus IX71) equippedwith a triple-line total internal reflection (TIR)condenser (Olympus) and a back-illuminatedelectron multiplied (EM) CCD camera (iXonDU897D, Andor Technology) was used. Thesample was TIR illuminated via a 150x magnifi-cation objectivewith a numerical aperture of 1.45(UAPO 150 x /1.45 TIRFM Olympus). Rho11 wasexcited by a 561 nm laser (CrystaLaser) at 0.95mW(~32 W/cm2) and DY647 by a 642 nm laser(Omicron) at 0.65 mW (~22W/cm2). For fluores-cent detection a spectral image splitter (Dual-View, Optical Insight) with a 640 DCXR dichroicbeam splitter (Chroma) combined with a 580/40(Semrock, Rho11 detection) and a 690/70 (Chroma,DY647 detection) bandpass filters has been useddividing each emission channel into 512 × 256pixel. For each cell image stacks of 150 frameswere recorded with temporal resolution of ap-proximately 32 ms/frame.

Single molecule analyses

The workflow of single molecule analyses isoutlined in fig. S6A. Single molecule localizationwas assessed by using multiple-target tracing(MTT) algorithm (63) and tracking was per-formed, as described previously (64). Becauseimmobile molecules (typically ~10% of the entiredataset) weremostly caused by either non-specificbinding of labeled NBs to the coverslip or byendosomal uptake of receptors, these were iden-tified by spatiotemporal cluster analysis (65) and




Dow

nloaded from


removed from the dataset. Dual-color co-trackinganalysis was carried out as described recently(20). Prior to each experiment, Rho11 and DY647channels were aligned with sub-pixel precisionbased on a calibration with multicolor fluores-cent beads (TetraSpeck microspheres 0.1 mm,Invitrogen). Individualmolecules detected in bothspectral channels within the same frame andwithin a distance threshold of 100 nm were con-sidered as co-localized. In order to reconstructco-locomotion trajectories (co-trajectories) fromthe identified population of co-localizing par-ticles, the tracking algorithm was applied tothe co-localization dataset. Co-trajectories with aminimum length of 10 consecutive steps (320 ms)were considered as receptor dimers. The relativefraction of co-locomotion was calculated with theabsolute number of trajectories from both chan-nels and was corrected for stochastically doublelabeling with the same fluorophore within adimer. The relative fraction of co-locomotingparticles was assessed with respect to the ab-solute number of tracked particles in both chan-nels and corrected for stochastically double-labeleddimers by the same fluorophore species:

rel. co – locomotion =

2 � ABAþ B �

1

2 � AA þ B� �

� BA þ B� �h i

Where A, B, and AB are the absolute numberof trajectories observed for Rho11 and DY647 aswell as co-trajectories, respectively (first part ofthe product determines the relative co-locomotionwhile the second part determines the correctionfactor for stochastically double labeling).

Human primary cell culture and analysis

CD34+ hematopoietic stem and progenitor cellsfrommobilized peripheral bloodwere purchasedfrom the Fred Hutchinson Cancer ResearchCenter. Primary cells were cultured as previouslydescribed (18). Cells were cultured with one offollowing ligands: EPO (3 U/mL), A_R3 (10 nM),A_angle_R4 (1 mM), A_angle_R5 (1 mM), A_angle_R6 (1 mM), C_R3 (10 nM), C_angle_R4(100 nM), C_angle_R5 (100 nM), C_angle_R6(100 nM), A_dist_R5 (333 nM), and A_dist_R7(1 mM). Cell concentration was measured usingan automated cell counter (Beckman Coulter).To evaluate erythroid differentiation, cells werestained with CD71-Fluorescein Isothiocyanate(FITC), CD11b/ CD41a-Phycoerythrin (PE), andCD235a-Allophycocyanin (APC). PropidiumIodide(PI; 1:1,000) was used to discriminate live anddead cells. All antibodies were used at a 1:20 di-lution, unless otherwise noted.For phospho-STAT staining, differentiated

primary erythroid cells were starved in cytokine-free media (1% BSA in IMDM) for 4 hours andstimulated either with EPO (3U/mL), A_R3(10 nM), or A_dist_R5 (333 nM) for 1 hour.Treated cells were fixed and permeabilized aspreviously described (18). Then, cells werestained with Alexa Fluor-647 Mouse Anti-STAT5(pY694; 1:20 dilution), Alexa Fluor-647 Mouse

Anti-STAT3 (pY705; 1:20 dilution), or Alexa Fluor-647 Mouse Anti-STAT1 (pY701; 1:20 dilution).Data were acquired using either BD Accuri C6Cytometer or Canto instruments (Becton, Dick-inson and Company) and all analysis was per-formed using FlowJo.

REFERENCES AND NOTES

1. M. Akdis et al., Interleukins (from IL-1 to IL-38), interferons,transforming growth factor b, and TNF-a: Receptors, functions,and roles in diseases. J. Allergy Clin. Immunol. 138, 984–1010(2016). doi: 10.1016/j.jaci.2016.06.033; pmid: 27577879

2. G. M. Delgoffe, P. J. Murray, D. A. Vignali, Interpreting mixedsignals: The cell’s cytokine conundrum. Curr. Opin. Immunol.23, 632–638 (2011). doi: 10.1016/j.coi.2011.07.013;pmid: 21852079

3. J. B. Spangler, I. Moraga, J. L. Mendoza, K. C. Garcia,Insights into cytokine-receptor interactions from cytokineengineering. Annu. Rev. Immunol. 33, 139–167 (2015).doi: 10.1146/annurev-immunol-032713-120211; pmid: 25493332

4. K. Kaushansky, Lineage-specific hematopoietic growth factors.N. Engl. J. Med. 354, 2034–2045 (2006).

topological control of cytokine receptor signaling induces ... · clone, n3c-e2, was expressed in...

Documents