Background/Purpose: Our proprietary variant Ig domain™ (vIgD™) platform creates novel, therapeutically-applicable protein domains with tailored specificity and affinity. These vIgDs are created through directed evolution of immunoglobulin superfamily (IgSF) proteins, which are key components of the immune system that include well-known family members such as PD-1, PD-L1, and CTLA-4. CD28 and Inducible T-cell Costimulator (ICOS) are two related costimulatory molecules within the IgSF which are expressed on T cells and interact with CD80/CD86 and ICOS ligand (ICOSL), respectively. Both play critical roles in T cell activation and adaptive immunity. We have used our vIgD platform to generate human ICOSL vIgD-Fc proteins capable of binding both ICOS and CD28, blocking the interaction of these costimulatory molecules with their respective receptors. Methods: ICOSL vIgD-Fc molecules were evaluated in vitro in mixed lymphocyte reactions (MLR) generated using negatively-selected human pan T cells mixed with activated human monocyte-derived dendritic cells, and in vivo in standard mouse models of delayed type hypersensitivity (DTH), collagen-induced arthritis (CIA), and human-mouse xenograft PBMC-NSGTM graft versus host disease (GvHD). Results: ICOSL vIgD-Fc fusion proteins containing variant ICOSL domains significantly attenuate T cell activation in vitro as assessed by suppressed proliferation and cytokine production in MLR. They also reduce mouse DTH reactions in vivo (as previously shown). ICOSL vIgD-Fc molecules mediate significant disease reduction, matching or exceeding CTLA-4-Ig comparators targeting only the CD28 pathway in mouse CIA, and in the human PBMC-NSGTM GvHD model. ICOSL vIgD-Fc inhibits multiple parameters of disease in the CIA mouse model, including reducing paw swelling and inflammatory cell infiltrates, lowering serum levels of inflammatory cytokines and anti-collagen autoantibodies, and decreasing follicular helper T cells (Tfh), B cells, and activated T cells in the paw-draining lymph nodes. Conclusion: Efficacy in vitro and in vivo of ICOSL vIgD-Fc is superior to wild-type ICOSL domains due to the induced alterations in affinity for cognate ligand (ICOS) and through specifically directed changes in ICOSL vIgD-Fc’s ability to bind additional counter-structures (i.e. CD28). Thus, vIgDs like these ICOSL variants can be developed to acquire unique biochemical properties that can potentially significantly enhance their therapeutic utility as immunomodulatory agents. This vIgD therapeutic platform has broad potential to enhance the activity of biologics in treatment of autoimmune and other disorders driven or subject to modulation by IgSF proteins, such as cancer and infectious diseases. Preclinical development of ICOSL vIgD-Fc has been initiated to support clinical studies.

A Dual ICOS/CD28 Antagonist ICOSL Variant Ig Domain (vIgDTM) Potently Suppresses Mouse Collagen-Induced Arthritis and Human Xenograft Graft vs. Host Disease (GvHD) Stacey R. Dillon, Katherine E. Lewis, Ryan Swanson, Lawrence S. Evans, Michael G. Kornacker, Steven D. Levin, Martin F. Wolfson, Erika Rickel, Susan J. Bort, Sherri Mudri, Aaron M. Moss, Michelle A. Seaberg, Janhavi Bhandari, Sean MacNeil, Joe Hoover, Mark W. Rixon, and Stanford L. Peng

Alpine Immune Sciences, Inc. 201 Elliott Avenue West, Seattle, Washington, USA

Abstract

Figure 1: The vIgD Platform

Fig 2: vIgDs can be Formatted Specifically for Various Therapeutic Applications

Fig 3: CD28 and ICOS Mediate T Cell Costimulation

Fig 4: Simultaneous Affinity Maturation of ICOSL Towards Two Receptors

Fig 5: ICOSL vIgD-Fc, an ICOS/CD28 Dual Antagonist for Autoimmunity/Inflammation

Fig 6: ICOSL vIgD-Fc Inhibits Lymphocyte Proliferation & Cytokine Secretion In Vitro

Fig 8: ICOSL vIgD-Fc More Effective than Abatacept in the CIA Mouse Model of RA

IL-6 and TNFa was measured in serum collected at 17 or 24 days post-collagen boost in the CIA model described in Fig 8. Paw-draining lymph nodes (LN) were collected at the end of the study and cells were analyzed by flow cytometry. ICOSL vIgD-Fc treatment significantly reduced both IL-6 and TNFa in the serum, and reduced B cells, activated CD44+CD4+ or CD8+ T cells, and follicular helper cells (Tfh) in the LN, as compared to PBS and/or Fc control treatments.

• A variant Ig domain (vIgD) platform generates novel immunomodulatory IgSF-based biologic therapeutics with higher affinity and increased multiplicity of ligand binding, translating into superior preclinical efficacy in vitro and in vivo

• In human MLR assays, ICOSL vIgD-Fc’s repeatedly demonstrate superior efficacy in vitro to blockade of the CD28 (vs belatacept, abatacept) or ICOS (vs WT ICOSL-Fc) pathways alone for inhibition of cellular proliferation, as well as intracellular cytokine production and secretion

• ICOSL vIgD Fc’s performed as well as, or better than, CD28- or ICOS-only pathway blockade in vivo in DTH, humanized GVHD, and collagen-induced arthritis mouse models

• ICOSL vIgD-Fc’s inhibit multiple parameters of disease in the collagen-induced arthritis mouse model, including reducing paw swelling and inflammatory cell infiltrates, lowering serum levels of inflammatory cytokines and anti-collagen autoantibodies, and decreasing Tfh cells, B cells, & activated T cells in the paw-draining lymph nodes

• The vIgD therapeutic platform has broad potential to enhance the activity of biologics in treatment of autoimmune and other disorders driven or subject to modulation by IgSF proteins, such as autoimmunity, cancer, and infectious diseases

• Preclinical development of ICOSL vIgD-Fc (ALPN-101) is underway to support clinical studies

Fig 9: ICOSL vIgD-Fc Inhibits Paw Inflammation in CIA More Effectively than Abatacept

Fig 10: ICOSL vIgD-Fc Reduces Inflammatory Cytokines, TFH, B Cells, & Activated T Cells in CIA

Summary and Conclusions

Bead and flow cytometric selection

vIgD

Fc-fusion protein

generation

Counter-structure binding

Functional assays

IgSF ECD Yeast Display Libraries

IgSF protein

Therapeutic protein

Limited counter-structures with low/moderate affinity

Random or targeted mutagenesis of ICOSL ECD

Parental ICOSL • Binds ICOS

Selections: rICOS and rCD28

Binding assays • Flow cytometry • Octet (affinity)

MLR • Proliferation • IFNγ

production

Screen yeast outputs for improved binding to rICOS and rCD28

Sequence yeast outputs and identify unique variant hits

Transient 293 or CHO production followed by Protein A purification

Tailored counter-structures with improved/high affinity

CAR-T (TIP) Autoimmune

ICOSL ECD

Fc

ICOSL vIgD

Oncology

mAb

Costimulatory Agonist

Costimulatory V-mAb

Fc

Target 1

Target 2

vIgD Multi-Checkpoint

Antagonist Costimulatory Agonist

Fc

Localizer

Localized vIgD

vIgDs from Alpine’s directed evolution platform may have multiple therapeutic formats • Fusion proteins (Fc or mAb) with antagonistic or tumor-localizing agonistic activity • Cell-displayed (TIP) or secreted (SIP) versions for enhancement of adoptive cellular therapies

T cells express the costimulatory molecules CD28 and ICOS, which interact with CD80/CD86 and ICOS-L respectively, on antigen presenting cells (APC). In lymphoid organs, professional APC (i.e. dendritic cells, macrophages, and B cells) express CD80, CD86, and ICOSL and engage CD28+/ICOS+ T cells. Activated T cells can then differentiate into effector cells such as CD8+ cytotoxic T cells (CTL), IL-17A/F-secreting CD4+ Th17 cells, or CD4+ follicular helper (TFH) cells. TFH expressing CD40L engage B cells in lymphoid follicles and release cytokines (e.g. IL-21) inducing differentiation of B cells to antibody (Ab)-secreting plasma cells. Plasma cells can produce tissue-damaging Abs like rheumatoid factor (RF) and anti–citrullinated peptide antibodies (ACPA) in humans, and anti-collagen (CII) Abs in mice, which can form immune complexes and deposits in the joints and other tissues. ICOSL can also be expressed on non-professional APCs, leading to T cell activation in non-lymphoid tissues and further damage to the tissues and joints. Our goal was to use the vIgD platform to isolate an ICOSL molecule that could bind and block both CD28 and ICOS and co-inhibit these critical T cell costimulatory pathways for the treatment of autoimmune and inflammatory disorders.

Dual-Ligand Affinity Maturation of a Single Random ICOSL vIgD Library Yeast were transformed with a single random vIgD library and affinity matured by selection with two cognate receptors. Individual receptor binding to bulk yeast populations are shown, including two rounds of flow cytometric selection. Improvements to both ligands over wild-type (WT) ICOSL were noted at 1st Gen (F1), and were further improved at 2nd Gen (F2). MFI, mean fluorescence intensity. -1 0 1 2

0

200

400

600

800

CD28 Binding

CD28 Fc log[ nM ]

MFI

WT ICOSLF1F2

-2 -1 0 1 2

200

600

1000

1400

ICOS Binding

ICOS Fc log[ nM ]

MFI

WT ICOSLF1F2

APC T cell

ICOSL vIgD

CD80/CD86

CD28

ICOS

ICOSL

CD80/CD86 CD28

ICOS ICOSL

Activating Blocking

Fc

ICOSL

A) Proliferation in the human MLR (left) was determined by quantitating the percentage of CFSE-labeled cells diluting CFSE over time. Effects of ICOSL vIgD-Fc’s and controls on CD8+ T cells & CD4+ T cells are presented (top left). Intracellular cytokine staining was performed on cells stimulated with PMA/ionomycin in the presence of brefeldin A & monensin (bottom left). B) Purified human CD4+T cells and allogeneic B cells were CFSE labeled and incubated for 7 days in the presence or absence of ICOSL Fc-fusion proteins. Cells were collected, stained for CD4, CD19, and CD38 and analyzed by flow cytometry. Effects on the number of CD38+ and divided (CFSE-) B cells (CD4-CD19+) are shown (mean and standard deviation of assay triplicates).

0.01 0.1 1 10 10015

25

35

45

CD8+ T-cells

% D

ivid

ed C

ells

0.01 0.1 1 10 100

15

25

35

CD4+ T-cells

% D

ivid

ed C

ells

0.01 0.1 1 10 1005

10

15

20

25

IL-21+ CD4+

Protein [ nM ]

% P

ositi

ve

0.01 0.1 1 10 10015

20

25

30

IFN-+ CD4+

Protein [ nM ]

% P

ositi

ve

0

500

1000

1500

2000

LegendLegendLegendLegendLegend

Fc Control WT ICOSL ICOSL vIgD B cells alone CD4+ cells alone

Allogeneic B cell/CD4+ T cell Co-cultures

Num

ber o

f CD

38+

/ C

FSE-

B c

ells

A B

Fig 7: ICOSL vIgD-Fc More Effective than Belatacept in Human PBMCNSG™ Model

Mean (+SD) total paw scores:

Day 7 post-boost Day 18 post-boost0

200

400

600

Mou

se a

nti-m

ouse

CII

IgG

(g/

mL)

PBSFc ControlAbatacept1st Gen ICOSL vIgD-Fc

**

* p <0.05 by1-way ANOVA

Serum mouse anti-mouse collagen IgG:

0 5 10 15 20 25 300

2

4

6

8

Day post-boost

mea

n su

m p

aw s

core

PBSFc ControlAbatacept (WT CTLA-4-Fc)1st Gen ICOSL vIgD-Fc

Doses

***

p=0.065 (ICOSL vIgD vs belatacept)

0 5 10 15 20 25 30 35 40 45 500

25

50

75

100

Day

Perc

ent s

urvi

val

Saline

WT ICOSL-Fc

ICOSL-Fc vIgDBelatacept (Nulojix)

Last Dose

p<0.001 (ICOSL vIgD or belatacept vs Saline or WT ICOSL-Fc)

0 5 10 15 20 25 30 35 40 45 500

1

2

3

4

5

6

7

8

Day

Mea

n D

AI s

core

Disease Activity Index (DAI)

Last Dose

p<0.001 (ICOSL vIgD or belatacept vs Saline or WT ICOSL-Fc)

p=0.035 (ICOSL vIgD vs belatacept)

Survival

A B

C D

0.5 mm

* *

Paws were collected at the end of the CIA study described in Fig 8, and stained with H&E. Shown are representative histopathologies from (A, B) unaffected ICOSL-vIgD-Fc-treated forelimbs, (C) inflamed Fc control-treated hindlimb, and (D) inflamed abatacept-treated hindlimb. Note the presence of severe inflammatory pannus (asterisks) present in the Fc control- and abatacept-treated paws, but absent in ICOSL vIgD-Fc-treated specimens.

Serum IL-6 Serum TNFa

PBS - d. 1

7

Fc - d 17

Abatace

pt - d 17

ICOSL-Fc -

d 17

PBS - d. 2

4

Fc - d 24

Abatace

pt- d 24

ICOSL-Fc -

d 240

200

400

600

800

pg/m

L

*

PBS - d 17

Fc - d 17

Abatace

pt - d 17

ICOSL-Fc -

d 17

PBS - d. 2

4

Fc d 24

Abatace

pt - d 24

ICOSL-Fc -

d 240

10

20

30

40

pg/m

L

** *

** *

PBS Fc

Abatace

pt

ICOSL-Fc

naive

0

1

2

3

4

% C

D4

T ce

lls

**

PBS Fc

Abatace

pt

ICOSL-Fc

naive

0

10

20

30

% C

D4

T ce

lls

******* TFH =

CD25- CD4+ PD-1+ CXCR5+

PBS Fc

Abatace

pt

ICOSL-Fc

naive

0

20

40

60

80

% li

ve ly

mph

ocyt

es

*

* p< 0.05 ** p< 0.01 *** p< 0.001 **** p< 0.0001 (by 1-way ANOVA)

B cells as % live lymphocytes

CD44+ CD4+ T cells CD44+ CD8+ T cells TFH cells as % of CD4+ T cells

PBS Fc

Abatace

pt

ICOSL-Fc

naive

20

25

30

35

40

% C

D8

T ce

lls

*

Inject chick collagen

II/CFA (tail)

Boost: Inject chick collagen

II/IFA (tail)

0 1 2 3 4 5 6 7 Weeks Study end

Dosing

Study Design:

* p<0.05 for 1st Gen ICOSL vIgD-Fc vs. abatacept ** p<0.001 for 1st Gen ICOSL vIgD-Fc vs. PBS (by 2-way repeated-measures ANOVA)

Study Protocol: On Day –1, female NOD.Cg-Prkdc<scid> Il2rg<tm1Wjl>/SzJ (NSGTM) mice (9/group) were irradiated (100 rad) & administered 10 mg human gamma globulin SC. On Day 0, mice received 10x106 human PBMC, IV. IP dosing began and continued 3x weekly (M, W, F) through Day 37: 100ug WT ICOSL-Fc or ICOSL vIgD-Fc & molar equivalent (75ug) belatacept. On Day 14, all mice were bled, and engrafted human CD45+ cells were phenotyped by flow cytometry. Study was terminated on Day 49. Endpoints included survival, body weight loss, and disease activity index. Study was performed at JAX Labs (Sacramento, CA).

Professional APC Lymph Node

Synovial Fibroblasts, Epithelial Cells, etc. Non-Lymphoid Tissue

CD80/86

CTLA-4 ICOS TCR CD28

CD80/86

ICOSL

MHC

ICOS TCR

ICOSL

MHC

T cell T cell

Effector T cell (CD8+ CTL, Th17, etc)

TFH B cell Plasma cell

Disease-Associated Antibodies (RF, ACPA, anti-CII, etc.)

sIg

IL-21R IL-21

TNFa, IL-1 IL-6, IL-8

IL-12

Secreted Ig

CD40L CTLA-4

ICOS TCR

CD28 CD80/86 MHC

CD40 CD80/86 ICOSL

WT ICOSL 1st Gen vIgDs Human IgG Belatacept 2nd Gen vIgDs

3rd Gen vIgDs