1. The CLBS20 Manufacturing Procedures Isolate Tumor initiating Cells

2. The Tumor Initiating Cells Are Immunogenic

Background 3. The Patient Immune Profile Was Altered by Treatment

4. Markers That Typify Non-Responders, or Predict Survival in the CLBS20 Treatment Group

Conclusions

Immunological modeling and testing

Mechanism of Action of CLBS20 A Patient-Specific Immunotherapy Targeting Metastatic Melanoma

Gabriel Nistor

Aleksandra Poole

Lu Chen

Hans Keirstead

Society for Melanoma Research 2015 Congress

November 18 - 21, 2015 San Francisco

Marriott Marquis San Francisco, California

Methods: • Xenograft in immune-deficient mice • Exomic analyses

Methods: • Exomic analyses and HLA typing • Neoantigenic affinity determination

The investigators are developing an innovative approach to cancer treatment targeting the cells responsible for tumor growth and metastasis, known as tumor-initiating cells (TIC). The therapy uses purified autologous TICs, derived from a patient’s own tumor as the antigen source to induce or enhance an anti-tumoral immune response. Using the immune system to target TICs may provide a complementary or adjunctive therapy that would result in better long-term control, and possibly even a cure, of certain cancers.

This approach has shown consistent and compelling results in sequential open-label and a randomized Phase 2 trials for advanced stage melanoma, and is now being tested in a Phase 3 trial called the Intus Study. The dendritic cells loaded with purified and irradiated tumor initiating cells (CLBS20) were tested against antigen only, purified and irradiated tumor cells (control).

In order to determine whether isolated TIC populations contained antigenic cancer-related mutations, exomic sequencing was performed on tumor cell lines from 5 distinct melanoma tumors. Sequencing results, combined with epigenetic data, indicated the presence of cancer and metastasis-related driver mutations with highly antigenic properties. Identification of these mutated antigens allows us to further study specific T-cell activation using mixed lymphocyte reaction assays.

We analyzed patient serum samples before and 1 week after 3 weekly injections of CLBS20, using comprehensive high throughput screening for proteins related to immune response. The analysis of cytokines demonstrated a decrease of serum tumor markers in parallel with the reduction of immune markers compared to the control arm. The results suggest a combination of humoral and cell immune responses. A set of biomarkers associated with poor response in the treatment arm was identified. These findings demonstrate a complex immune response following administration of CLBS20 immunotherapy and is of critical importance for effective prediction of patient response to therapy.

• Tumorigenicity and exomic analyses of patient derived melanoma cell lines isolated using our methods confirmed the tumor initiating, highly antigenic nature of these cells.

• The serum analyses demonstrated a distinct immune response upon administration of CLBS20. Based on existing data we can link the treatment with CLBS20 with an immunological response that is different from the natural evolution seen in the control group.

• At 3 weeks from first dose, the data suggest the involvement of a Th17 cytotoxic population. The model is verified in vivo by increased IL23 and IL17F and in vitro by IL17A and a RORγt positivity increase after MLR with tumor loaded dendritic cells.

• Shorter survival in CLBS20-treated patients was associated with very high levels of PD1.

• Other mechanisms of response are not excluded. An earlier time point for Th1/CD8 CTL or a later for immunoglobulin response was not analyzed.

• Although suggestive, the findings will be validated in the current Phase 3 clinical trial.

• PD1 has only partial impact on the response to CLBS20 as shown by the OS difference between the low survival groups of the treatment and control arms. Pre-existing immunological response to increased antigen exposure may be inhibited by PD1, however new populations of effector T-cells could respond to new TAAs presented by DCs in non-inhibited manner.

Patient derived melanoma purified lines were transplanted in nude mice. A subcutaneous low dose was administered in one flank, a high dose in the other flank.

We analyzed the affinities of the non-synonymous mutations to the same patient’s HLA system, to determine antigenic properties of the isolated tumor initiating cells

• The neo-affinities were calculated by comparing the mutated peptide sequence with the specificity pockets of the HLA

• A neo-antigenicity score was calculated by integrating the mutation sequence, patient variants, neo affinity values, databases and models of immune response, all MHC Class I and Class II data points

Retrieved palpable tumors were mostly positive for the neural crest marker CD271. (Green: anti-human staining (Ku80); Red: CD271; Blue: cell nuclei (DAPI)

• In total, 90 genes were commonly mutated in all analyzed samples

• A group of 14 exomic point mutations were identical to all 5 patient samples in the 90 mutated genes

• The large amount of common mutations suggest stem cell characteristics

• Analysis of driver mutations confirms the involvement of cell cycle regulatory mechanisms that are associated with gain of function

Patient derived melanoma xenograft in immune-deficient mouse model

Determination of antigenic mutations in patient-derived, purified melanoma cell lines

Exomic analyses of patient-derived, purified melanoma cell lines

Patient cell line

Tumor present/total transplants

Mean tumor size for the entire group

(higher dose) (mm3)

Mean tumor size for the entire group

(lower dose) (mm3)1982 5/5 2361.7 (SD 2106.2) 1176.4 (SD 1160.2)1983 2/5 29.2 (SD 65.3) 106.4 (SD 238.0)1936 5/5 573.9 (SD 266.6) 403.3 (SD 268.7)

Large tumor xenograft in immuno-deficient mouse. Highly proliferative, infiltrative tumor tissue, with neovascularization and absence of necrotic areas. (Hematoxilin and eosin staining, 4x)

RESULTS: • Neo-affinities of exomic non-synonymous mutations to all HLA classes were identified

in all analyzed samples

• For each HLA, 10-30 mutations presented high affinity scores, with immunogenic potential

biological process

cellularprocess

single-organismprocess

single-organismcellularprocess

single-organismmetabolic

process

carbohydratemetabolic

process

macromoleculemetabolic

processsystem development

cell-celladhesion

cell-celladhesion

via plasma-membraneadhesionmolecules

homophiliccell adhesion

via plasmamembraneadhesionmolecules

organic substancemetabolic

process

primary metabolicprocess

cellularmetabolic

process

anatomicalmetabolic

developmentcell adhesion

developmentalprocess

biologicaladhesion

metabolicprocess

single-organismcarbohydrate

metabolicprocess

glycosylation

macromoleculeglycosylation

protein glycosylation

protein O-linked glycosylation

O-glycanprocessing

cellularprotein modi�cation

process

post-translationalprotein modi�cation

cellularmacro molecule

metabolicprocess

macromoleculemodulation

protient modi�cationprocess

cellularprotien metabolic

process

protient metabolicprocess

nervous system development

HG19human

referencegenome

Tumorexome

Normalexome

Antigenicproperty

Defectivepresentation

HLA HLASomatic

Mutations

Algorithm

Examples of mutations with affinities to two different HLA classes

from the same patient * Based on HLA

binding affinity IC50 <150 nM (Sidney et al.

2001, Cai et al. 2012)

Venn diagram of the exomic mutations from purified tumor samples

Methods: • Investigation of serum biomarkers from:- Tumor and immune system - Before treatment and 4 weeks from baseline

Methods: • The control and treatment (CLBS20) groups were divided at median survival to compare responders (high survivors) to non-responders (low survivors)

• Interaction models can be constructed from published knowledge and the measured patient values of individual stimulus/effect markers

• Based on observations in the treatment group (CLBS20), we constructed pathway models targeting survival as a dependent variable

• Linear regressions are common mathematical tools aimed to verify/validate a model

Baseline evaluation:• Tumor markers were 20-40

times higher than normal

• Inflammation markers were proportionally 80-150 times higher than normal

Observation: • The randomization was

biased against the CLBS20 arm with higher tumor and inflammatory markers

• Tumor markers progressed less in the high survivors

• Inflammation is proportionally lower in the high survivors

• The data is consistent with the natural progression of cancer

• Unlike the control arm, the changes in various pathways suggest an interventional effect of the CLBS20 treatment

• Specifically, some inflammatory pathways positively correlate with survival (higher average values in the high survival group): Inflammatory proteins, dendritic cells, T-reg, Th17, NKT, and B cell stimulation

• The lower survival group, a decrease of inflammatory markers despite moderate increase of tumor markers, suggests that this group may have an inhibited inflammatory response, possible explained by higher serum levels of PD1.

Post-treatment evaluation (4 weeks)• Tumor and inflammation

markers continued to increase in the control arm suggesting natural tumor progression

• In the treatment (CLBS20) arm, tumor markers increased only slightly, while inflammation markers slightly decreased from baseline

• The marker analyses were performed according to pathway membership

• The differences between the treatment arms are apparent in:

• Melanoma and other tumor markers

• All inflammatory pathways

Control arm:• Tumor and inflammation markers increased from baseline

Treatment arm:• Tumor markers were only slightly increased from baseline

• The majority of inflammatory pathways decreased from baseline

• A few pathways only slightly elevated from baseline, suggestive of MOA

Statistical evaluation confirms a significant increase of tumor and inflammatory markers in the control arm and a significant decrease of tumor and inflammatory markers in the CLBS20 arm

*P values are calculated with Wilcoxon signed rank test for dependent variables. FDR are calculated using Benjamini–Hochberg method

• Variables were grouped in subsets of 4-10 biomarkers that are specific for biological pathways

• The grouping considered two categories: stimulus markers (input) and effect markers (output)

• Given that the effect markers may represent stimulus markers for other pathways, logical modeling of dependencies can be constructed

• Logical grouping and targeted analyses reduce FDR

Analyses of patient serum data pre-treatment indicate a positive correlation of tumor markers and inflammation

Evaluation of survivor groups in the control arm

Evaluation of survivor groups in the CLBS20 treatment armAnalyses of patient serum data post-treatment indicate an active effect of CLBS20 compared to control

Evaluation of specific pathways, pre-treatment

Evaluation of specific pathways, post-treatment

Pathway Analyses

Immune system

Tumor

Tumor destruction

Tumor prevalence

In�ammation

Tumor markers

In�ammation

Tumor markers

Model of tumor in�ammation

CLBS20 Immune-therapy

Control arm Tumor marker Inflammatory markersLAPTGFb1 IL2Ra TARC uPAR

p* .021 .030 .005 .008FDR% 14.5% 29.9% 9.3% 6.3%Fold change at week 4 1.12 3.33 1.99 1.32

CLBS20 arm Tumor markers Inflammatory markersAngiostatin HGFR S100b GCSF IL1RII IL2 IL23 IL6R

p* .025 .016 .028 .028 .035 .007 .018 .039FDR% 7.4% 22% 8.5% 25.4% 32% 5.1% 10.6% 28%Fold change at week 4 -1.04 -1.09 -1.21 -1.10 -1.08 -1.22 1.15 -1.08

RESULTS: • The immune profile of the melanoma patients was altered by CLBS20 treatment

• The control arm followed the natural evolution of cancer, with progression of tumor markers and associated inflammation

• The CLBS20 treatment is associated with reduced tumor progression and reduced inflammation, with some of the markers at statistical significance

• The Th17 (IL23-associated) pathway was increased after CLBS20 treatment at week 4, suggesting a possible involvement in MOA

RESULTS: • Compared to the control group, the changes in tumor markers and inflammatory

pathways of the treatment group indicate an interventional effect of the CLBS20

• The low survivors of the CLBS20 treatment group did not follow the natural evolution of the cancer as in the control group. These patients presented elevated PD1 at weeks 0 and week 4, that could interfere with CLBS20 effectiveness. Thus the serum level of PD1 could be used as a predictive marker for survival and response in CLBS20 treatment.

• Other tumor and angiogenicc markers with possible predictive value on survival and treatment response: EGFR, HGFR, TGFb1, TGFb3, CA153, CA199, NSE, GRO/GROa, IL1ra, VEGFR1, VEGFD, TIMP1

• Inflammation markers with possible predictive value on survival and treatment response: IFNg, IL12p70, IL8, IL7, IL13, IL15, IL4, IL5, IL17F, IL23, IgG3, FAS, MCSF

• Future studies on larger numbers of patients are required to validate the findings

The baseline analysis of PD1 serum level of the patients enrolled in the Phase 2 randomized trial revealed that the low survival group (LS) of the CLBS20 arm presented significantly (p=0.019) higher values comparative to the high survival group (HS). The significative difference was maintained between the values measured at 4 weeks from baseline (p=0.015)

Increased PD1 serum levels correlate with lower survival in the treatment group

Model of DC-Th17-Tumor pathways interaction based on week 4 immune profile

Naïve T cells Pro-Th17

Regulatory Th17

In�ammatory Th17 Th17/Th1-like

Monocytes

In-tumor DC

In-vitro DC

Regulatory DC

In�ammatory DC

IL4, GMCSFAntigen(s)

IL23p19+IL12p40

IL23IL2IL21

IL17, IL17FIFNg, TNFa

PD1B2M

IL6IL12IL1b

IL10TGFbPD1IL23p19

IL10TGFb

Tumor growth

IL23p19

Regulatory positive loop

Cytotoxic positive loop

hypothesis

Arm B SE Wald *Sig. Exp(B)

Control

IL23_W4 1.733E-06 2.64E-06 .431 .511 1.000IL1b_W4 4.284E-04 6.76E-04 .402 .526 1.000

IL12p40_W4 7.822E-04 1.56E-03 .251 .616 1.001IL17F_W4 -1.435E-06 1.89E-06 .574 .449 0.999

CLBS20

IL23_W4 5.185E-05 2.21E-05 5.521 .019 1.000

IL1b_W4 -4.327E-03 1.78E-03 5.912 .015 0.996

IL12p40_W4 -2.540E-03 1.20E-03 4.480 .034 0.997

IL17F_W4 -3.423E-06 1.51E-06 5.140 .023 0.999

-3

-2

-1

0

1

2

3

4

5

1 10 100Calc

ulat

ed n

eoaffi

nity

scor

es

High <--- Normal peptide ---> Low

HLA A*26:01

Intracellular

Membrane

Secreted

Not identified

-3

-2

-1

0

1

2

3

4

5

1 10 100

HLA DRB1*10:01

Tolerogenicmutations

Immunogenic mutations *

Tolerogenicmutations

Immunogenic mutations *

1000 100000

-3

-2

-1

0

1

2

3

4

5

High <--- Normal ---

5

Tolerogenicmutations

Immunogenic mutations *

Tolerogenicmutations

Immunogenic mutations *

1 10 100 10000

molar affinity

1000 10000 1000001 10 100

Analysis of biological processes affected by common mutations identified O-linked glycosylation, homophilic cell adhesion, and nervous system development

-3

-2

-1

0

1

2

3

4

5

1 10 100Calc

ulat

ed n

eoaffi

nity

scor

es

High <--- Normal peptide ---> Low

HLA A*26:01

Intracellular

Membrane

Secreted

Not identified

-3

-2

-1

0

1

2

3

4

5

1 10 100

HLA DRB1*10:01

Tolerogenicmutations

Immunogenic mutations *

Tolerogenicmutations

Immunogenic mutations *

1000 100000

-3

-2

-1

0

1

2

3

4

5

High <--- Normal ---

5

Tolerogenicmutations

Immunogenic mutations *

Tolerogenicmutations

Immunogenic mutations *

1 10 100 10000

molar affinity

1000 10000 1000001 10 100

The Th17 subset was described as having both regulatory and cytotoxic activity on tumors, the widely accepted theory suggesting a pro-tumorigenic effect. Th17 cells can turn against the tumor if stimulated properly with inflammatory dendritic cells presenting tumor associated antigens (TAA). In the presence of new or increased TAA presentation the Th17 cells can convert to a cytotoxic phenotype based on plasticity, or by differentiation from naive Th17 cells. The model is verified by Cox proportional hazard regression using Th17 pathway as covariables, and by linear regression model, only in the CLBS20 arm

0

50

100

150

200

250

Tumor markers In�ammation

Fold

incr

ease

from

nor

mal

Baseline Tumor and Immune Status

Control (n=20)CLBS20 (n=18)

-10%

-5%

0%

5%

10%

15%

20%

25%

30%

Tumor markers In�ammation

Perc

ent c

hang

e fr

om b

asel

ine

The e�ect of CLBS20 at 4 weeks from baseline

Control (n=20)CLBS20 (n=18)

Control (n=20)

CLBS20 (n=18)

0

50

100

150

200

250

300

350

Fold

incr

ease

from

nor

mal

Baseline Tumor and Immune Biomarkers

Control (n=20)CLBS20 (n=18)

-20%

-10%

0%

10%

20%

30%

40%

Perc

ent c

hang

e fr

om b

asel

ine

The e�ect of CLBS20 at 4 weeks from baseline

-10%

0%

10%

20%

30%

40%

50%

60%

Perc

ent c

hang

e fr

om b

asel

ine

Changes in the control group after 4 weeks Below median survivors Above median survivors

-40%

-30%

-20%

-10%

0%

10%

20%

30%

40%

50%

Perc

ent c

hang

e fr

om b

asel

ine

The e�ect of CLBS20 at 4 weeks from baseline

Below median survivors Above median survivors

RESULTS: Tumorigenicity studies and exomic analyses suggest that the CLBS20 manufacturing process results in a relatively homogenous cell population of tumor initiating cells

0 10 20 30 40 50 600.0

0.2

0.4

0.6

0.8

1.0

DC-TC

Control (TC)

Months

Prop

ortio

n Su

rviv

ing

Randomized Phase 2 Trial CLBS20 vs control in metastatic melanoma (Dillman, R.O., et al. 2012, J Immunother)

Randomized Phase 2 Trial CLBS20 vs. control 2007-2011 (NCT00436930)

Patients were consented and randomized to one of two treatment arms, and 42 received their assigned treatment. An interim analysis was performed. There were no differences between the two arms in patient characteristics or prior therapies.

The clinical results are consistent with the concept of immunizing against TIC.

COX regression using PD1 as co-variate, shows significant effect only on the CLBS20 arm at baseline and after treatment

Time B SE Wald Sig. Exp(B)

Control armBaseline -0.0000044 0.0000045 .965 .326 0.999996Week 3 -0.000005 0.000005 1.001 .317 0.999995

CLBS20 armBaseline 0.0000051 0.0000019 6.935 .008 1.000005Week 3 0.000006 0.000002 6.710 .010 1.000006

Melanoma patient’s PBMCs were examined for IL-17-A and IL-17-F expressing CD4+ cells (A) and for CD4+RORγt+ (B) through flow cytometry under the following control conditions: untreated PBMCs, PBMCs treated with PMA (50 ng/ml) and ionomycin (1 μg/ml) overnight. We tested for the same parameters PBMCs co-cultured with autologous tumor-cell-loaded dendritic cells (CLBS20) (1:3, PBMC to CLBS20 ratio), PBMCs co-cultured with autologous CLBS20 (1:3, PBMC to CLBS20 ratio) with co-stimulatory factors including CD40L (500 ng/ml), anti-CD28 antibody (100 ng/ml), and anti-CD49d antibody (100 ng/ml).

Only the presence of tumor loaded dendritic cells causes the increase of IL17A positive cells and increase in the RORγt+ CD4 population in autologous MLR

1

10

100

1,000

10,000

100,000

1,000,000

0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60

PD1

seru

m le

vel (

pg/m

L)

Median survival (months)

PD1 serum level of patients enrolled in the Phase 2 randomized trial

49,108 pg/mL

52,109 pg/mL

150,196 pg/mL

6,122 pg/mL

Control- LS Control- HSCLBS20- LS CLBS20- HS

Healthy subjects: 595 pg/mL

9.86 42.4418.64 52.96 group median survival

0

50,000

100,000

150,000

200,000

250,000

Before treatment After treatment

PD1

conc

entr

atio

n (p

g/m

L)

Serum PD1 levels of patients enrolled in the Phase 2 randomized trial

High survivors - CLBS20 Low survivors - CLBS20

High survivors - Control Low survivors - Control

p=0.019 p= 0.015