Please do not remove this page

A Translational Approach to Enzyme-Mediated Retinal Lipofuscin Removal for Atrophic Age-Related Macular Degeneration and Stargardt's Disease

Moody, Kelsey James https://scholarship.miami.edu/discovery/delivery/01UOML_INST:ResearchRepository/12355252850002976?l#13355471410002976

Moody, K. J. (2019). A Translational Approach to Enzyme-Mediated Retinal Lipofuscin Removal for Atrophic Age-Related Macular Degeneration and Stargardt’s Disease [University of Miami]. https://scholarship.miami.edu/discovery/fulldisplay/alma991031447210902976/01UOML_INST:ResearchR epository

Open Downloaded On 2021/09/26 10:17:17 -0400

Please do not remove this page

UNIVERSITY OF MIAMI

A TRANSLATIONAL APPROACH TO ENZYME-MEDIATED RETINAL LIPOFUSCIN REMOVAL FOR ATROPHIC AGE-RELATED MACULAR DEGENERATION AND STARGARDT’S DISEASE

By

Kelsey James Moody

A DISSERTATION

Submitted to the Faculty of the University of Miami in partial fulfillment of the requirements for the degree of Doctor of Philosophy

Coral Gables, Florida

May 2019

©2019 Kelsey James Moody All Rights Reserved

UNIVERSITY OF MIAMI

A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy

A TRANSLATIONAL APPROACH TO ENZYME-MEDIATED RETINAL LIPOFUSCIN REMOVAL FOR ATROPHIC AGE-RELATED MACULAR DEGENERATION AND STARGARDT’S DISEASE

Kelsey James Moody

Approved:

______Zafar Nawaz, Ph.D. Feng Gong, Ph.D. Professor of Biochemistry and Associate Professor of Biochemistry Molecular Biology and Molecular Biology

______Michal Toborek, M.D., Ph.D. Sanjoy Bhattacharya, Ph.D. Professor of Biochemistry and Professor of Ophthalmology Molecular Biology

______Ralf Landgraft, Ph.D. Guillermo Prado, Ph.D. Associate Professor of Biochemistry Dean of the Graduate School and Molecular Biology

______Gerold Feuer, Ph.D. Scientific Consultant, HuMurine

MOODY, KELSEY JAMES A Translational Approach to Enzyme- (Ph.D., Biochemistry and Molecular Mediated Retinal Lipofuscin Removal for Biology) Atrophic Age-Related (May 2019) Macular Degeneration and Stargardt’s Disease

Abstract of a dissertation at the University of Miami.

Dissertation supervised by Professor Zafar Nawaz. No. of pages in text. (116)

Despite significant resource expenditures and decades of focused research, chronic age-

associated diseases have not been met with improvements in patient outcomes that reflect

the successes against infectious disease over the last century. This observation calls into

question contention approaches to drug discovery for diseases of aging, and more

importantly, their molecular targets. Herein, the strategies for engineered negligible

senescence (SENS) framework for drug discovery -- an approach to treating aging

broadly based on damage mitigation -- is presented. A translational program for age-

related macular degeneration is detailed as illustration of this approach. Collectively, this

work validates the academic and commercial viability of the SENS damage repair

platform and presents a novel therapeutic modality that may eventually treat the global leading cause of vision loss in the elderly.

Dedication To all future generations that may take for granted the cures we strive to develop today.

iii

Acknowledgment

This work would not have been possible without foundational training and steadfast support from my family, research team, mentors, and investors:

Annette Deyo Bill Hunt Meegan Sleeper Adam Blanden Mark Yandell Brandon Moyer Aaron Wolfe Barry Moore Kelly Moody Cornelis Wortel Mike Kope Miles Moody Nick LeClair Lisa Fabiny Doug Hagrman Kris Grohn John Schloendorn Eric Zluhan Stephanie Martens Nick LeClair Danique Wortel Nichole Fish David Gobel Donald Levine Zach Thomas Roger Bagg William Tooke Kathleen Kelly Leon Apel Tom Curle Anthony Bianchi Gerold Feuer William Faloon Scott Campbell Doyle Lab Tomas Burl David Reed Rohrer Lab

For the work described in chapter 2, the authors would like to thank SENS

Research Foundation, Center State CEO, and our private investors for their assistance in

funding this work, and to SENS Research Foundation for the original conception of this

approach via the LysoSENS program.

For the work described in chapter 3, the authors are thankful to Maximus Peto and

Aubrey de Grey for providing valuable feedback for this manuscript. This work was

supported by grants from the Foundation and Longecity (formerly

Immortality Institute).

For the work described in chapter 4, we are grateful to Gerold Feuer, Maximus

Peto and Aubrey de Grey for providing valuable feedback for this manuscript, and to the staff at the Fayetteville Free Library for 3D printer training and facility access. This work was supported by grants from the Life Extension Foundation and Longecity (Immortality

Institute).

iv

Disclosures

For the work described in chapter 2, the authors declare the following competing financial interest(s): All authors are employees or interns of for-profit companies or affiliate companies that own and commercialize the described technology (LysoClear,

Inc.) excepting R.D., B.R., E.O., and J.T. R.D. receives consistent grant support from

LysoClear, Inc. and Ichor Therapeutics, Inc. to support his work on A2E and macular degeneration. A.R.B., A.J.W., and K.J.M. have a substantial equity positions in

LysoClear, Inc. and Ichor Therapeutics, Inc., both of which may profit from the technology.

For the work described in chapter 3, KJM holds equity positions in the following for-profit stem cell therapy companies; Ichor Therapeutics, Inc., ImmunePath, Inc., and

Advanced Cell Technologies, Inc. EZ and DW hold equity positions in Ichor

Therapeutics, Inc. GF holds equity positions in Humurine Technologies, Inc. and Ichor

Therapeutics, Inc. The authors declare no further conflicts.

For the work described in chapter 4, KM holds equity positions in several for- profit stem cell therapy companies, including Ichor Therapeutics, Inc., ImmunePath, Inc., and Advanced Cell Technologies, Inc. EZ and DW hold equity positions in Ichor

Therapeutics, Inc. The authors declare no other disclosures.

v

TABLE OF CONTENTS

Page

LIST OF FIGURES ...... vii

LIST OF TABLES ...... viii

Chapter

1 INTRODUCTION ...... 1 Strategies for Engineered Negligible Senescence (SENS)…………………… 2 A Brief Review of Age-Related Macular Degeneration………………………. 4 Narrative Introduction ...... 10

2 RECOMBINANT MANGANESE PEROXIDASE REDUCES A2E BURDEN IN MOUSE MODEL OF STARGARDT’S DISEASE ...... 15

3 LEAN START-UP: A COMPREHENSIVE CASE STUDY IN THE ESTABLISHMENT OF LABORATORY INFRASTRUCTURE ...... 65

4 AUTOMATING HESC DIFFERENTIATION WITH 3D PRINTING AND LEGACY LIQUID HANDLING SOLUTIONS ...... 86

5 CONCLUSION ...... 103

WORKS CITED…………… ...... 108

vi

LIST OF FIGURES

Figure 1.1. Classes of age-associated damage ...... 12 Figure 1.2. SENS damage-repair approach ...... 13 Figure 1.3. Summary of the Visual Cycle ...... 14

Figure 2.1. wtMnP and rMnP activity in vitro ...... 48 Figure 2.2. rMnP activity in cell-free and cell culture models ...... 50 Figure 2.3. rMnP activity in vivo ...... 52 Figure 2.S1. Monosaccharide chromatograms of rMnP performed by UC San Diego GlycoAnalytics Core ...... 55 Figure 2.S2. DMP activity of wtMnP and rMnP ...... 56 Figure 2.S3. A2E degradation by rMnP in the presence and absence of oxygen 57 Figure 2.S4. Breakdown of A2E by (A) wtMnP and (B) rMnP by overnight incubation as a function of pH ...... 58 Figure 2.S5. Representative A2E breakdown kinetics with wtMnP and rMnP ... 59 Figure 2.S6. CD206 detection on (A) ARPE-19 and primary (B) RPE cells by flow cytometry ...... 60 Figure 2.S7. A2E concentration in Abca4(-/-) after 6 weekly doses of rMnP ..... 61 Figure 2.S8. Mouse weights for the efficacy study in 2.3B ...... 62 Figure 2.S9. ERG recordings for the 12-week 56 µg x 6 dose toxicity study ..... 63 Figure 2.S10. Normalized HPLC chromatograms (left) and absorbance spectra of the indicated major peak (right) ...... 64

Figure 3.1. Business Model Canvas ...... 85

Figure 4.1. Custom cell scraper and filter box labware ...... 93 Figure 4.2. Work surface configurations for automated liquid handling robot ... 94 Figure 4.3. Differentiation of human embryonic stem cells into monocytes ...... 96 Figure 4.4. Representative data by flow cytometry for monocyte differentiation 97 Figure 4.S1. Automated liquid handling robot performs cell culture with accuracy comparable to human technicians ...... 98 Figure 4.S2. Culture plate passage diagram ...... 100 Figure 4.S3. Liquid handler and enclosure ...... 101

vii

LIST OF TABLES

Table 2.S1. Side effects of rMnP intraocular injection after 6 weekly treatments (pooled) ...... 54

viii

CHAPTER 1: INTRODUCTION

Since the advent of antibiotics, vaccinations, and improvements in hygiene, modern medicine has been profoundly successful in managing infectious disease. In

1900, pneumonia, tuberculosis, and diarrhea were the leading cause of death in the

United States1. Few patients now perish from these formerly deadly afflictions. The resulting drop in infant and youth mortality led to a significant increase in the mean human lifespan, which was 78.6 years in 2016 in the United States2. Infectious disease has been replaced by chronic age-associated diseases in the United States, with heart disease and cancer accounting for approximately 40% of all deaths within the country3.

Despite significant resource expenditures and decades of focused research, these and other chronic age-associated diseases have not been met with proportional improvements in patient outcomes. This observation calls into question conventional approaches to drug discovery for diseases of aging, and more importantly, their molecular targets. Herein, the strategies for engineered negligible senescence (SENS) framework for drug discovery -- an approach to treating aging broadly based on damage mitigation -- is presented. A translational program for age-related macular degeneration is detailed as illustration of this approach. Finally, guidance is provided for establishing such a program de novo in a commercial setting. Collectively, this work validates the academic and commercial viability of the SENS damage repair platform and presents a novel therapeutic modality that may eventually treat the global leading cause of vision loss in the elderly.

1

2

Strategies for Engineered Negligible Senescence (SENS)

At its most basic level, aging can be thought of as the process by which damage

within a system leads to the eventual dysfunction of that system. Aging is a side-effect of existing, and the manifestations of aging within a system depend upon the environment and conditions in which the system operates.

For most biological organisms, aging can be thought of as localized and systemic dysfunction that eventually leads to pathology. Although a comprehensive review of prevailing theories of aging is beyond the scope of the work presented here and has been

reviewed elsewhere4, it is generally well accepted that aging is a side-effect of the

metabolic processes that allow an organism to survive in the first place5. Metabolism is highly efficient, yet imperfect, and over time small imperfections in metabolism lay down damage. After reaching a critical threshold, this damage can begin to cause pathology, which increases the risk of death to the organism.

Previous work by the biogerontologist Dr. Aubrey de Grey sought to establish a road map to therapeutically remediate human aging6–8. Broadly, de Grey described seven

categories of damage that accumulate with aging as a by-product of metabolism (Figure

1.1). They include, 1) cell loss and tissue atrophy, 2) cancerous cells, 3) mitochondrial

mutations, 4) death-resistant (senescent) cells, 5) extracellular matrix stiffening, 6)

extracellular aggregates, and 7) intracellular aggregates. Further, de Grey coined the term

“strategies for engineered negligible senescence”, or “SENS”, which refers to a series of

parallel therapies to repair each form of damage, with the eventual goal of returning

aging tissues to health and bringing back youthful function.

3

The SENS approach has several significant implications in the treatment of age-

associated disease, which to date have remained largely overlooked by drug developers.

First, many conventional treatments for chronic age-associated diseases require continuous, often lifelong administration. This is because they target metabolic processes directly (e.g. statins for hypercholesterolemia, insulin for type 2 diabetes), or treat downstream symptoms of damage (e.g. stents for vascular occlusion, blood pressure regulators). The SENS approach (Figure 1.2.) is an intermittent maintenance approach, aimed at periodically removing the offending damage before it accumulates to a level sufficient to promote pathology. Because most forms of aging damage take a lifetime to accumulate, therapies that target the damage need to work, but they may not necessarily need to work all that well to have significant clinical efficacy. Further, SENS therapies could allow for smaller therapeutic indices because they would be administered intermittently, rather than chronically to the patient.

Another implication of the SENS approach is the prevalence of orphan correlates.

For many damage classifications, examples of accelerated damage accumulation caused by genetic mutations exist. These accelerated accumulations can manifest as juvenile diseases, many of which are orphan diseases. In some cases, broad accelerated aging phenotypes are clinically apparent, as is the case with progeroid syndromes like Ataxia-

telangiectasia9, Werner syndrome10, Cockayne syndrome11, Trichothiodystrophy12, XPF-

ERCC1 progeroid syndrome13, and Ataxia-telangiectasia-like disorder 214. Other times,

the phenotype is restricted to elements of a specific adult-onset indication, such as

familial hypercholesterolemia15 or Stargardt’s macular degeneration16. Regardless of the

presentation, these afflictions provide drug developers with additional disease indications

4

for their products and opportunities for accelerated approval processes among other

benefits17. Further, juvenile onset diseases (by definition) typically progress faster and

present earlier than their age-associated counterparts, so pivotal clinical studies can be

completed faster than in an aged cohort. This assumes, however, that a sufficient patient

population exists for timely clinical trial enrollment -- a known hurdle for many clinical

programs and especially for rare diseases18–20. The presence of juvenile diseases resulting

from a given form of damage also offers significant confidence in the validity of that

damage as a molecular target for pharmaceutical intervention.

A brief review of age-related macular degeneration

Age-related macular degeneration is the leading cause of blindness in the elderly

Age-related macular degeneration (AMD) is the leading cause of blindness in

individuals over the age of 5021. The prevalence of AMD in North America is

approximately 20 million, and this number is expected to rise to approximately 30 million

by 2020. AMD can be subdivided into two forms, “dry” and “wet”. Dry or atrophic AMD represents 90% of all cases and manifests clinically as a gradual onset of blurriness and loss of visual acuity. Severe AMD, the neovascular or wet form, represents 10% of all cases and is a late-stage manifestation of the disease, characterized by pathological angiogenesis, which can lead to complete blindness22,23.

Stargardt’s macular degeneration (SMD) is an inherited juvenile onset form of

macular degeneration. The most common form of SMD, responsible for 90% of cases, is

caused by mutations in the ABCA4 gene, which is necessary for retinoid metabolism. The

prevalence of SMD is estimated to be between 1 in 8,000 to 1 in 10,00024.

5

The visual cycle is the process by which light is converted into electrical signals

for interpretation by the brain. Briefly, the cycle begins when 11-cis-retinal within retinal

pigmented epithelium (RPE) is transported to photoreceptors and combined with opsin to

form rhodopsin. When rhodopsin is photoisomerized, all-trans-retinal is released and

opsin is recycled. All-trans-retinal is reduced to all-trans-retinol then transported back to

the RPE, where is it converted back to 11-cis-retinal, completing the cycle. This process

is reviewed here25 and summarized in Figure 1.3.

The accumulation of aberrant retinoid aggregates and subsequent lipid accumulation (collectively called lipofuscin) in RPE lysosomes has been implicated in both AMD and SMD and is thought to be a side effect of the visual cycle. Briefly, when all-trans-retinal (ATR) is released from opsin, it can become linked to phosphatidyl ethanolamine (PE) as a Schiff-base (ATR-PE) in a condensation reaction. The flippase

ATP binding cassette transporter A4 (ABCA4) transfers ATR-PE into the extracellular space, and in the presence of only trace ATR, the reaction runs in reverse. This results in

ATR-PE hydrolysis into ATR and PE. However, if ATR-PE reacts with a second ATR

molecule before being flipped by ABCA4, both ATR molecules become condensed with

the ethanolamine and form A2-PE. A2-PE cannot be transported by ABCA4 and thus

remains part of the photoreceptor cell membrane until it is released as a rod outer

segment (ROS), after which ROS/A2-PE is endocytosed by RPE and delivered to the

RPE lysosomes. Lysosomal enzymes act on A2-PE, forming a variety of by-product, perhaps the best studied of which is N-retinylidene-N-retinyl-ethanolamine (A2E), reviewed here26.

6

Macular degeneration occurs as a side effect of the visual cycle

ABCA4 is highly efficient at removing ATR-PE, so the formation and accumulation of A2E and other lipofuscin retinoids within the lysosomes of RPE occurs

slowly over the human lifespan. At lower levels, these species are of negligible clinical

significance. However, growing lipofuscin burden over time eventually causes lysosomal

de-acidification, physical obstruction (A2E, for example, can represent up to 20% of RPE cells by mass), and direct cytotoxicity through detergent-like properties27–29. These insults

impair lipid metabolism and increase the release of free cholesterol30, leading to the

accumulation of additional intracellular lipofuscin and later, extracellular deposition,

which presents as extracellular drusen. Eventually, intracellular and extracellular insults

drive apoptosis of the RPE. Extracellular drusen and RPE cell death contribute to a pro-

inflammatory state characterized by complement activation, chronic inflammation,

carbohydrate and lipid oxidation, and advanced glycation end-product generation31–33.

The death of RPE also leads to the progressive loss of photoreceptors that rely upon them

for catabolism and retinoid recycling. The net result of this cascade is the gradual loss of

central vision, clinically termed "dry" or atrophic AMD.

As dry AMD progresses, regional dysfunction can occur. Pathological

angiogenesis originating from the choroid -- the blood vessels posterior to the RPE -- is a serious risk for patients with AMD. This so called "wet" or neovascular form of AMD can rapidly lead to complete blindness due to ocular damage resulting from swelling, inflammation, or hemorrhage. Secondary complications such as retinal detachment are not uncommon34.

7

SMD is a juvenile onset form of AMD. SMD patients have genetic defects in

ABCA4. This causes rapid accumulation of A2E because the condensation of a second

ATR to intracellular ATR-PE, usually a rare event, becomes common due to the failure

of ABCA4 to transfer ATR-PE into the extracellular space35. Disease progression is

similar to that of dry AMD patients, and SMD patients can become symptomatic as early

as ages 6-12, though severity is highly variable and is a function of the extent of ABCA4

dysfunction36.

Standard of care and emerging therapies are insufficient

Modifiable risk factors for AMD include smoking and diets with low

antioxidants. These factors are thought to be related to the high susceptibility of the

retina, particularly photoreceptors, to oxidative stress37.

The current standard of care for dry AMD was established by the Age-Related

Eye Disease Study (AREDS), a National Eye Institute sponsored clinical trial from 1992

to 2001 that evaluated the effects of high-dose vitamins C and E, beta-carotene, and zinc supplements on AMD progression and visual acuity in 3640 patients38. The study found

that progression to advanced AMD could be reduced by approximately 25% by regular

supplement consumption. AREDS2, a smaller follow-up study, modified the

recommendation to remove high-dose beta-carotene administration to smokers because of a higher risk of lung cancer in this cohort39.

Existing treatments for wet AMD are aimed at reducing neovascularization by

targeted inhibition of vascular endothelial growth factor (VEGF) pathway, typically using

monoclonal antibodies. Ranibizumab is the most commonly used and is marketed under

8

the trade name Lucentis by Genentech and Novartis. Bevacizumab is another AMD

therapy, and is marketed under the trade name Avastin by Genentech and Roche. Both

products are humanized mouse antibodies that target VEGF-A and are administered as monthly injections into the vitreous humor of the eye. They display comparable efficacy and safety profiles in clinical trials40.

Most emerging therapies for AMD are designed to interrupt the VEGF pathway.

Examples span gene therapy, receptor tyrosine kinase inhibitors, and steroids, among

others41. A notable exception to this trend is the generation of RPE cells de novo from

pluripotent stem cells or neural stem cells for use as a cell-based therapy42.

There are currently no available treatments for SMD. Sanofi has licensed the gene

therapy product StarGen from Oxford BioMedica. Ocata Therapeutics (formerly

Advanced Cell Technology) has generated RPE cells from human embryonic stem cells.

Phase I/IIa clinical trials for each of these therapies are ongoing (clinical trial identifier

NCT01367444 and NCT01345006, respectively).

Augmenting RPE lysosomes with exogenous enzymes may be an effective treatment

motif

Despite advances over the past 25 years, an unmet demand for safe and effective

treatments for AMD and SMD persists. Most current and emerging therapies are based on

inhibition of the VEGF pathway, which does not adequately address the underlying

pathology. New approaches to treat AMD and SMD are necessary.

In applying the SENS paradigm, age-related macular degeneration is an example

of damage #7, intracellular aggregates, which drive dysfunction and eventual death of the

9

RPE, leading to progressive loss of vision. A SENS inspired therapy would aim to

eliminate or remove the offending retinoids from the RPE cells to restore function.

Because AMD and SMD are fundamentally caused by the accumulation of lipofuscin in

the lysosomes of RPE cells, both could be thought of as age-related lysosomal storage

diseases. Enzyme replacement therapies for inherited lysosomal storage diseases, such as

glucocerebrosidase in Gaucher’s disease and alpha galactosidase in Fabry’s disease, have

been well studied and widely successful in both laboratory and clinical settings43,44.

However, only one endogenous enzyme has been described that is capable of

catabolizing lipofuscin retinoid species, myeloperoxidase (MPO)45. MPO is a lysosomal

protein canonically associated with neutrophil granulocyte response to microbial insults,

during which it is released into the extracellular space and exerts antimicrobial activity by

forming hypochlorous acid. Translation of MPO as an enzyme therapy for AMD and

SMD failed due to unacceptable cytotoxicity, presumably secondary to hypochlorous

acid formation.

Although no viable enzyme candidates have been described for endogenous

enzyme therapy, over half a dozen exogenous enzymes, mostly fungal peroxidases, have

been described that are capable of efficiently degrading A2E into non-toxic metabolites46.

Several show activity in the absence of hydrogen peroxide as a co-factor and across a

relatively wide pH range, supporting their utility in applications requiring activity within

even a de-acidified lysosomal compartment. Traditional enzyme replacement therapies rely on lysosomal targeting by endocytosis through either mannose receptor or mannose-

6-phosphate receptor47. Human RPE cells express both in vivo48,49. Producing

recombinant enzymes within yeast strains such as Pichia pastoris is known to result in

10

high-mannose glycans at N-linked glycosylation sites, enabling mannose receptor-

mediated endocytosis50. Herein, I present work demonstrating for the first time,

feasibility of removing lipofuscin retinoid species from RPE in vivo by intermittent

administration of a recombinant manganese peroxidase.

Narrative Introduction

A distinguishing feature of my time as a graduate student is my parallel growth as

an entrepreneur. My responsibilities during my graduate work extended to the viability

and development of my host institution in addition to acquiring all research funding to

support my work. I founded Ichor Therapeutics in my apartment as a medical student at

SUNY Upstate Medical University. When I began my graduate work, Ichor Therapeutics

was a small four-employee start-up operating out of a derelict hair salon in rural

LaFayette, NY. While completing the thesis research presented here, I grew the company

to 13,000 ft2 of pristine commercial office and laboratory space, and now employee over

50 staff. I am the named principal investigator on 38 research awards totaling over $20

million and had the pleasure of giving 17 invited presentations both nationally and

internationally. Importantly, this progress was made with an academic eye, and was the

subject of several publications during my graduate work describing early efforts towards

building and validating infrastructure for Ichor Therapeutics, the first published in the

Journal of Commercial Biotechnology as, “Lean Start-up: A Comprehensive Case Study

in the Establishment of Affordable Laboratory Infrastructure”, and the second published

in MethodsX as, “Automating hESC differentiation with 3D printing and legacy liquid

handling solutions.” These papers are presented herein as a roadmap for other graduate

11 students who have similar aspirations to commercialize the research they are pursuing as part of their graduate educations, particularly as such research pertains to translational programs for age-associated disease.

12

Figure 1.1. Classes of age-associated damage. The 7 classes of age-associated damage proposed by Dr. Aubrey de Grey to drive human aging, which include, 1) cell loss and

tissue atrophy, 2) cancerous cells, 3) mitochondrial mutations, 4) death-resistant

(senescent) cells, 5) extracellular matrix stiffening, 6) extracellular aggregates, and 7)

intracellular aggregates.

13

Figure 1.2. SENS damage-repair approach. Most conventional therapies target metabolic processes or attempt to mitigate symptoms of aging damages. The SENS approach can be viewed as a periodic maintenance approach aimed to remove the offending damage directly.

14

Figure 1.3. Summary of the Visual Cycle (obtained from Molday and Moritz51).

Vitamin A in the 11-cis-retinal form is transported from the RPE to the photoreceptor, where it binds with opsin to form rhodopsin, and enable vision. After activation with light, rhodopsin dissociates and vitamin A in the trans-retinal form is modified and shuttled back to the RPE, where it can be recycled into an active 11-cis-retinal form once again.

CHAPTER 2: RECOMBINANT MANGANESE PEROXIDASE REDUCES A2E

BURDEN IN MOUSE MODEL OF STARGARDT’S DISEASE1

SUMMARY

Macular degeneration is hallmarked by retinal accumulation of toxic retinoid species (e.g.

A2E) for which there is no endogenous mechanism to eliminate. This ultimately results in

progressive dysfunction and loss of vision either in advanced age for genetically normal

patients (age-related macular degeneration), or in adolescence for those with inherited

genetic mutations (Stargardt's disease). Here, we present a proof-of-concept study for an enzyme-based therapy to remove these retinoids, modeled on traditional enzyme replacement therapy. Recombinant manganese peroxidase (rMnP) is produced in Pichia pastoris. In vitro, we demonstrate that rMnP breaks down A2E and other lipofuscin fluorophores with limited cellular toxicity, and as this enzyme is mannosylated, it can be taken up into cells via mannose receptor-dependent endocytosis. In vivo we demonstrate that rMnP can significantly reduce the A2E burden when administered by intravitreal injections. Together, these data provide encouraging results towards the development of an enzyme-based therapy for macular degeneration and indicate the need for additional work to characterize the molecular mechanism of A2E breakdown and to improve the pharmacological parameters of the enzyme.

1 This chapter is based on the previous published work: Moody, K. J. et al. Recombinant

Manganese Peroxidase Reduces A2E Burden in Age-Related and Stargardt’s Macular

Degeneration Models. Res 21, 560–571 (2018).

15

16

INTRODUCTORY REMARKS

Age related macular degeneration (AMD) is the leading cause of blindness in the US with over 2 million cases currently and projected increases to over 5 million by 2050 52.

Although the mechanism of the disease is not fully understood, it is hallmarked by the lysosomal accumulation of undegradable retinoid fluorophores (e.g. all-trans-retinal dimer (ATR-di), A2E, oxyA2E, isoA2E, etc.) in the retinal pigmented epithelium (RPE)

as a byproduct of imperfect recycling of 11-cis retinal during the visual cycle 53. These

fluorophores seem to perturb lipid metabolism and cause a buildup of pro-inflammatory,

lipid-containing granules called lipofuscin. Both AMD as well as the inherited juvenile

macular degeneration (prevalence ~1/8000), Stargart's disease (SGD) are associated with

lipofuscin accumulation 54,55. Presumably, these granules then cause inflammation 56,

complement activation 57,58, and other biological responses that ultimately result in the

deposition of extracellular drusen followed by degeneration of the photoreceptor layers in

the macula, resulting in geographic atrophy 59. In both AMD and SGD, subsequent to

this central retinal damage, vascular endothelial growth factor (VEGF)-dependent

angiogenesis can cause rapid choroidal neovascularization (CNV), resulting in rapid

deterioration of vision 59.

All current FDA-approved treatments for AMD focus on the second, "CNV" stage

of the disease after visual decline has already begun, primarily by interfering with VEGF

signaling 60. For Stargart's disease and AMD with geographic atrophy, interfering with

non-vascular targets, such as lipofuscin accumulation, might be required. ALK-001 is a

deuterated vitamin A analog that resists dimerization (a critical step informing many of

the lipofuscin fluorophores) and ultimately lipofuscin accumulation 61. It is currently in

17 phase 2 clinical trials for Stargart's disease (NCT02402660). However, a shortcoming of such a therapy is that it does not remove existing lipofuscin, only slows its deposition

61,62. Thus, it is unclear if such a treatment could reverse the disease or merely slow its progression, potentially limiting treatment utility.

Other groups have sought to develop methods to break down lipofuscin fluorophores within the lysosomes after they have already accumulated to potentially pathologic levels. Enzyme replacement therapy (ERT) for lysosomal storage diseases presents a robust clinical strategy for treating accumulation of toxic, undegradable metabolites in lysosomes 63. Specifically, a given missing enzyme required in the lysosomes is generated in appropriate cell lines, decorated with either mannose or mannose-6-phosphate (M6P), and then administered to the patient where it is targeted to the lysosomes via the mannose or M6P receptor endocytosis pathways. While there exists a plethora of ERT enzymes that use this strategy, for AMD and SGD there is an additional challenge in that there is no native enzyme that breaks down lipofuscin fluorophores, likely because they generally do not accumulate to toxic levels until patients are well into their 60's and 70's, long after nature could provide a solution in our evolutionary history 52.

Schloendorn 64 and Sparrow 65 previously described several fungal peroxidases capable of degrading A2E, the most prominent lipofuscin fluorophore, in cell-free assays.

Tested enzymes included manganese peroxidase (MnP), versatile peroxidase (VP), and horseradish peroxidase (HRP), as well as the non-peroxidase enzyme laccase (LAC). The subject of this paper, MnP, canonically uses heme chelated iron and hydrogen peroxide to create oxidized heme radicals and Mn(III) from Mn(II), which subsequently oxidize a

18

variety of substrates. The mechanism by which it breaks down A2E is currently

uncharacterized. Further development of these enzymes was abandoned when lysosomal

targeting via mannose receptor endocytosis could not be achieved in their cell model 66.

Using the human enzyme myeloperoxidase (MPO) decorated with M6P, Yogalingam et.

al. were able to degrade lysosomal A2E in culture, but were limited by cellular toxicity

67.

Here, we present a proof-of-concept study for the use of both wild-type MnP from

Phanerochaete chrysosporium (wtMnP) (which is natively mannosylated) and recombinant MnP produced in Pichia pastoris (rMnP) (which is hypermannosylated) to break down A2E and other lipofuscin fluorophores in cell-free systems, reduce A2E levels and associated toxicity in A2E-loaded cells, as well as reduce the A2E burden in

ABCA4-/- mice, a model of SGD. This work presents significant progress on the use of heterologous enzymes to degrade lipofuscin, but requires additional work to optimize the enzyme and its delivery.

RESULTS wtMnP and rMnP break down A2E in vitro

To establish the ability of wild-type manganese peroxidase from Phanerochaete chrysosporium (wtMnP) and recombinant manganese peroxidase produced in Pichia pastoris (rMnP) to break down A2E in vitro, wtMnP was obtained from commercial sources and rMnP was purified in-house. Both proteins migrated on SDS-PAGE with higher molecular weights than predicted based on their primary sequences and exhibited a molecular weight shift when treated with PNGase F, consistent with N-linked

19

glycosylation (Fig 1a). Additionally, rMnP migrated as a diffuse smear before PNGase

treatment and collapsed into a much tighter band when treated with PNGase F, consistent

with the N-linked hyper-mannosylation typical of proteins produced in Pichia. An analysis of N-linked glycan content of rMnP confirmed mannose as the primary sugar

with ~26 mannose monomers and ~1 N-acetylglucosamine monomer per enzyme on

average, confirming hypermannosylation relative to the wt enzyme (1 mannose

monomer/enzyme in crystal structure, PDB 3M5Q) (Fig. S1a). We additionally

performed O-linked glycan profiling and found ~19 glucose monomers and ~3 N-

acetylglucosamine monomers per enzyme on average, indicating O-linked

hyperglycosylation relative to the wt enzyme as well (Fig. S1b). Peroxidase activity of

both enzymes was confirmed by 2,6-dimethoxyphenol (DMP) colorimetric assay,

indicating that they are in a functional conformation (Fig. S2).

To test the ability of MnP to break down A2E, we incubated purified A2E with

wtMnP and monitored the disappearance of A2E absorbance at 440 nm (Fig. 1B) as well

as the A2E peak by HPLC (Fig. 1C). The longer wavelength 440 nm peak was chosen to

minimize interference with UV absorbing substances or potential scattering artifacts. As

expected, the protein was able to degrade A2E. Notably, the reaction was unaffected by

the presence or absence of H2O2, indicating a different mechanism from its canonical

peroxidase activity. Degradation of A2E by rMnP did, however, require oxygen, as the

reaction was completely inhibited in an argon-purged atmosphere, indicating an oxidative type of degradation (Fig. S3).

To evaluate the pH dependence of A2E degradation activity for the enzymes, the relative initial velocity of A2E breakdown was measured as a function of pH (Fig. 1D).

20

Non-canonical enzyme activity related to A2E breakdown was compared to typical peroxidase activity using DMP as the probe. While the pH optimum for wtMnP for the degradation of both A2E and DMP was below pH 6.0, the recombinant enzyme showed a substantial alkaline shift in pH optimum. The pH optimum for DMP activity shifted from~5.7 (wtMnP) to ~7.2 (rMnP), whereas the optimum pH for A2E degradation shifted from ≤4.5 (wtMnP) to a double maximum of ~4.7 and ~6.8 (rMnP). This alkaline shift in pH dependence was confirmed by an overnight degradation trial assayed by HPLC, where wtMnP showed maximal A2E degradation at ~4.8, and rMnP ≥7.2 (Fig. S4). This shift demonstrates a substantial effect of the altered glycosylation on the activity of the enzyme. Importantly, while pH optimum for wtMnP is acidic with little to no activity at

cytosolic or extracellular pH, rMnP activity was equal to that of wtMnP at lysosomal pH

but exhibits additional activity at cytosolic or extracellular pH.

We then measured the kcat and KM of the two enzyme preparations for the breakdown of A2E by measuring the initial rate of A2E disappearance as a function of

A2E concentration (Fig. 1E). The two enzymes had similar constants, with kcat and KM of

0.0042 min-1 and 12 μM for wtMnP, and 0.0060 min-1 and 44 μM for rMnP. However, the kinetic curves of the two enzymes exhibit important differences (Fig. S5). After an initial burst phase of several minutes, the wtMnP stabilized to a linear phase as expected and remained linear for hours. The rMnP, on the other hand, exhibits a burst phase as wtMnP, but instead of stabilizing to a linear phase would slowly lose activity. This slow loss in activity cannot be attributed to simple enzyme instability, as the enzyme was routinely left at 20 ˚C for several hours without noticeable loss in peroxidase activity, indicating a substrate-dependent activity loss. As a result, the ability of the recombinant

21

enzyme to break down A2E might be diminished relative to wt. Nevertheless,

recombinant MnP was chosen as the final candidate for the following reasons:

Recombinant MnP can break down A2E, is highly mannosylated which should facilitate

cellular uptake in our approach, can be reproducibly purified to homogeneity, and

behaved consistently in our biological assays (data not shown). In comparison, the wt

preparations varied widely from batch to batch in our biological assays (data not shown),

even from the same supplier, perhaps because of the difficulty in obtaining an

isotypically pure sample reproducibly from wild-type sources. Thus, despite its favorable

pH profile, sustainable linear phase kinetics, and overall higher activity in vitro, the wt

enzyme was not suitable for additional studies.

Lipofuscin substrate spectrum of rMnP

As lipofuscin is composed of several structurally related retinoids, the relative

ability of rMnP to degrade 7 of the most abundant retinoids was tested by incubating the

indicated substrate with varying concentrations of enzyme at 37 ˚C overnight and

quantifying the remaining substrate by HPLC relative to BSA control. Heat denatured

enzyme was not used as a control because rMnP would precipitate before losing activity

on heating, and chemically denatured enzyme was not used because rMnP would regain

significant activity upon removal of the denaturant. Recombinant MnP was able to break

down all of the tested lipofuscin fluorophores with EC50s ranging from 1.1 μM for all- trans-retinol dimer (ATRdi) to 20 μM for A2E (Fig. 2A). A2E was chosen as the endpoint biomarker for the biological work presented here because it is the highest abundance in lipofuscin, the most resistant to degradation, and the most widely studied

22

lipofuscin fluorophore 68. It should be noted that there is an open question regarding the

role of A2E in AMD, and its use here as a biomarker for AMD (in addition to SGD) is

not based on the assumption that it is causative in the disease, but that it is a reasonable

surrogate for the yet to be definitively identified causative lipofuscin component(s) 69.

Absence of mannose receptor in RPE cell models

The strategy for delivering rMnP to the lysosomes of RPE cells leverages mannose-receptor dependent endocytosis. Mannose receptors are robustly expressed on human RPE in vivo and the presence of mannose-specific receptors has been shown on the apical side of rat RPE cells 70. However, poor uptake was observed in previous

studies where enzyme delivery to RPE (ARPE-19) by mannose receptor endocytosis was

attempted 71. We hypothesized that the difficulty in achieving uptake into cells was a

result of deficiencies in mannose receptor expression in culture. Using flow cytometry,

we found that APRE-19 cultures did not express mannose receptors (CD206) (Fig. S6A),

and primary human RPE cells exhibited a passage-dependent downregulation of mannose

receptors (Fig. S6B). Please note that the commercial vendors supplying primary human

RPE typically split cells in culture for several passages before cryopreservation. We

found that most primary human RPE cells obtained through these sources do not express

mannose receptor upon arrival (4/5 vials tested), making them unsuitable for efficacy and

toxicity studies of a mannosylated enzyme.

23

rMnP uptake and degradation of A2E in cells

Wu et al. successfully reduced A2E concentration in pre-loaded ARPE-19 cells

with horseradish peroxidase (HRP) using the Bioporter system, a lipid-based reagent that

encapsulates proteins and non-specifically delivers them intracellularly 65. Absent a

readily obtainable RPE cell culture model with intact mannose receptor expression, we

chose this model to conduct our cell-based efficacy and toxicity studies.

Cell viability in the presence of rMnP with and without Bioporter at 24 h and 48 h was evaluated (Fig. 2B). The enzyme was well tolerated at concentrations up to ~100 µM, irrespective of the presence of Bioporter. In ARPE-19 cells that were pre-treated with 10

μM A2E, rMnP efficiently removed A2E when delivered using Bioporter (Fig. 2C).

Accordingly, rMnP treatment also prevented A2E-mediated cytotoxicity as determined by Annexin V / PI staining (Fig. 2D). Together, these results indicate that rMnP possesses limited toxicity and can clear A2E from cells if delivered intracellularly.

To confirm that rMnP can be internalized by mannose receptor-dependent endocytosis, we used the cell line RAW264.7, a transformed mouse macrophage cell line previously reported to maintain mannose receptors in culture 72. Briefly, we incubated the

cells in monolayer with either FITC-labeled rMnP or FITC-labeled mannosylated-BSA

(man-BSA) in the presence or absence of mannose-receptor competing ligand, mannan.

We then detached the cells at the indicated times, and measured fluorescence by flow cytometry (Fig. 2E). The fluorescence of the rMnP treated samples increased significantly over time, indicating that rMnP was being taken up by the cells. Uptake was similar to that of man-BSA, and uptake of both rMnP and man-BSA was inhibited by the presence of mannan, indicating that the uptake is dependent on the mannose receptor.

24

Intraocular half-life and efficacy of rMnP

To measure the vitreal half life of rMnP, we administered a single dose (112 μg)

to each eye of C57BL/6 mice by intravitreal injection and monitored the concentration by

direct ELISA of the vitreous (Fig. 3A). We found a first-order, single exponential decay

with a half life of 9.6 h. We then conducted a multi-dose efficacy study in the ABCA4-/-

mouse, a genetic model of Stargardt’s macular degeneration on a C57BL/6J background

that exhibits accelerated and elevated accumulation of A2E concomitant with slow retinal

degeneration 73. Aged ABCA4-/- mice (72 wks) were treated with 6 intravitreal doses of

sterile PBS or rMnP (24 μg eye-1 dose-1 or 112 μg eye-1 dose-1 ) spaced one week apart

(Fig. 3B). We harvested eyes 24 h after the final dose, extracted the A2E, and quantified

the levels by HPLC. There was a significant decrease (~31%) in the 112 μg group

relative to control after post-hoc analysis, and a trending decrease in the 24 μg group. To

capture the effect of an intermediate dose, we conducted a second study with PBS versus

56 μg eye-1 dose-1, and again found a significant decrease (Fig. S7). The combined data uncovered a linear decrease in A2E/eye as a function of dose, with a fit that allows us to estimate the effective A2E reduction of ~6 fmol μg-1 eye-1 dose-1 (Fig. 3C).

Toxicity evaluation of rMnP

Because MnP is heterologous to humans, we were concerned about immunotoxicity upon intraocular injection despite the eye being considered an immuneprivileged area 74. There was no significant weight loss in the 6-week efficacy

studies injecting animals with 24 or 112 μg rMnP that would indicate systemic

25

inflammation or other constitutional impairment (Fig. S8). After 6 intraocular injections

mice in the PBS group experienced cataracts (~52 %) and hemorrhage (~3 %). Mice

treated with rMnP experienced cataracts (p < 0.01) and microphthalmia (p < 0.001) at

elevated rates relative to control, but no additional hemorrhage (Table S1).

To evaluate functional toxicity to the eye, we performed a 12-week, bi-weekly

dosage study (6 x 56 μg dose) and conducted electroretinography (ERG) and histology

(Fig. S9, 3D). Sham-treated and PBS-treated mice showed significant reduction in ERG

amplitudes, indicating that the repeated intraocular injections themselves are functionally

damaging. However, the rMnP-treated animals were almost completely blind, indicating that the treatment caused significant additional functional damage to the retina.

Nevertheless, histology of eyes in the untreated, PBS, and rMnP group show appropriate architecture of the retinal layers with no significant alterations in the thickness of any individual layer or the retina as a whole (Fig. 3D,E). However, there with a significant number of immune cells in the vitreous of both the PBS and rMnP treated samples (Fig.

3D). To evaluate immunogenicity of rMnP, we longitudinally measured serum rMnP reactive antibody levels by ELISA in ABCA4-/- mice treated in parallel with the 112 μg

efficacy group. We found a repeat-dose dependent increase in anti-rMnP antibodies (Fig.

3F). We then performed a multi-dosage study, evaluating 24 μg dose-1 eye-1 and 112 μg

dose-1 eye-1 after 4 weekly doses. We found similar absolute levels to the longitudinal

study at 4 weeks, and no significant differences between the 24 and 112 μg groups,

indicating that the IgG response is not dependent on the amount of protein administered

in the concentration ranges used here (Fig. 3G). To evaluate if the circulating antibodies

make it into the eye, we repeated the 4-dose study with 112 μg dose-1 eye-1, and measured

26

antibody levels in the serum and in the vitreous. We found the expected increase in serum antibody levels, but failed to find a significant difference in the vitreous antibody levels

(Fig. 3H). However, it should be noted that the vitreal antibody levels tested are at or

very near the lower limit of quantification for the method used (~1 ng mL-1), so this negative should be interpreted with caution. Nonetheless, even if there is an increase in antibody in the vitreous below the lower limit of detection, the level is at least 2-orders of magnitude lower than that found in the serum. Together, these data indicate that while rMnP does impart some level of functional impairment in the retina of treated animals, it does not grossly disrupt the retinal layers. Further, these data do not differentiate between the plausible mechanisms of direct enzyme toxicity and local immune response

(discussed below).

DISCUSSION

Lipofuscin accumulation in the lysosomes of RPE cells and subsequent damage to the macula are hallmarks of AMD in humans. This paradigm is similar to traditionally understood lysosomal storage diseases, which are often treated by replacing a missing lysosomal enzyme with an exogenously expressed recombinant version, which is transported to the lysosomes via mannose or mannose-6-phosphate receptor mediated endocytosis. A key difference in the case of lipofuscin in the retina is that there is no endogenous human enzyme that can break down its major components. In this study, we present the first in vivo evidence demonstrating the breakdown of major lipofuscin fluorophores in the retina by administration of an exogenous, mannosylated enzyme.

27

Previous work by Schloendorn et. al. identified a series of peroxidases capable of

breaking down the highest-abundance retinal lipofuscin retinoid molecule, A2E 64. In

addition to A2E, we found that a recombinant manganese peroxidase produced in Pichia

pastoris (rMnP) efficiently breaks down the 7 most abundant lipofuscin retinoids, which

is advantageous considering the ongoing controversy regarding the identity of the

causative toxic lipofuscin species in AMD 69. Notably, it seems to do so through a

mechanism different from its peroxide function, possessing a different pH profile and

lacking a requirement for peroxide. Other groups have described H2O2-independent, O2-

dependent peroxide mechanisms previously, and studies to elucidate this mechanism are

ongoing in our lab 75.

In cell culture, rMnP efficiently clears A2E and does so without apparent cellular

toxicity at concentrations up to ~100 μM, orders of magnitude less toxic than previous

attempts to clear A2E from cells using myeloperoxidase and mannose-6-phosphate

targeting 67. Uptake of rMnP can be inhibited by mannan, indicating that it occurs through mannose receptor-mediated endocytosis, a lysosomal targeting pathway conserved in RPE of human and mouse 76–79.

When administered intravitreally, rMnP reduces A2E burden by 31% in 6 weeks.

This compares favorably to existing approaches e.g. ALK-001, a drug in phase II trials that reduced lipofuscin burden in mice by 50% in 9 months 61. An important distinction is

that rMnP breaks down existing lipofuscin, rather than current approaches aimed at

reducing the rate of new A2E formation. In theory, this could extend the window to begin

treatment to after the disease has presented itself clinically (which is current clinical practice), rather than requiring the development a screening test to identify at-risk

28 individuals before they have symptoms to be effective. The vitreal half-life of rMnP is

9.6 hours, which is considerably shorter than standard of care antibodies ranibizumab

(2.51 days) and bevacizumab (6.99 days), suggesting rapid internalization by target cells

80.

Using a therapeutic with a short half-life to break down a target like retinoids in lipofuscin, which continually build up over a lifetime, may seem counterintuitive.

However, because the approach can clear existing toxic retinoids, and the accumulation of those retinoids is slow, at least for AMD, one treatment course (or periodic treatment decades apart) will likely be sufficient to stop or reverse the disease. That said, a longer acting formulation may be beneficial for the juvenile SGD, in which A2E accumulates more rapidly. In that case, protein engineering or even encapsulation may improve the efficacy and extend the treatment interval.

While intravitreal injection of rMnP does not seem to induce systemic toxicity based on the lack of weight loss, it does seem to lead to function loss as evidenced by the

ERG results. Importantly, this did not seem to result from destruction of the photoreceptor layer or overall retinal architecture, and is additionally complicated by the large degree of functional impairment by virtue of intraocular injection in the mouse eyes using standard equipment and techniques. This suggests that larger animals may be a more suitable model system for this type of treatment, and additional follow-up studies are warranted to deconvolute damage resulting from the procedure itself and damage due to rMnP. That said, there are significant differences between rMnP and control groups with regard to both efficacy and toxicity using the same techniques, which argues strongly for effects mediated by the enzyme itself.

29

While the current mechanism of toxicity related to rMnP treatment is unclear, the

two most likely candidates are immune-mediated and directly from the enzyme. Repeat

dosing led to a detectable systemic IgG response, presumably due to leakage of rMnP

into the systemic circulation. While no anti-rMnP IgG were detected in the vitreous, this

does not necessarily preclude the possibility of IgG-mediated ocular damage. IgGs can directly activate the complement system via the classical or lectin pathway leading to complement-dependent cytotoxicity. Or, IgGs can bind to their specific receptors (Fc-

receptors; FcR) to trigger antibody-dependent cell-mediated cytotoxicity. Both of these mechanisms could impair RPE function from the choroidal side, indirectly impacting photoreceptor cell health. Looking for downstream activation of these pathways may be able to provide clearer answers. Additionally, there are a significant number of immune cells in the vitreous of eyes of treated animals, suggesting perhaps a cell-mediated immune response. However, this is complicated by the presence of immune cells in PBS treated animals as well, suggesting the physical treatment is a major contributor to the response. Additional mechanistic work in larger animals in which intraocular injections are easier to perform, will be required to adequately investigate the immune response.

Compared to its wild-type counterpart, rMnP has a glycan mediated alkaline shift in pH

optimum and high activity at physiologic pH. It stands to reason that this would result in

activity and potentially damage cells and tissues outside the lysosome. Additionally,

because our results demonstrate that the enzyme requires O2, it is possible that the enzyme could cause oxidative damage to sensitive neural or photoreceptor cells, thus diminishing vision. Hypermannosylation also somewhat diminishes the A2E degradation activity of rMnP at lysosomal pH, thus increasing the amount of enzyme that must be

30

administered for equivalent efficacy, which could amplify the toxicities we see. It is

possible that glycoengineering could restore wild-type pH profile, attenuating repeat dose

toxicity and restoring wt-like kinetics and lysosomal pH activity, greatly enhancing in

vivo efficacy. Because glycans are also critically involved in immune system recognition,

humanizing the glycosylation of the enzyme could also assist with any immune-related

toxicities.

Together, our results establish clear proof-of-concept for the use of heterologous enzymes to degrade RPE lipofuscin following an enzyme replacement therapy approach.

However, additional work is required to maximize efficacy and definitively identify the causes of repeat dose toxicity and to attenuate them.

METHODS

Reagents

Wild-type manganese peroxidase (wtMnP) was obtained from Jena Bioscience

(Jena, Germany) (cat # JBEN201L). P. pastoris strain with a genomically integrated recombinant manganese peroxidase (rMnP) expression construct was a generous gift from the Christine Kelly group at Syracuse University (Syracuse, NY). A2E and other lipofuscin fluorophores were synthesized in-house as previously described with minor modifications and purified by HPLC (see Supplemental Methods) 81–84. Bioporter protein

delivery reagent was purchased from Gelantis (San Diego, CA). D-mannose-BSA (14 atom spacer) was purchased from Dextra Laboratories, Ltd. (Reading, UK). Zeocin was purchased from InvivoGen (San Diego, CA). CellTiter-Blue viability kit was purchased from Promega Corporation (Madison WI). PTM1 salts were purchased from Sunrise

31

Science Products, Inc. (San Diego, CA), HRP-conjugated polyclonal rabbit-anti-6xHis antibody was purchased from Abcam (ab197049) (Cambridge, MA). BCA kit was purchased from Gene and Cell Technologies (Vallejo, CA). LAL endotoxin test kit was purchased from GenScript (Piscataway, NJ). PNGase F was purchased from New

England Biolabs (Ipswitch, MA). Cell culture media, fetal bovine serum (FBS), phosphate buffered saline (PBS), plates, and other standard cell culture reagents were purchased from Corning (Corning, NY). Annexin-V/PI Apoptosis Detection Kit was purchased from BioLegend (San Diego, CA). Flow cytometry staining buffer, antibodies, and other reagents for flow cytometry were purchased ThermoFisher Scientific

(Waltham, MA). Cell lines were purchased from ATCC (Manassas, VA). Primary cells were purchased from Lonza (Basel, Switzerland). Protein chromatography resins were purchased from GE Healthcare Bio-Sciences (Marlborough, MA), except for Ni-NTA agarose, which was purchased from Qiagen (Valencia, CA). HPLC columns were purchased from Waters Corporation (Milford, MA). 1-Step™ Ultra TMB-ELISA

Substrate Solution, SuperSignal ELISA Femto Substrate, HRP conjugated mouse-anti- myc (clone 9E10), unconjugated mouse-anti-6xHis (4E3D10H2/E3), and HRP- conjugated goat-anti-mouse were purchased from Thermo Fisher Scientific (Waltham,

MA). All other chemicals were purchased from Millipore Sigma (Burlington, MA).

rMnP expression

The coding sequence for manganese peroxidase from Phanerochaete chrysosporium lacking its n-terminal secretion signal tag was cloned into pGAPZα-A in- frame with the n-terminal α-factor secretion sequence and c-terminal myc and 6xHis tags,

32

and genomically integrated into P. Pastoris. The resultant strain was maintained at 25 ˚C

by weekly restreaking on YPDS plates (65 g/L YPD agar, 1 M D-sorbitol, 100 μg/mL

Zeocin). For protein expression, a 1 L starter culture in a 2.8 L baffled flask (YPD + 100

μg/mL zeocin + 1 drop antifoam 204) was inoculated with 5 isolated colonies and grown for 24 h in a dark shaking incubator at 250 RPM at 30 °C. Starter cultures were pelleted by centrifugation in sterile containers, resuspended in a small volume of fresh YPD, and transferred to a 14 L Eppendorf bioreactor with 8 L of the following minimal media: phosphoric acid - 26.7 ml/L, calcium sulfate dihydrate - 0.93 g/L, potassium sulfate -

18.2 g/L, magnesium sulfate heptahydrate - 14.9 g/L, potassium hydroxide - 4.13 g/L, glucose - 40 g/L (1 L of 320 g/L of glucose was autoclaved separately and added to the reactor), antifoam 204 (3 mL), zeocin - 100 mg/L (0.2 μm filtered into the bioreactor before inoculation), hemin - 0.1 g/L (autoclaved separately). PTM1 trace salts were prepared according to the manufacturer's instructions, and 4.38 mL/L 0.2 μm filtered and added to the bioreactor. Before inoculation, the bioreactor was covered with aluminum foil (to prevent the degradation of zeocin), brought to pH 6.0 with 30% ammonium hydroxide, and brought to 30 °C. Temperature, dissolved oxygen, and pH were tracked during growth. Agitation was maintained at 800-1200 RPM (DO stat >20%), temperature at 30 °C, pH at 6.0, and aeration maintained at 10-20 SLPM (DO stat >20%). A solution of 30% ammonium hydroxide was used to maintain pH through the base pump. When the

DO spiked at the end of the batch phase (~24 h after inoculation), 50 % glucose + 12 mg/L PTM1 trace salts was added at a rate of 0.5-1 mL/min (adjusted down to keep DO >

20 %) and the culture grown for an additional 12 hours.

33

Protein purification

Yeast cultures were pelleted by centrifugation and the supernatant pH adjusted to

8.0 with NaOH. The sample was then purified by single-step NiNTA chromatography

(Equilibration/wash buffer 50 mM NaPO4 pH 8.0, 300 mM NaCl, elution buffer 200 mM

NaOAc pH 4.6, 300 mM NaCl), dialyzed exhaustively against 50 mM NaPO4 pH 7.4,

150 mM NaCl, sterile 0.2 μm filtered, and concentrated in ethanol sterilized gamma- irradiated spin concentrators. Concentration was determined by BCA assay with BSA standard, and protein flash frozen at -80 ˚C until use. Endotoxin content of purified protein was 0.01-0.1 EU/mg by LAL gel clot test run according to the manufacturers instructions.

Glycoanalysis

For gel shift, samples were treated with PNGase F using the denaturing protocol

according to the manufacturer's instructions and run on SDS-PAGE. Monosaccharide

analysis was performed by the UC San Diego GlycoAnalytics Core. For N-linked glycan

analysis, briefly, a 100 μL sample of 1 mM rMnP was flash-frozen in liquid nitrogen and

sent to GlycoAnalytics (UC San Diego School of Medicine). The sample was processed

in two sequential steps: 1) Preparation of N-glycans: N-linked glycan was removed using

recombinant PNGase F. The released N-glycans were applied to a tandemly connected

Sep-Pak C18 and porous graphitic carbon (PGC) cartridge equilibrated in water. PGC

bound glycans were eluted with 30% acetonitrile containing 0.1% TFA in water,

lyophilized, and used for further analysis in the second step. 2) Monosaccharide analysis

by HPAEC-PAD: Previously-purified sample (2.5 µg) was treated with 100 μL of 2 M

34

TFA at 100 °C for 4 h, followed by removal of excess acid by dry nitrogen flush. To

ensure complete removal of acid, the samples were co-evaporated twice with

isopropanol. Finally, the samples were dissolved in water and analyzed by HPAEC-PAD

(High-Performance Anion-Exchange Chromatography – Pulsed Amperometric

Detection) using a CarboPac PA-1 column (250 mmx 4 mm).

For O-linked glycans, 50 µg of glycoprotein sample was used for monosaccharide

analysis. Briefly, the sample was hydrolyzed using 2 N TFA at 100 °C for 4 h followed

by removal of excess acid using dry nitrogen flush. The sample was then dissolved in

known amount of Milli-Q water and composition analysis was done using HPAEC-PAD

(Dionex-Thermo ICS-3000). CarboPac PA-1 column (4 mm x 250 mm) was used to

profile the monosaccharide and gradient of 100 mM NaOH and 250 mM of NaOAc were

used for chromatography. The area and retention time of the monosaccharides were

reported by comparing with authentic standards.

Lipofuscin fluorophore degradation assay

A2E degradation was monitored by both HPLC and UV-Vis spectrophotometry.

For HPLC, A2E in the indicated buffer (50 mM NaOAc, 50 mM NaPO4, 150 mM NaCl,

2% DMSO, 0.2% TritonX-100, pH 5.5 unless otherwise stated) was incubated at 37 ˚C for 4 h with continuous shaking with the indicated concentration of enzyme or BSA control, and the remaining A2E measured by HPLC peak integration. HPLC conditions are detailed below for A2E and in Supplemental Methods for other lipofuscin fluorophores. Degradation of all other lipofuscin fluorophores was measured using this method. For UV-Vis, A2E was incubated with enzyme in the same buffer in quartz

35

cuvettes or clear-bottomed 96-well plates, and the A2E concentration monitored by

disappearance of absorbance at 440 nm over time. Activity was determined by linear fits

4 of the initial kinetic curves. A2E concentration was determined using ε440 nm = 2.2 x 10

M-1cm-1. Absorbance measurements were taken on a Thermo Fisher Scientific (Waltham,

MA) Evolution 201 UV-Vis spectrophotometer, or a Molecular Devices (San Jose, CA)

SpectraMax i3 plate reader with similar results.

DMP activity assay

2,6-Dimethoxyphenol (DMP) assay was conducted as previously described with

minor modifications 85. Briefly, the indicated concentration of enzyme was incubated

with the indicated buffer (200 mM NaOAc, 0.4 mM MnSO4, 0.005% Tween20, 1 mM

DMP, pH 4.6 unless otherwise noted) in clear 96 well plates and the reaction initiated

-4 with 9 x 10 % H2O2. Absorbance was monitored at 469 nm in a Molecular Devices (San

Jose, CA) Spectramax i3, and activity determined by linear fits of the initial kinetic

curves.

Cell culture

ARPE-19 and RAW 264.7 were maintained on tissue culture treated vessels in

DMEM containing 10 % FBS, 100 U/mL penicillin, 100 µg/mL streptomycin, and 0.25

µg/mL amphotericin B. Cells were grown in 5% CO2 at 37 °C with >80% relative

humidity. Experiments were performed using 3-7 day post-confluent ARPE-19 (passage

<40) and 75-90% confluent RAW 264.7 (passage <15). Sub-culturing was performed using trypsin/EDTA solution as recommended by ATCC.

36

Primary human RPE cells were maintained on tissue culture treated vessels in

RtEGMTM Retinal Pigmented Epithelial Cell Growth Medium (Lonza, Basel Switzerland) containing 100 U/mL penicillin, 100 µg/mL streptomycin, and 0.25 µg/mL amphotericin

B. For initial thawing, media was also supplemented with 2 % FBS. Cells were grown in

5% CO2 at 37°C with >80% relative humidity. Experiments were performed using 80-

90% confluent cells at the passage indicated.

A2E extraction and quantification

A2E Extraction from eye tissue – 4 eyes were extracted using a modified version of a previously reported extraction method from age matched mice post treatment with rMnP 86. Eye sets were added to a glass tissue grinder with 1 mL of a 2:1 v/v mixture of chloroform:methanol and homogenized over 15 minutes. The homogenate was moved to a 15 mL Falcon tube and the homogenizer washed with 3 mL of ddH2O, which was then added to the 15 mL falcon tube containing the eye homogenate. Samples were vortexed for 1 min, centrifuged at 3000 x g for 20 min, after which the lower organic phase was removed and added to a 1.5 mL microcentrifuge tube. 2 mL of a 1:1 v/v mixture of dichloromethane:methanol was then added to the eye homogenate and the samples were vortexed for 1 min. Samples were then centrifuged at 3000 x g for 20 min and the lower organic phase was removed and added to the same 1.5 mL microcentrifuge tube. Samples were evaporated to dryness under vacuum at 37 ˚C. The residue was then resuspended in

100 μL of methanol. 30 μL of this sample were then injected into a Waters 2695 solvent manager HPLC system with an attached 2996 photodiode array detector and separated as detailed in below.

37

A2E Extraction from ARPE-19 cells – Cells were pelleted in 2 mL microcentrifuge tubes by centrifugation and resuspended in 325 μL ddH2O and 830 μL

methanol:glacial acetic acid (98:2 v/v), by brief vortexing. Each sample was left to

incubate for 5 min before adding 375 μL dichloromethane. Samples were mixed well by

vortexing and centrifuged on a tabletop centrifuge for 5 min at max speed. The

dichloromethane layer was then removed to a new microcentrifuge tube, and the sample

was extracted a second time with 410 μL dichloromethane. Extractions were pooled and

evaporated under vacuum at 37 ˚C. Samples were then resuspended in 200 μL methanol, and 10 μL injected into the HPLC.

A2E HPLC method

Samples were analyzed using a Cosmosil AR-II 5C18 4.6 x 150 mm 5 μm particle

size column at a flow rate of 1.00 mL/min using the following mobile phase composition:

A – Water + 0.1% TFA, B – Methanol + 0.1% TFA. For A2E samples extracted from eye

tissue, a gradient method was performed such that the initial conditions were set at 20%

A/80% B, these conditions were held for 4 min before starting a gradient to 100% B over

15 min, which were then held for 4 min. The elution time of A2E under these conditions

was ~16.4 min. Monitoring was performed at 430 nm. For analysis of A2E from cell

culture or in cell free assays, separation was performed isocratically using 10% A/90% B.

Cell viability assay

Cell viability was assessed using CellTiter-Blue according to manufacturer instructions. Briefly, CellTiter-Blue Reagent was added to cell media and mixed by

38

repeat pipetting. Samples were then incubated at 37 °C for 1-4 hours and read by

fluorescence λex/ λem = 560 nm/590 nm using a SpectraMax i3 microplate reader

(Molecular Devices, San Jose, CA).

Annexin V / PI

Apoptosis/necrosis determination was performed using Annexin-V/PI Apoptosis

Detection Kit according to manufacturer instructions. Briefly, cell samples were washed

with PBS and resuspended in either PBS or Annexin V binding buffer. 5 µL of FITC-

Annexin V and 10 µl propidium iodide solution was added to each sample (~105 cells/sample). Samples were incubated for 15 min at room temperature in the dark then analyzed by flow cytometry using an Accuri C6 flow cytometer (BD Biosciences, San

Jose, CA).

Mannose receptor (CD206)

Cells were washed in PBS and resuspended in PBS or Flow Cytometry Staining

Buffer. 5 µl Fc receptor binding inhibitor antibody and 5 µl (125 ng) anti-human

CD206(19.2)-PE or mouse IgG1 kappa isotype control-PE was added to each sample

(~105 cells/sample). Samples were incubated for 15-30 min at room temperature in the

dark, washed three times with PBS, and analyzed on an Accuri C6 flow cytometer (BD

Biosciences, San Jose, CA).

39

Bioporter-delivered A2E breakdown assay

A2E was dissolved in DMSO to create a 10 μM stock solution. BioPORTER

Protein Delivery Reagent was prepared according to manufacturer’s instructions. Briefly,

BioPORTER Reagent was dissolved with methanol and an appropriate volume transferred to an Eppendorf tube and evaporated. rMnP was diluted in PBS was added to hydrate the dried BioPORTER Reagent, and finally serum free media was added to obtain a final concentration of 50 μM rMnP.

A2E solution was added to ARPE-19 cells at a final concentration of 10 µM and incubated for 24 h in the dark. Cells were then washed five times with PBS and

BioPORTER/rMnP complex was added. Cells with complex were incubated for 24 h in

5% CO2 at 37 °C with >80% relative humidity in the dark. A2E was extracted then

measured by HPLC as described above.

Mannan competitive uptake assay

FITC labeled samples were prepared by adding 5 μL of 100 mM FITC to the cap

of a 1.5 mL microcentrifuge tube containing either 100 μL 1 mM rMnP or 10 mg/mL D-

mannose-BSA, vortexing gently, then incubating for 2 hours in the dark at room

temperature. FITC labeled samples were then desalted 2x using a PD10 rapid desalting

column into PBS with calcium and magnesium. FITC and BSA concentrations were

determined by measuring absorbance at 494nm and 280nm and extinction coefficients of

75,000 M-1cm-1 and 48,000 M-1cm-1, respectively with a correction factor of 0.3 for FITC

A280 nm/A494 nm. rMnP concentration was measured using BCA assay as described.

40

Mannan powder was dissolved in DPBS containing calcium and magnesium to create a

49 mM stock solution and sterile filtered.

To measure uptake, media was removed from all cells in a confluent 24-well

plate. Cells were treated with DPBS with calcium and magnesium containing 1 μM BSA-

FITC or rMnP-FITC with and without 4.9 mM mannan. At timepoints 0, 30, 60, 90, 120,

and 180 min, wells were washed gently in PBS three times then detached by vigorous

pipetting. 25,000 events were collected by flow cytometry, and the mean FITC

fluorescence measured.

ELISA

For direct ELISA of rMnP in the vitreous, vitreal fluid pulled from C57/BL6 mice

at the indicated time point after treatment was diluted 1:100 in PBS and adsorbed to

untreated clear 96-well well plates (100 μL) overnight. Samples were washed 4 x in PBS

(200 μL) and blocked for 1 h in PBS + 1% BSA + 0.05% Tween 20 (200 μL), washed 4 x

in PBS, incubated with 2 μg/mL HRP conjugated mouse-anti-myc (clone 9E10) for 2 h in blocking buffer, washed 4x in PBS, and developed with TMB. Absorbance values were normalized to their maximum value and the value from naive mouse vitreous, and fit with a single exponential decay. For baited ELISA for measuring circulating IgG titers, untreated white polystyrene plates were incubated with 1 μM rMnP in PBS (100 μL) overnight at 4 °C, washed 4 x in PBS, blocked for 2 h in PBS + 1% BSA + 0.05% Tween

20 (200 μL), washed 4 x in PBS, incubated with 100 μL serum or vitreous from mice

(1:100 dilution) for 1 h, washed washed 4 x in PBS, incubated with 20 ng/mL HRP- conjugated goat-anti-mouse for 1 h (100 μL), washed 4 x in PBS, and exposed with

41

SuperSignal ELISA Femto Substrate. Luminescence was measured after 1 min in a

SpectraMax i3 by collecting all wavelengths in luminescence mode. Sample

concentrations were calculated by normalizing to a standard curve generated with

unconjugated mouse-anti-6xHis (4E3D10H2/E3) fit with a 4-parameter logistic

regression. All ELISA steps were conducted at room temperature.

Husbandry

Mice were purchased from The Jackson Laboratory (Sacramento, CA). Animal studies were approved by the Ichor Therapeutics Institutional Animal Care and Use

Committee for studies conducted at Ichor exclusively, or both the Ichor and Medical

University of South Carolina Institutional Animal Care and Use Committee. Animals

were maintained in a specific pathogen free facility with a 12-hour light/dark schedule.

Standard mouse chow and chlorinated RO water were provided ad libitum. Animals were

monitored daily.

Unless otherwise indicated, study animals received intravitreal injections in both

eyes with the indicated treatment or saline for control under sedation with isoflurane. For

intravitreal injection, Animals were placed and maintained under inhalant anesthesia with

isoflurane anesthetic. Proparacaine was placed on each eye for local analgesia. The sclera

was then perforated with a 29 G needle just posteriorly to the limbus. 2 μL (or 1 μL in the case of the toxicity study) of enzyme or saline was then introduced with a 34G Nanofil syringe. Triple antibiotic ophthalmic ointment was placed on the eye. Animals were also given a single subcutaneous injection of meloxicam for generalized pain control at the

time of administration (5 mg/kg). Animals then recovered from anesthesia. As needed by

42

the study, animals were bled by tail bleed and humanely euthanized at designated points

by CO2 inhalation and cervical dislocation.

Electroretinography

Electroretinography (ERG) recordings were performed as previously described 87.

In short, mice were dark-adapted overnight and anesthetized with xylazine and ketamine

(20 and 80 mg/kg). Pupils were dilated with phenylephrine HCL (2.5%) and atropine

(1%). ERGs were recorded with the UTAS-2000 (LKC Technologies, Inc., Gaithersburg,

MD) system, using a Grass strobe-flash stimulus. Stimuli consisted of 10 µs single- flashes at a fixed intensity (2.48 cd*s/m2) under scotopic conditions. ERG measurements were performed before (baseline ERG) laser photocoagulation, and afterwards on day 6.

A-wave amplitudes were measured from baseline to the a-wave trough, whereas b-wave amplitudes were measured from the a-wave trough to the peak of the b-wave.

Histology

After ERG analysis, eyes were harvested, embedded in cryoembedding compound, frozen and stored at -80 °C until sectioned. Prior to sectioning, embedded

eye blocks were transferred to a pre-cooled cryotome at -20°C and allowed to acclimate for at least 20 min. Blocks were affixed to specimen holders with OCT Compound and trimmed until the area of interest was completely exposed. 5 µm thick sections were cut, mounted to poly-L-lysine coated slides, fixed in 95% EtOH for 30 s, and washed in PBS for ~5 min. Eye sections were then stained in hematoxylin for 5 min, washed in DI water, incubated in 1% MgSO4 and 0.067% NaHCO3 bluing reagent for 30 s, washed in DI

43 water, stained in eosin for 15 s, and washed in DI water. Sections were then dehydrated through 95% EtOH, two changes of 100% EtOH, and cleared through two changes of xylene substitute before adding Organo/Limonene Mounting Medium, coverslipping, and sealing.

SUPPLEMENTAL METHODS

Synthesis of lipofuscin components

A2PE & NRPE – A2PE and NRPE were synthesized as described previously 83. A mixture of 78 mg all trans retinal (ATR) and 94 mg 1,2-Dipalmitoyl-sn-glycero-3- phosphoethanolamine in 3.4 mL chloroform was stirred for 72 hours in the dark at 37 ˚C in the presence of catalytic amounts of glacial acetic acid (8 μL). Solvent was removed under vacuum and the sample was purified via silica chromatography using a chloroform/methanol + Trifluoracetic acid (86%/14% + 0.1% TFA) mobile phase followed by further purification via HPLC. Chromatographic conditions used a 4.6 x 150 mm cosmosil AR-II 5C4 5 μm particle size analytical column performed at 1 mL min-1 via a gradient of ddH2O + 0.1% TFA (Solvent A) and methanol + 0.1% TFA (Solvent B).

The initial conditions of the gradient were (20% A/ 80% B), these conditions were held for 4 minutes, followed by a 15 minute gradient to 100% B which was held for 4 minutes at 100% B. Detection was performed using a Waters 2996 photodiode array detector at

440 nm, the chromatography system used was a Waters 2695 solvent manager system.

A2E/ isoA2E - A2E and iso-A2E were synthesized as described previously 82. A mixture of 195 mg ATR and 21 μL ethanolamine in 7 mL ethanol was stirred for 72 hours in the dark in the presence of catalytic amounts of glacial acetic acid (18 μL). Solvent was

44 removed under vacuum and the sample was purified via silica chromatography using a chloroform/methanol + Trifluoracetic acid (92%/8% + 0.1% TFA) mobile phase followed by further purification via HPLC. Chromatographic conditions used a 10 x 250 mm cosmosil AR-II 5C18 5 um particle size semi preparative column performed at 6.0 mL min-1 isocratically using a mobile phase composed of 12% Water + 0.1% TFA: 88%

Methanol + 0.1% TFA. Detection was performed using a Waters 2996 photodiode array detector at 440 nm, the chromatography system used was a Waters 2695 solvent manager system.

ATRdi – All trans Retinal dimer (ATRdi) was synthesized by the method of Peter J. E.

Verdegem et. al. 88. 255 mg ATR in Tetrahydrofuran was stirred in the presence of 30 mg

NaOH for 3 h. The sample was then quenched by dropwise addition of saturated ammonium chloride. The sample was then extracted 3 times with diethylether, and washed with saturated sodium chloride. The pooled samples were then dried with magnesium sulfate and gravity filtered. Samples were dried under vacuum and purified via silica gel chromatography using a diethylether/hexane (20%/80%) mobile phase followed by further purification via HPLC. Chromatographic conditions used a 4.6 x150 mm cosmosil AR-II 5C18 5 μm particle size analytical column performed at 1.0 mL min-

1 isocratically using a mobile phase composed of 100% acetonitrile. Detection was performed using a Waters 2996 photodiode array detector at 440 nm, the chromatography system used was a Waters 2695 solvent manager system.

45

Oxy-A2E – Oxy-A2E was synthesized by exposing a sample of purified A2E in methanol to long-wavelength UV-light (4W 365 nm handheld fluorescent transilluminator) for 1 h, followed by purification by HPLC. Chromatographic conditions used a 4.6 x 150mm cosmosil AR-II 5C18 5 μm particle size analytical column

-1 performed at 1 mL min via a gradient of ddH2O + 0.1% TFA (Solvent A) and methanol

+ 0.1% TFA (Solvent B). The initial conditions of the gradient were (40% A/ 60% B), these conditions were held for 4 min, followed by a 25 min gradient to 100% B which was held for 4 minutes at 100% B. Detection was performed using a Waters 2996 photodiode array detector at 440 nm, the chromatography system used was a Waters 2695 solvent manager system.

Identity confirmation of synthesized compounds

HPLC chromatograms and UV-Vis absorbance spectra shown in Figure S10.

A2E(N-Retinylidene-N-retinylethanolamine). Prepared following the synthesis and purification procedure listed. H NMR (MeOD, 600 MHz): δ 8.53 (d, J = 6.8 Hz, 1H),

7.99 (dd, J = 15.2, 11.5 Hz, 1H), 7.93 (dd, J = 6.8, 2.0 Hz, 1H), 7.86 (d, J = 1.9 Hz, 1H),

7.13 (dd, J = 15.0, 11.4 Hz, 1H), 6.76 (d, J = 15.1 Hz, 1H), 6.69 (s, 1H), 6.61 (d, J = 15.2

Hz, 1H), 6.55 (d, J = 16.1 Hz, 1H), 6.42 (d, J = 11.5 Hz, 1H), 6.35 (d, J = 15.9 Hz, 1H),

6.28 (d, J = 16.1 Hz, 1H), 6.25 (d, J = 11.3 Hz, 1H), 6.18 (d, J = 16.1 Hz, 1H). RMS

(ESI+) calculated for C42H58NO: 592.4518, found: 592.4502.

46

ATR-dimer. Prepared following the synthesis and purification procedure listed. H NMR

(CDCl3, 600 MHz): δ 9.46 (s, 1H), 6.96 (dd, J = 15.0, 11.5 Hz, 1H), 6.81 (d, J = 6.1 Hz,

1H), 6.41 (d, J = 15.1 Hz, 1H), 6.35 (dd, J = 15.3, 11.1 Hz, 1H), 6.30 (d, J = 16.0 Hz,

1H), 6.19 (d, J = 11.7 Hz, 1H), 6.16 (m, 2H), 6.08 (d, J = 15.9 Hz, 1H), 6.02 (d, J = 15.9

13 Hz, 1H), 5.99 (d, J = 10.2 Hz, 1H), 5.85 (d, J = 15.2 Hz, 1H). C NMR (CDCl3, 151

MHz) δ 192.4, 144.6, 144.3, 141.9, 139.9, 138.6, 138.3, 138.2, 138.1, 137.7, 135.4,

133.0, 130.6, 130.5, 130.1, 129.7, 129.2, 129.1, 126.7, 124.2, 123.0. HRMS (ESI+)

+ calculated for C40H54O, H : 551.4247, found: 551.4237.

A2PE. Prepared following the synthesis and purification procedure listed. (ESI+)

calculated for C77H125NO8P: 1222.9142, found: 1222.9151.

NRPE. Prepared following the synthesis and purification procedure listed. RMS (ESI+)

calculated for C57H101NO8P: 958.7264, found: 958.7222

Iso-A2E. Prepared following the synthesis and purification procedure listed. RMS

(ESI+) calculated for C42H58NO: 592.4513, found: 592.4505

Effect of O2 on rMnP breakdown of A2E

A volume of 50 mM sodium acetate/50 mM sodium phosphate/150 mM NaCl, 0.2% v/v

Tween 20 buffer (pH 5.5), was transferred into a small round bottom flask and sonicated for 60 min. The flask was then wrapped in foil and rMnP was transferred at concentrations of 0, 4, 8, or 12 µM. The buffer, peroxidase mixture was then degassed

47

using three iterations of freeze-pump-thaw under argon gas on a Schleck-line. A2E in

100% DMSO at a concentration of 835 µM degassed solution was then added to the round bottom flask via syringe injection to a final concentration of 45 µM. 800 µL of the prepared solution was then rapidly transferred via syringe to a septa-sealed quartz cuvette and electronic absorption spectroscopy was performed between 225-800 nm over 60 min, with a scan every min. The maximum absorbance of A2E at 434 nm was used to follow oxidation. Control experiments were conducted by performing the experiment as described above without degassing, using an argon environment or using a sealed cuvette.

48

Figure 2.1. wtMnP and rMnP activity in vitro. A) SDS-PAGE of wtMnP and rMnP with and without PNGase F treatment. B) Representative kinetic A2E degradation traces by wtMnP in the presence and absence of H2O2. [A2E] = 44 μM, [wtMnP] = 1 μM,

[H2O2] = 0.001%, room temperature. C) HPLC traces of A2E in the presence and absence of wtMnP after overnight incubation. [A2E] = 44 μM, [wtMnP] = 1 μM D)

Normalized reaction velocity of wtMnP and rMnP for DMP and A2E activity as a function of pH. Conditions were as defined in Methods but with 100 mM acetate/100 mM bis-tris buffer. Data are mean ± SEM (n = 3). Data are normalized to 1 at the pH of

49 highest activity. E) Michaelis-Menten curves for wtMnP and rMnP for breakdown of

A2E at room temperature. [Enzyme] = 5 μM.

50

Figure 2.2. rMnP activity in cell-free and cell culture models. A) Degredation of purified lipofuscin fluorophores by HPLC peak integration relative to BSA control.

Concentration of each fluorphore was 10 µM. Data are mean ± SEM (n = 3). B)

Cytotoxicity curves of rMnP, BioPORTER (BP), and rMnP + BP. For rMnP + BP, 1 μL

BP was used. Data are mean ± SEM (n = 3). C) A2E concentration in ARPE-19 cells with and without rMnP treatment. BP was used in all samples. Data are mean ± SEM (n =

6). Data were evaluated by one-way ANOVA with Holm-Sidak post hoc analysis. ns = not significant, ***p<0.001. D) Cell viability as measured by Annexin V / PI flow cytometry in A2E loaded cells with and without rMnP as in panel C 48 h. Data are mean

51

± SEM (n = 6) E) Mean fluorescence of RAW264.7 cells over time after treatment with

FITC labeled rMnP in the presence or absence of 4.9 mM mannan. FITC-labled man-

BSA was used as a positive control, and only one trial is shown to confirm functioning of

the system. rMnP data were evaluated by 2-way ANOVA (time and mannan treatment).

Data are mean ± SEM (n = 3).

52

53

Figure 2.3. rMnP activity in vivo. A) Vitreal half-life or rMnP measured by ELISA.

Data are mean ± SEM (n = 5). B) A2E concentration in Abca4(-/-) after 6 weekly doses of rMnP. Data points are A2E concentration pooled from 4 eyes (2 mice) presented as pmol eye-1. C) Pooled results from efficacy studies in Fig. 3B and Fig. S7. Data are mean

± SEM. D) Representative H & E stained frozen sections from eyes from mice in the

ERG toxicity study shown in Fig. S9. E) Quantification of retinal layer thickness from

H&E frozen sections of eyes from mice in the ERG toxicity study shown in Fig. S9. Data

are mean ± SEM using single eyes from individual mice. IPL = inner plexiform layer,

INL = inner nuclear layer, OPL = outer plexiform layer, ONL = outer nuclear layer, PL = photoreceptor layer, RPE = retinal pigment epithelium, overall = total thickness. F)

Serum concentration of anti-rMnP antibody as a function of dose measured by ELISA

with rMnP baited plates. Each data point is a single mouse. Dosing was 1 intraocular

injection per week G) Serum antibody concentration after 4 weekly intraocular injections

of rMnP measured by ELISA with rMnP baited plates. Data are mean ± SEM (n = 19).

H) Serum and vitreal antibody concentration after 4 weekly intraocular injections of rMnP measured by ELISA with rMnP baited plates. Data are man ± SEM (n = 5).

Samples were compared by t-test. **p<0.01.

54

Table 2.S1. Side effects of rMnP intraocular injection after 6 weekly treatments

(pooled). Numbers of animals and percentages shown below. Data evaluated using Chi-

Square. **p<0.01, ***p<0.001.

55

Figure 2.S1. Monosaccharide chromatograms of rMnP performed by UC San Diego

GlycoAnalytics Core. A) N-linked monosaccharides and standards. B) O-linked

monosaccharides and standards. By comparing peak integrals to protein weights, we can

estimate ~26 N-linked mannose monomers, 1 N-linked N-acetylglucosamine monomer,

19 O-linked glucose monomers, and 3 N-acetylglucosamine monomers per rMnP molecule.

56

Figure 2.S2. DMP activity of wtMnP and rMnP. Both preparations of protein have peroxidase activity. Conditions were conducted as described in Methods and assayed in a

96 well plate reader in absorbance mode. wtMnP seems to exhibit a burst phase followed by a linear phase, while rMnP shows a slight lag.

57

Figure 2.S3. A2E degradation by rMnP in the presence and absence of oxygen. A2E breakdown was measured by disappearance of the absorbance peak at 434 nm as detailed in Supplemental Methods. There is no significant difference between A2E only control and A2E + Enzyme in the absence of oxygen. No H2O2 was included in these reactions.

58

Figure 2.S4. Breakdown of A2E by (A) wtMnP and (B) rMnP by overnight incubation as a function of pH. 44 uM samples of A2E were incubated with 10 μM of the indicated enzyme at the indicated pH under the conditions used in Fig 1D overnight at room temperature, and the A2E remaining measured by HPLC. Data are mean ± SE (n =

3).

59

Figure 2.S5. Representative A2E breakdown kinetics with wtMnP and rMnP. wtMnP activity bursts then remains linear, rMnP activity bursts and then plateaus.

Enzyme concentrations were 1 μM for wtMnP, 5 μM for rMnP.

60

Figure 2.S6. CD206 detection on (A) ARPE-19 and primary (B) RPE cells by flow cytometry. Staining conditions were as described in Methods. ARPE-19 cells have no detectable CD206, and primary human RPE cells lose the receptor as a function of passage.

61

Figure 2.S7. A2E concentration in Abca4(-/-) after 6 weekly doses of rMnP. Data points are A2E concentration pooled from 4 eyes (2 mice) presented as pmol eye-1. Data

were evaluated using a students t-test.

62

Figure 2.S8. Mouse weights for the efficacy study in Fig. 2.3B. No significant reduction in weight was seen during the course of the study. Data are mean ± SEM.

63

Figure 2.S9. ERG recordings for the 12-week 56 µg x 6 dose toxicity study. Data are mean ± SEM. (n = 6 per group). Sham, PBS, and rMnP treated all exhibit decreased amplitudes, with rMnP treated exhibiting the smallest amplitude.

64

Figure 2.S10. Normalized HPLC chromatograms (left) and absorbance spectra of

the indicated major peak (right). Samples are A) A2PE, B) NRPE, C) A2E, D) Iso-

A2E, E) ATRdi.

CHAPTER 3: LEAN START-UP: A COMPREHENSIVE CASE STUDY IN THE

ESTABLISHMENT OF LABORATORY INFRASTRUCTURE2

SUMMARY

Historically, innovation in the biotechnology sector has relied to a large extent on

the expensive infrastructure provided by universities or large pharmaceutical companies.

This prohibitive start-up expense is the basis of why garage-style biotechnology

entrepreneurs are exceedingly rare as compared to their software and high-tech

counterparts. Recent consolidation among pharmaceutical companies and the release of

next generation research equipment has produced an affordable surplus in the secondary

equipment markets, reducing the barrier to entry posed by equipment expenses. We

examine the biotechnology start-up Ichor Therapeutics, Inc., and review strategies that

the founding team has successfully employed to establish an affordable laboratory and

reduce research expenses. Corporate structuring strategies to reduce risk and provide

stability are also discussed.

INTRODUCTORY REMARKS

Historically, innovation in the biotechnology sector has relied to a large extent on the expensive infrastructure provided by universities or large pharmaceutical companies.

2 This chapter is based on the previous published work: Grohn, K. J. et al. Lean Start-up:

A Comprehensive Case Study in the Establishment of Affordable Laboratory

Infrastructure. Journal of Commercial Biotechnology 21, (2015).

65

66

This prohibitive start-up expense is the basis of why garage-style biotechnology entrepreneurs are exceedingly rare as compared to their software and high-tech counterparts. Biotechnology entrepreneurs also face the additional challenge of inflated reagent and consumable pricing. This stems from the proprietary nature of many research products, and as a result of most research being ultimately supported by public funds.

In recent years, consolidations among pharmaceutical companies and the release of next generation research equipment has led to a surplus of pre-owned equipment in the secondary market. The equipment surplus has substantially reduced the barrier to entry imposed by limited equipment access. In the present case study, we examine the biotechnology start-up company Ichor Therapeutics, Inc., and review strategies that the founding team has successfully employed to establish an affordable laboratory, promote information sharing among team members, reduce research expenses, and guide scientific discovery. We then discuss corporate structuring strategies used by the company to reduce risk and provide stability.

BACKGROUND

Ichor Therapeutics, Inc. was founded in May 2013 with a grant from the Life

Extension Foundation, a private entity that supports scientific and medical research related to the prevention of degenerative disease. The primary focus of the company is to address a known bottleneck in the field of regenerative medicine89: deriving hematopoietic stem cells (HSC) from human pluripotent stem cells. Briefly, HSC are a type of adult stem cell that resides in the bone marrow and maintains the hematopoietic system, which includes all immune and blood cells, throughout life. Bone marrow and

67

cord blood transplants are useful in clinical practice to treat a wide range of diseases

because of the presence of HSC within grafts. After transplantation, HSC migrate into

and repopulate the hematopoietic system of the host. Unfortunately HSC are extremely

rare, representing only 0.05% of bone marrow cells, so a chronic supply shortage

persists90. Developing a scalable manufacturing process to produce HSC from pluripotent

stem cells would address this unmet medical need.

A start-up company focused on stem cell research is an excellent case study

because the infrastructure requirements are extensive. While many labs require basic

laboratory equipment for mammalian cell culture (incubators, laminar flow hood,

inverted microscope), molecular biology (shaker incubator, electrophoresis equipment,

refrigerated centrifuge), and analytics (flow cytometer, microplate reader, fluorescence

microscopy), Ichor also required liquid handling robotics for medium-throughput

screening of differentiation protocols and a vivarium (suitable for housing severe

combined immunodeficient mice) to assess the function of its cellular products in vivo.

BASIC EQUIPMENT AND CONSUMABLES PRECUREMENT

One of the highest barriers to entry for a biotechnology start-up is the significant cost of establishing a basic laboratory, as the acquisition and maintenance cost of equipment has historically been prohibitive. The last decade has been marked by consolidation of large pharmaceutical companies and the closure of many early stage biotech companies. However, this market volatility has nurtured a healthy secondary equipment market that can help to overcome this barrier91. Secondary markets include

offerings at online auctions and through used equipment vendors. Regardless of where

68 equipment is obtained, buyers should confirm the availability of user manuals, technical schematics, replacement parts, and free software before making a purchase. Failure to do so may result in unexpected post-acquisition expenses. For example, a used Molecular

Devices Vmax absorbance reader can be purchased online for as little as a few hundred dollars. However, these instruments rarely include software, which must be purchased from the manufacturer at a cost of $4,119.00 (Molecular Devices, 2013, personal communication).

Reagents used during the normal operations of a biotech company present another major cost to potential entrepreneurs. Commonly used supply companies have universities and government funded labs as their primary market, and as a result of those labs buying in bulk to negotiate discounts on consumables, the list price of those same consumables have increased. The authors highly recommend comparing prices between small specialist supply companies, as the list prices from these sources can be significantly lower than the list price of larger suppliers.

PURCHASING AT AUCTION

Online auction websites generally provide the largest savings when buying equipment, but offer no guarantee of item quality or customer support. The variable nature of online bidding means that item cost can be extremely dynamic at different times, even on the same website. For example, purchasers at Ichor observed that winning bids for an identical product at auction ranged from $50.00 to $3,750.00. Therefore, groups that purchase through auction should study winning bids for similar items over time to set a realistic bid ceiling and identify the best deals. Using this information for

69

proxy bidding -- a process by which a maximum bid for an item is set by the bidder, and

then the auction website automatically increases the bid up to this limit in response to

other bidders -- is particularly useful to help bidders avoid overpaying for highly

competitive items.

The inability to purchase essential items on-demand, uncertainty about equipment quality, and lack of warranty, make auction purchasing better suited for teams with flexible timelines for equipment procurement and the expertise to repair and service equipment in-house. Groups attempting this strategy must accept that some items will inevitably need to be thrown away. However, the cost savings over time should more than justify the losses, provided a responsible bidding strategy is adopted throughout92.

Importantly, some universities have policies that prohibit investigators from

purchasing equipment at auction, either directly or indirectly. This issue can be further

complicated when the equipment purchased qualifies as a fixed asset. SUNY Research

Foundation defines a capital asset as, “A single item with an acquisition cost of $5,000 or

more and has a useful life beyond one year.” An investigator may purchase three

damaged units for $5,000 each at auction, and ruse parts from two of the units to repair

one unit with a refurbished value of $30,000. However, it may be difficult for the

investigator to throw away the two units that were scrapped for parts because of

Foundation policy. Even donated assets present a potential problem because they are

assigned the fair market value at the time of acquisition. Finally, most institutions require

investigators to obtain quotes from a minimum of three vendors for equipment purchases.

Collectively, it is advisable for academic investigators to contact their purchasing

70

department for specific policy information relating to auction purchasing and other forms

of equipment procurement before bidding.

PURCHASING THROUGH USED EQUIPMENT VENDORS

Equipment resellers are a faster, more reliable, but generally more expensive

means of acquiring equipment as compared to purchasing at auction93. Some used

equipment vendors employ in-house technicians to refurbish used laboratory equipment to factory specifications before putting the product up for sale, while others simply acquire cheap equipment at auction then directly resell to customers without servicing or recertification. Because of this, the quality of goods purchased through resale vendors can be superior to those found at auction, but ultimately each vendor must be assessed on a case-by-case basis. One major advantage over auction purchasing is that many used equipment vendors offer extended warranties on refurbished equipment, reducing purchasing risk for the buyer. Further, fewer institutional policies exist that prohibit academic investigators from purchasing from used equipment vendors as compared to purchasing at auction (SUNY Upstate Medical University Purchasing Department, 2013, personal communication).

As with other retail companies, stagnation of inventory for used equipment vendors is often undesirable because of the real cost associated with equipment storage94,95. Because of this, these vendors are generally motivated to move inventory

quickly and may be willing to part with items for far less than the advertised price,

especially if an offer is made on an unpopular product. Vendors will often accept low

offers when several items are purchased at once for the same reason. Previous price

71

reductions on equipment are generally a good indicator that the interest in the item is low,

and the vendor may be willing to sell at a reduced price to move inventory.

By default, purchasers at Ichor make starting offers of not more than 50% of the

asking price when purchasing from the website of a used equipment vendor. Even these

offers are only made after carefully studying pricing trends and product availability

through auction and other vendors. Approximately half of these "low ball" offers are

accepted immediately, which supports the usefulness of haggling for start-up companies

looking to stretch limited seed capital.

INFORMATION TECHNOLOGY AND SPECIALTY EQUIPMENT

COMPUTER WORKSTATIONS

Integrating personnel into a cohesive team is a persistent challenge at any company, and this challenge can be exacerbated during periods of rapid growth in a start-

up environment. Streamlining information sharing and communication between

researchers and administrative staff is essential; not only to promote efficiency within the

team, but also to reduce growing pains as the company expands over time. For early stage

start-up companies, hiring information technology support staff or project managers to

improve laboratory efficiency can be prohibitively expensive. Fortunately, there are

several free off-the-shelf solutions available, depending on the specific needs of the team.

Ichor uses several of these platforms to manage its workflow, including Dropbox,

Evernote, Zotero, and Quartzy. Although these tools may not be appropriate for some

applications, such as those involving sensitive patient data, they can suit the needs of

many laboratories. Of note, it is generally advisable to implement these solutions early

72 while the team is small, rather than later, when the adoption of change may be more difficult to encourage.

Dropbox is a free cloud-based storage service that automatically syncs data between computers and devices96. Ichor created a company Dropbox account (free up to

2 GB storage space, $9.99 per month for up to 100 GB) and installed Dropbox on all workplace computers. All team members are given their own folder to organize as they see fit. A team folder was also created that contains pertinent information, such as spreadsheet templates for common calculations and data acquisition, user manuals for equipment, and company documents, such as expense reports. Syncing data across all workplace computers increases the availability of core infrastructure because data can often be obtained on the computer attached to the instrument, but analyzed later on a general-use workstation. Because Dropbox is cloud-based, physical backups of files are generally not as necessary. Dropbox and similar services are ideal for labs with modest or low hard drive requirements, but may not be suitable with computationally intensive projects involving large data storage.

Evernote is a free cloud-based notebook client that automatically syncs data between computers and devices97. Evernote is customizable and is capable of handling a variety of different file types and sizes. Users are able to annotate imported spreadsheets and images, and can also attach raw data to each annotation. This helps to streamline management within the laboratory. Supervisors may review the notes and results of a team member, but can readily access the raw data for their own interpretation as needed.

A personal Evernote account is free, and integrated business accounts can be added for a

$10/user per month fee. Each team member at Ichor has personal and business notebooks,

73 the latter of which is shared with other company employees. Notebooks are synced automatically across all devices. An employee's information can be maintained in the business notebooks, but company policy permits team members to make copies of non- confidential information, such as basic protocols, in their personal notebooks for later use. This policy is particularly helpful for temporary employees, such as interns or collaborators, who come to the company for technical training. When any employee leaves the company, their business account is archived and can no longer be viewed by them, but the account remains accessible for current employees to reference.

Zotero is a free cloud-based tool that organizes peer-reviewed journal articles using a searchable interface98. It automatically syncs data between computers and devices. Content may be collected and organized by each individual user. Zotero also supports a group feature where users can share information with one another through a central repository. Ichor uses a Zotero group to streamline document sharing among its team members. Training new employees is simplified through the use of a “new hires” folder, which contains literature reviews and protocol collections of relevance to the company workflow. New hires are able to copy this literature into their individual libraries and annotate documents all within the software. Importantly, Zotero can also integrate with common word processing applications like OpenOffice and Microsoft

Word, automating bibliography and in-text citation formatting during the preparation of manuscripts for publication or grant submission.

As the needs of a laboratory become more sophisticated, management software like Quartzy, a cloud-based platform, can be used to centralize order requests, track inventory, store laboratory records, and schedule equipment use. Vendor supplied

74 programs can also be useful to augment workflow by centralizing service requests.

LabLinker, for example, allows researchers to schedule services such as DNA sequencing and primer synthesis.

Collectively, there is a wide selection of affordable information technology solutions to promote laboratory efficiency and improve communication. Ichor regularly surveys its team members at laboratory meetings to identify workflow bottlenecks, and then uses this information to seek out technology solutions to address them. Team members then provide critical feedback of new solutions during trial periods before a decision is made to rollout the software company-wide. This cycle has helped Ichor to establish a company culture where efficiency is valued and promoted.

LABORATORY AUTOMATION

Laboratory automation solutions may represent one of the most underutilized and cost-effective benefits of engaging in used equipment purchases. Liquid handling robots in particular have broad applications in the laboratory. They are adept at performing repetitive tasks, and generally have less random error than human technicians99.

Ichor utilizes traditional embryoid body formation as a preferred method of inducing pluripotent stem cell differentiation, and purchased a used Biomek 2000

(Beckman Coulter, USA) at auction (GoIndustry, DoveBid, USA) to automate the process. Briefly, embryoid body formation involves detaching pluripotent stem cells from their growth surface, and transferring them to culture dishes with low adherence in the presence of media containing factors that promote differentiation into desired cell types.

Because the new growth surface has low adherence, the pluripotent stem cells self-

75

aggregate to each other rather than the plate surface, forming spheres termed embryoid

bodies. The size of the embryoid body influences the differentiation process, so it is

important to control this variable when optimizing differentiation conditions100,101.

To enable sterile work including cell culture, a customized semi-sterile enclosure

was constructed for the Biomek 2000 using basic materials from a home improvement

store at a cost of less than $550.00. The functionality of the robot was greatly enhanced

by the development of several custom 3D printed tools, which cost less than $5.00 each

to print. To validate the system and demonstrate proof-of-concept for its utility in embryoid body formation, production of CD14+CD45+ monocytes from pluripotent stem cells using an embryoid body method was successfully automated on the Biomek 2000

(Eric Zluhan, 2014, manuscript in preparation).

Collectively, liquid handling robots have the potential to dramatically improve workflow and reduce labor expense in a lean start-up environment. They can be substantially augmented to support unique applications with a little creativity and minimal capital. If planned carefully, automated methods can be designed in a modular format that provides short-term value by performing basic processes, yet enables convenient module integration for more complex applications in the medium or long term.

VIVARIUM

Assessing the function of human pluripotent stem cell derived products in vivo represented the most demanding infrastructure requirement for the company. Human cells cannot be evaluated in standard laboratory mice because they will be rejected by the host immune system. Instead, severe combined immunodeficient (SCID) mice need to be

76 used. These genetically engineered mice have deficient immune systems that permit engraftment of human cells; but by extension, they are also hypersensitive to otherwise benign pathogens and must be housed in specialized clean rooms to avoid death from opportunistic infection102.

To accommodate this need, Ichor built a customized 11' x 9' clean room that utilized a two-tiered positive pressure system. Vinyl flooring (catalog number 219836,

Lowes, USA) was installed and the room was then subdivided with standard 2" x 4" studs

(catalog number 6005, Lowes, USA) into three smaller rooms, including a viewing room, a gowning room, and a clean room. The rooms were electrically wired on two circuits, one controlling LED lights installed for basic lighting, the other controlling germicidal fluorescent light bulbs (catalog number 299305, Philips, USA) installed in under cabinet light fixtures (catalog number 7108, Lights of America, USA) for disinfection. The walls were constructed with reflective insulated sheets (catalog number 15358, Lowes, USA) and an observation window, sliding glass door, and interior door were also installed. Air was filtered with HEPA allergen removers containing carbon filters (catalog number

550266396, Wal-Mart, USA) and piped into the rooms through galvanized heating duct.

Three units were installed for the clean room, and one for the changing room to create a tiered positive pressure gradient. A blow-off was also installed in the clean room. The clean room and gowning room were sealed with silver foil tape (catalog number 225505,

Lowes, USA) to maintain pressure and control air flow.

To promote a pathogen free environment, each room (walls, floor, and ceiling) is disinfected with 70% isopropyl alcohol and UV light every two weeks. To assess relative air sterility, LB agar petri dishes are placed uncovered inside and outside each room for

77

45 minutes then covered and grown overnight at 37 °C. Colonies are then scored and

recorded.

Although the clean room design may not be appropriate for all applications,

available data suggest it is effective at protecting the company's SCID mice from

infection. In over 9 months of operation, clean room LB plates have not grown a single

bacterial colony, and no detectable or discernable infection has been observed in

laboratory mice. The cost to construct the vivarium was $2,259.52($879.52 materials +

$1,380.00 labor) and up to 15 cages can be conveniently housed in the clean room. This

is in stark contrast to an estimated $43,164- $45,144 a commercial-grade vivarium of similar size103.

BUSINESS MODEL

It is well known that most biotechnology companies inevitably fail because of the

high risk associated with clinical research and development programs. Surprisingly, few

founding teams take this fact into consideration when developing their business plans104.

For small start-up companies, cash flow may be detrimentally turbulent. When an early

stage start-up company runs out of operating capital, its assets are often liquidated and

the resulting capital is returned to investors. Losing basic laboratory functionality can

prematurely terminate an otherwise viable venture and it can take many months to rebuild

necessary infrastructure, even after raising new capital. At Ichor, the preservation of

laboratory access has been prioritized. To accomplish this, Ichor uses multiple corporate

entities to manage its business and research programs. These entities reflect a mixture of

78 traditional high-risk biotechnology research and development, but are stabilized by more conservative business structures.

Ichor Therapeutics, Inc. functions as a contract research organization. Research and development activities, including employee payroll, are performed through this entity. Ichor Therapeutics, Inc. operates the online store WeCellStuff.com through a

DBA to obtain wholesale pricing on reagents and consumables. As a contract research organization, Ichor Therapeutics, Inc. can perform work for hire in addition to its intramural research, which helps to offset overhead. This strategy has a long history of use by biotechnology companies at all stages of development.

In 2014, Ichor Therapeutics, Inc. diverged its capital assets to a separate corporate entity, Ichor Laboratory Solutions, Inc., which leases laboratory equipment. As a leasing company, Ichor Laboratory Solutions, Inc. is able to utilize more conservative financing, such as low interest debt financing, and is not dependent on grants, research contracts, or dilutive investment to support its operations. Because the Ichor team is skilled in asset procurement, equipment leasing can be used to increase revenues or support other companies and entrepreneurs of strategic value.

Real estate in Central New York is inexpensive as compared to other regions in the United States. Ichor Therapeutics, Inc. has partnered with Kelsey Moody &

Associates, LLC, which is owned and operated by Ichor’s CEO. Through this agreement,

Kelsey Moody & Associates, LLC can issue convertible notes instead of collecting rent, allowing Ichor Therapeutics, Inc. not only to persist, but remain operational during periods of insolvency. Through this partnership, Ichor Therapeutics, Inc. also provides various tenants shared access to its research facilities. Although indirect, including a real

79 estate component to the broader company structure has stabilized Ichor Therapeutics, Inc. and allows the founding team to make strategic decisions that focus more on the medium and long term, rather than short term, success of the company. In recent years, graduate students have become more focused on entrepreneurial ventures and careers in industry, rather than the pursuit of traditional academic appointments. A strategy involving real estate acquisition may be particularly well suited for young graduate students who expect to complete many years of study in one location, and lend necessary stability as they build out their own biotechnology start-ups.

In an effort to reduce the burden of high consumable and reagent pricing for its research and development activities, Ichor has established an online store

WeCellStuff.com, which is a distributor for several manufacturers. Although online sales provide a small basal level of revenue for the company, functionally, it permits Ichor to receive wholesale pricing on these items for its own intramural research programs at considerable savings.

One consideration of using a multiple company approach is that investment deals are complicated. A company with active contract research activities, leasing, and real estate may actually deter investors who want the flexibility of investing only at the level of a specific research program. To overcome this obstacle, Ichor has designed its business structure to support the incorporation of subsidiary intellectual property holding companies. Investment funding is received at the level of the subsidiary, and Ichor

Therapeutics, Inc. (the contract research entity) is contracted by the subsidiary to perform the research, while the subsidiary retains resulting intellectual property. This structure provides numerous advantages to all parties. The investor benefits because start-up

80

expenses are reduced, they can invest at the level of a specific research program, and they

are not subject to any liabilities or other risks associated with other ongoing activities

(equipment leasing, real estate, etc.). Forced liquidation of company assets is not a factor, so the company achieves its goal of preserving access to a functioning laboratory.

Employees can be repurposed to work on other contract projects during periods of turbulent cash flow, enabling the founding team to maintain a competent workforce.

It is important that entrepreneurs understand that a multiple company structure

should be planned for early, but should be executed slowly over time. The cost to

maintain a corporation can range from approximately $250 - $5,000 per year (depending

on corporation type, state fees, and the extent to which accounting and legal services are

sourced). When dealing with several corporate entities, a new entrepreneur should be

careful to balance corporate needs with financial realities. For example, capital assets and

research intellectual property were contained within Ichor Therapeutics, Inc. during the

early stages until the business was sufficiently mature to benefit from diverging these

components into distinct companies. Making this sort of move too early can put a new

venture at risk of being “nickel-and-dimed” to death. Spin-off companies should contribute to growth and reduce risk, not contribute to either.

For entrepreneurs who want to focus exclusively on technology development or lack sufficient capital or assets to benefit from the aforementioned strategies, community laboratories are an attractive alternative. These laboratories can range from small do-it- yourself (DIY) hobbyist labs, the so-called biotechnology “hacker spaces”105, to large

institutional biotechnology incubators106. DIY-shared spaces can be rented for low

monthly membership fees, whereas institutional incubators rent out dedicated suites that

81

contain both office and wet laboratory space. To subsidize the expense of maintaining

expensive core facilities, many academic universities offer per run or per hour pricing for

equipment use, so even the most expensive equipment is often readily available for use.

Collectively, the savvy entrepreneur can creatively utilize these and similar solutions to

overcome the accessibility barrier, despite significant financial constraints.

EMERGING BUSINESS MODELS

At the most basic level, the focus of Ichor has been to create a network of

companies that can sustain a pre-clinical research and development pipeline. Other

companies, at all stages, have found opportunities to establish sustainable infrastructure.

Functionally, this is a balance of reducing operational expenses and increasing revenue.

One of the more interesting and increasingly popular emerging motifs is to apply

laboratory automation technologies to improve experiment reliability, drive down costs,

and offer early-stage revenue. This strategy is distinct from most life science ventures,

which typically do not have cash flow in the early stages of development.

Emerald Therapeutics and Transcriptic are two biotech start-up companies with a focus centered on automation. Both of these companies offer automated lab services, which run customized protocols designed by researchers through the company website.

A large variety of tasks have already been automated, including common techniques like

PCR, transfection, chromatographic tests, DNA preparation, and RNA extraction among others, with more manually intricate techniques such as x-ray crystallography and patch clamp recording being developed. The value of the automated laboratory model not only comes from the ability to perform experiments more quickly and accurately than a human

82 technician could for the same price, but also adds a valuable revenue component to the company. Founders of both Emerald Therapeutics and Transcript claim that a task requiring a technician to perform liquid handing for a months’ time could be compressed into a week. Relevant information is also recorded at each step of the process, making replicating the experimental conditions far easier and making the end results more reliable.

Laboratory automation is a powerful tool, but it is still limited by logistics and issues with customer relations. Interestingly, some of the most common technical issues relate to tasks that would be simple when performed by a human technician. For example, the need to move samples in and out of cold storage and incubation chambers, or the uncapping and recapping of different containers can be challenging for a robotic arm.

Another challenge is managing the storage and maintenance of many different test specimens and special materials shipped to the worksite from users. Advertising services to potential customers is difficult, and often requires proactive outreaching to university labs. As with any emerging service, researchers may be cautious to risk large sums of limited funds. Transcriptic and Emerald Therapeutics attempt to address these flaws through custom engineering devices to handle logistical tasks, and by using modular workstations to allow their operations to adjust to a sporadic workflow.

The end product of automated laboratory services is simply a compilation of user experimental data, which could eventually support a broader market base than previous service models. This could have a transformative effect internationally, as areas with poor infrastructure could use these services to perform experiments that would otherwise be unfeasible.

83

ASSET IDENTIFICATION FOR CORPORATE STRUCTURING

Because every start-up is unique, providing a comprehensive “how to” guide for

creating a successful company is not feasible. However, the ability to effectively identify

and capitalize on corporate assets is perhaps the most important trait shared by each

company discussed previously. The business model canvas is a tool the authors of this

manuscript recommend to entrepreneurs at all stages to identify and focus on critical

business activities. The original Osterwalder Business model canvas (Fig 3.1) contains 9 components that show how a company intends to generate revenue. These include, 1) value propositions, 2) customer segments, 3) channels, 4) customer relationships, 5) revenue streams, 6) key activities, 7) key resources, 8) key partners, and 9) cost structures. Several variants of have been made based on the open source Osterwalder

Canvas, but most still focus on these original 9 blocks. The business model canvas is an excellent tool for designing a new venture, illustrating your business model, and for refocusing an existing business model to be more efficient.

CONCLUSION

Traditional sources of seed capital for high-tech and software startups, such as friends and family, have historically been insufficient to support the financial demands of biotechnology ventures. But as barriers to entry are eroded, biotechnology as an industry is beginning to move from centralized institutions to the garage. Ichor began in the living room of its founder, who at that time was a medical student. The company has since grown and expanded into a string of companies that balance the risk of research and

84 development with conservative, sustainable enterprises. It is the hope of the authors that this manuscript provides some guidance for aspiring entrepreneurs and shows, by example, that garage-style biotechnology startups are not only possible, but also viable.

85

Figure 3.1. Business Model Canvas107. The Business Model Canvas is a planning tool used to summarize the various segments of the organization as they relate to the activities of the business.

CHAPTER 4: AUTOMATING HESC DIFFERENTIATION WITH 3D PRINTING

AND LEGACY LIQUID HANDLING SOLUTIONS3

SUMMARY

Historically, the routine use of laboratory automation solutions has been prohibitively expensive for many laboratories. As legacy hardware has begun to emerge on the secondary market, automation is becoming an increasingly affordable option to augment workflow in virtually any laboratory. To assess the utility of legacy liquid handling in

stem cell differentiation, a used liquid handling robot was purchased at auction to automate a stem cell differentiation protocol that gives rise to CD14+CD45+ mononuclear cells. To maintain sterility, the automated liquid handling robot was housed in a custom constructed HEPA filtered enclosure. A custom cell scraper and a disposable filter box were designed and 3D printed to permit the robot intricate cell culture actions required by the protocol. All files for the 3D printed labware are uploaded and are freely available. A used liquid handling robot was used to automate an hESC to monocyte differentiation protocol. The robot-performed protocol induced monocytes as effectively as human technicians. Custom 3D printed labware was made to permit certain cell culture actions and are uploaded for free access.

3 This chapter is based on the previous published work: Zluhan, E., Kelly, K., LeClair,

N., Wortel, D. & Moody, K. Automating hESC differentiation with 3D printing and

legacy liquid handling solutions. MethodsX 3, 569–576 (2016).

86

87

METHODS DETAIL

Here, we assessed the utility of a legacy liquid handling robot at performing a stem cell differentiation protocol that requires intricate and accurate movements. An automated liquid handling robot was purchased used at auction ($100 from eBay.com) and was programmed to perform a stem cell differentiation protocol initially developed by

Wilgenburg et al. that gives rise to mononuclear cells108. Briefly, the automated liquid

handling robot is controlled by a computer and consists of a robotic arm that controls

modular pipette attachment tools. Like standard pipettes, the automated liquid handling

robot’s pipetting tools are used to transfer liquid from one location to another. The robot

also has a gripper tool that can manipulate culture dishes and other lab equipment on the

work surface while housed in a custom sterile enclosure. We developed and used custom

3D printed tools to aid the robot in the differentiation protocol and provide the 3D files

others to download.

Modifications to hESC differentiation protocol for use with the liquid handling robot

Minor modifications were made to the protocol by Wilgenburg et al to optimize it for use

on the automated liquid handling robot:

Wilgenburg et al. plated 20 embryoid bodies from a microplate to a 35mm well of a 6-

well as part of the protocol. However, the multichannel pipette of the automated liquid

handling robot has 8 positions, so the robot is instead programmed to transfer 16

embryoid bodies to each well for convenience.

Wilgenburg et al. culture hESCs on mouse embryonic fibroblast feeder layers but for this

method were maintained on stem cell culture plates with mTeSR-1 since this culture

88

system is more amenable for automation and appears comparable to the co-culture

method (data not shown).

Custom tools were made to accommodate the robots physical limitations.

Trituration combined with enzymatic detachment solution works for detaching loosely

adherent cells such as HEK 293 cells, but hESCs strongly adhere to the vessel surface

and require a cell scraper. The robot is not equipped with a cell scraper tool, however, its

gripper arm can be programmed to readily pick up and manipulate customized pieces of lab equipment. We exploited this functionality and designed and 3D printed a custom cell scraper with 6 positions so that each scraper fit into one well of a 6-well plate for parallel processing (Fig. 1A).

We also 3D printed a custom disposable “filter box” that allowed the automated liquid handling robot to filter the cells as described in the protocol by Wilgenburg et al (Fig

1B). The filter box was designed to allow the robot’s pipette to dispense cells through a filter and into a main reservoir (155 mL capacity), which could be accessed by the pipette tool to continue the protocol. To utilize as much of the robot’s work surface as possible, the filter box was designed in thirds so that it also contained two small reservoirs for holding phosphate buffered saline (PBS) for washes and Accutase for enzymatic detachment solution (the two chambers each hold ~35 mL). The filter box also features a holster for a micro centrifuge tube, which was used to collect an aliquot of cells to determine cell concentration. The custom filter box that contained necessary components for the protocol saved significant work surface space. The equipment that would normally occupy 3-4 work surface modules was reduced to 1 module.

89

The cell-scraper and filter box were designed in Google SketchUp® 3D modeling software. We found it convenient to design custom tools that are based on the dimensions of a typical microplate in the x,y dimensions (~127.76mm X 76.49mm) The schematics were exported as an .stl file and printed on a MakerBot Replicator® 2 3D printer with white poly lactic acid (PLA) plastic. Standard 3D printed items typically have some degree of porosity and not liquid proof. This can be an issue for continued reuse in sterile conditions. To address this, items can be coated in a waterproof sealant such as epoxy resin or silicone. Also, though we used disposable PLA items here, we have had success treating ABS printed items with acetone, which melds the filament layers and makes the item liquid proof. The 3D printed tools were sterilized by UV light before use and the filter box was discarded after use. The .stl files for these designs are available for download in the supplementary materials.

Automating the hESC differentiation protocol by Wilgenburg et al.

Note: We recommend verifying the robot’s pipetting and cell culture accuracy with an inexpensive cell line such as HEK 293 cells before proceeding with hESC culture (Fig.

S1)

Four work surface configurations were used to support the protocol on the automated liquid handling robot (Fig. 2). Each work surface layout is used for specific step of the differentiation protocol (refer to Fig. 3A).

Two wells of hESC (ES-701, BioTime, USA) starter cultures were manually seeded on

Synthemax 6-well plates CLS3978, Sigma-Aldrich, USA). When cells were 70-80% confluent, the robot was programmed to passage the cells from 2 to 6 wells. This resulted in 2 new starting wells and 4 wells to be used for the monocyte (MNC) differentiation

90

(Fig. S2). A single well of a 6-well plate yields enough cells for at least two 96-well plates for embryoid body formation. A single 96-well plate yields two 6-well plates for monocyte differentiation with 16 embryoid bodies in each well.

Custom movement scripts were programmed in the robot’s software (Bioworks) for non- native movements. These are sequences of simple commands to move the robot arm in the x, y or z direction, manipulate the gripper arm and aspirate or dispense with the pipette. The scripts can be recorded and called via the user interface. Custom movement scripts were particularly useful for using the custom labware. For cell passaging, the robots gripper arm was programmed to pick up the cell scraper and scrape the wells.

Then the robot’s pipette tool was programmed to move to four corners of each well for trituration to ensure cell/media removal.

Note: Robots should be calibrated before recording scripts and before performing protocols to ensure accurate execution.

The robot was able to perform all of the steps of the hESC differentiation protocol but was not equipped with a centrifuge for the 96-well embryoid body formation step and required the user to manually centrifuge the plate. However, it is possible to equip the robot with a centrifuge if desired.

METHOD VALIDATION

Comparison of robot and human technicians performing the hESC differentiation protocol

The robot and a human technician performed the monocyte differentiation protocol (Fig.

3A) and their performance was compared. The robot-cultured and human-cultured cells

91 and embryoid bodies had similar morphology throughout the differentiation process (Fig.

3B). Cell viability was also compared and was comparable (data not shown). After 33 days in culture, cells in suspension were harvested, treated with anti-CD45 and anti-

CD14 and assessed by flow cytometry. Wilgenburg et al. reported an 8.97%

CD14+CD45+ monocyte yield 33 days after embryoid body formation. Here the human technician yielded 5.22% (±2.42%) and the robot yielded 8.38% (±5.69%) CD14+CD45+ cells per well (Fig. 4). These results are similar to those reported by Wilgenburg et al.

CD14-CD45-, CD14-CD45+, and CD14+CD45+ percent yields were not significantly different between a human technician and the robot (n = 5 for all and p < 0.43, p < 0.92, and p < 0.29, respectively). Taken together, the liquid handling robot is able to perform a complex stem cell differentiation protocol with comparable monocyte yields to human technicians.

ADDITIONAL INFORMATION

Reproducibility is critical when using sensitive stem cell-based assays. Automation provides a solution to side-step user error and standardizes methods. It is important to budget time for a development period before implementing an automated platform into existing protocols. We optimized a number of handling variables including culture conditions, passaging methods and incubation times in order to optimize the protocol by

Wilgenburg et al. for the robot’s use. It is important to take care in programming liquid handling robots and in preparing starting reagents to avoid downstream errors. For this protocol in particular, stem cell seeding density and pluripotency is key for success.

Calibration and development steps are also critical to obtain accuracy and consistency in

92 automated protocols. Several custom tools were designed for this protocol, but other protocols may have different requirements. For example, 3D parts may require waterproofing for repeated use in sterile conditions. There are also online stores (e.g. shapeways.com) that offer less common printing materials that may be more suited to different uses.

Making custom labware and other files freely available can reduce development phases and increase collaboration between laboratories. Here we have uploaded the 3D printing files for others to download and modify as they see fit. Other labs can reproduce this study or change the files directly to suit their own automation needs. We propose that this type of open-source mentality and use of legacy hardware can shrink certain barriers to entry for budding research and development teams.

93

Figure 4.1. Custom cell scraper and filter box labware. (A) Schematic of cell scraper in 3D modeling computer program (left) and image of 3D printed final product (right).

The cell scraper is designed to simultaneously scrape six wells for cell detachment. (B)

Schematic of filter box and lid in 3D modeling computer program (left) and image of 3D printed final product (right). The main chamber holds media and cell filtrate. The two small reservoirs hold PBS and Accutase, the micro centrifuge tube holder holds a tube used for obtaining cell concentration.

94

Figure 4.2. Work surface configurations for automated liquid handling robot. (A)

Blank deck layout of automated liquid handling robot. The robot comes with 8 deck slots

(A2-A5, B2-B5); 2 additional slots were added on each side (A1, A6, B1, B6) to increase work surface area. (B) Work surface layout for embryoid body formation step (Day 0).

Operator places 6-well and 96-well plates on work surface, 2mL of PBS and Accutase in small reservoirs, and 5mL spin media in main reservoir. Custom filter box at B2 is defined as a 6-well plate. A 6-well plate with confluent hESCs and an ultra low adherence round bottom 96-well plate are manually placed on work surface at B3 and B4.

95

(C) Work surface layout for 96-well 50uL media change of embryoid bodies (Day 2).

Operator places 4.8 mL spin media in reservoir on B3. (D) Automated liquid handling robot work surface layout for 96-well to 6-well transfer of embryoid bodies (Day 4).

Operator places 96-well (B2) and 6 well plate (B3) on work surface and 13.2 mL of mononuclear cell media in reservoir (B4). (E) Automated liquid handling robot work surface layout for 6-well 2/3 media change of mononuclear cells (Day 9). Operator places

6-well plate with mononuclear cells (B3) on work surface and 12 mL monocyte media in reservoir (B2).

96

Figure 4.3. Differentiation of human embryonic stem cells into monocytes. (A)

Timeline for hESC differentiation. (B) (1) Undifferentiated hESC before passaging, (2) after being plated and centrifuged in an ultra low attachment 96 well plate, and (3) just prior to transfer to a tissue culture treated 6-well plate. (4-5) Embryoid bodies adhere to culture vessel then give rise to non-adherent monocytes (MNCs).

97

Figure 4.4. Representative data by flow cytometry for monocyte differentiation. (A)

Representative flow cytometry plots for CD14+ CD45+ hESC derived monocytes.

Specific antibodies (left) vs. isotype controls (right). (B) Technician and automated liquid

handling robot results for hESC-derived monocytes harvested on day 33 are double positive for CD14 and CD45. n = 5 for all and p < 0.43, p < 0.92, and p < 0.29 respectively using student’s two-tailed t-test.

98

Figure 4.S1. Automated liquid handling robot performs cell culture with accuracy comparable to human technicians. (A) To compare pipette accuracy, the robot and human technicians performed serial dilutions of Giemsa stain into water and a microplate reader measured the absorbance of each dilution. Absorbance values were averaged for 8 wells for each concentration measured in the dilution series. Concentration and standard deviation are shown below. (B) The automated liquid handling robot demonstrated comparable cell passaging yields of HEK 293 cells compared to human technicians.

Graphical data are the average cell counts ± S.D. of the 6 passaged wells. ANOVA

99 analysis of the passaging efficiency showed there is no statistical significant difference between the three technicians and the biomek p > 0.13.

100

Figure 4.S2. Culture plate passage diagram. (A) Stem cell culture plates containing hESC colonies are passaged at a 2:6 ratio. This strategy restores the starting material with each passage, and provides a 6-well plate with 4 wells for use in differentiation. (B) hESC colonies in 4 stem cell culture plate wells are detached, diluted, then plated in an ultra low adherent 96 well plate to form embryoid bodies. (C) 8 embryoid bodies, corresponding to 8 wells of a 96 well plate, are pooled in each 35mm well of a 6-well dish for the final steps of differentiation. Each 96 well plate yields 2 6-well plates.

101

Figure 4.S3. Liquid handler and enclosure. A.) Automated liquid handling robot in custom sterile enclosure. B.) Representative sketch of automated liquid handling robot in sterile enclosure; its components: (1) Biomek 2000 automated liquid handling robot, (2) work surface, (3) main light, (4) UV lights, (5) waste bin, (6) air purifier, (7) control panel for air purifier, (8) lights and automated liquid handling robot on/off switch, (9) duct for air purifier.

102

Link to .stl files: https://www.dropbox.com/sh/yqti0k59akcrwbq/AADWfKm6hQdX6v2dOJcjWTIma?dl=

0

CHAPTER 5: CONCLUSION

Age-related macular degeneration (AMD) is the leading cause of vision loss in

patients over the age of 50. Like other age-associated diseases, new therapeutic

modalities and approaches must be pursued to address the immense unmet clinical need

for effective treatments. In applying the SENS damage repair and mitigation paradigm to age-related macular degeneration as a case example, the relevant metabolic process is the visual cycle, which recycles retinoids (vitamin A) to enable vision. The target “damage” consists of trace quantities of retinoid derivatives that cannot be catabolized by the body,

and form as by-products of the visual cycle. As they accumulate, they drive dysfunction of lipid metabolism within the essential support cells of the eye – retinal pigmented epithelium (RPE). This manifests as the accumulation of intracellular lipofuscin, consisting of retinoid and lipid derivatives, and later, the extracellular deposition of drusen, a medley of esterified cholesterol, phosphatidylcholine, lipid containing particles,

and a variety of RPE proteins. The SENS paradigm would categorize age-related macular

degeneration as an example of damage #7, intracellular aggregates, which drive

dysfunction and eventual death of the RPE, leading to progressive loss of vision. A SENS

inspired therapy would aim to eliminate or remove the offending retinoids from the RPE

cells to restore function. As is the case with many aging damage accumulations, a

juvenile form of age-associated macular degeneration, called Stargardt’s disease, exists.

In Stargardt’s disease, retinoid derivatives accumulate at an accelerated rate due to

mutations in the ABCA4 flippase, a critical enzyme in the visual cycle. The existence of

Stargardt’s disease offers significant commercial opportunities to develop a drug for both

103

104

forms of macular degeneration and offers greater confidence in pursuing the removal of intracellular retinoid derivatives as a valid molecular target.

I was inspired by the clinical successes of conventional enzyme replacement therapies for the treatment of congenital lysosomal storage disorders, such as Cerezyme for Gaucher’s disease and Fabrazyme for Fabry disease, among others. Drawing upon the body of literature surrounding these drugs, I designed and executed a research program aimed at establishing proof-of-concept for the use of a recombinant manganese peroxidase to remove intracellular lipofuscin from RPE cells, an approach I term enzyme augmentation therapy (EAT), an extension of conventional enzyme replacement therapy

(ERT). In this body of work, I presented the results of those efforts, which demonstrate that augmenting mammalian lysosomal function with exogenous enzymes to increase catabolic potential is feasible and may be translationally viable. I demonstrated that a recombinant manganese peroxidase can enter ocular target cells, the RPE, via a mannose receptor endocytosis mechanism. Both in vitro and in vivo, catabolism of lipofuscin retinoids was readily observed. This approach differs substantially from existing treatment modalities which target the visual cycle directly (i.e. visual cycle modulators) or chase downstream symptoms and pathologies (i.e. VEGF pathway and complement inhibitors).

Moving from a wild-type manganese peroxidase to a recombinant form produced in the yeast strain P. pastoris led to hypermannosylation of the enzyme. Quite unexpectedly, this caused a shift in the pH optima, whereby lipofuscin retinoid catabolic activity was elevated at physiologic pH, rather than at the acidic pH of RPE lysosomes where the enzyme needs to function. Although this phenomenon did not seem to prevent

105

robust clearance of lipofuscin retinoids in vivo and showed no indication of cytotoxicity

in RPE cell cultures, it did lead to apparent functional impairment of photoreceptors in

vivo. There are a few reasons why this may have occurred. First, efficient catabolism of

retinal retinoids could produce transient blindness by preventing the generation of

rhodopsin in the visual cycle. This would not be unsurprising given the short intervals

between administration of the enzyme. Alternatively, physiologic activity could be

driving toxicity directly. This explanation is inconsistent with cytotoxicity studies

conducted on the ARPE-19 cell line. However, in vivo we observed morphological changes in the photoreceptor layer in the enzyme treated group, but not vehicle control, suggesting the photoreceptor layer could be more sensitive to manganese peroxidase mediated cytotoxicity than RPE cells. Our observation that manganese peroxidase activity against DMP was substantively different than its activity against A2E as a substrate led us to hypothesize that the mechanism of action for DMP cleavage and A2E catabolism are distinct. Our current data indicate that A2E catabolism is oxygen dependent and rely on superoxide (data not shown, manuscript in preparation). It is well established that photoreceptors are highly sensitive to oxidative damage, so it is possible that RPE cells are simply more tolerant of oxidative insults. Collectively, these observations are very encouraging for further development of a manganese peroxidase to remove retinal lipofuscin. Glycoengineering to reduce the number of mannose residues on O-linked or N-linked glycans should restore a wild-type pH profile, which would considerably reduce activity at physiologic pH and mitigate toxicity to photoreceptors.

This would, at the same time, increase activity at lysosomal pH, improving enzyme efficacy in the lysosomal compartment where retinal lipofuscin aggregates.

106

In addition to glycoengineering and modification of the primary amino acid

sequence of manganese peroxidase to generate an optimal pH activity profile, additional

work must be performed regarding chemistry, manufacturing, and controls (CMC).

Heterogeneity in glycosylation state could lead to variability in the performance of

manganese peroxidase as a clinical product. The stability of recombinant manganese

peroxidase has not been investigated, nor has optimization been done surrounding culture

media, which includes a variety of requisite co-factors. P. pastoris has the advantage of being integrating, so stable cell lines can be produced. However, given the observed shortcomings of hypermannosylation, alternative expression systems may be preferred and should be investigated.

Finally, while my current work demonstrates proof-of-concept that a manganese peroxidase can eliminate retinal lipofuscin in vivo, only the ABCA-/- mouse model was

used in the work presented. Murine strains broadly have limited applications in the study

of age-related macular degeneration and Stargardt’s disease, particularly for formulations

administered by intravitreal injection. Mice do not have anatomic macula, and the

accumulation, distribution, and composition of retinal lipofuscin significantly differs

from what is known about human equivalents. Further anatomical variations complicate

study design and analysis. The mouse eye is small, so intravitreal injections which are

simpler in a larger target become very technically challenging on smaller scale. Repeat

injections are highly damaging, confounding interpretation of toxicity endpoints. The

murine lens is also significantly larger than that of a human, so intravitreal injections

must either pierce the retina itself, or pierce the lens, leading to cataracts, which prohibit

107

many forms of functional assessment, like OCT and fundus imaging. Collectively, these

observations mandate further toxicological assessments in larger model organisms.

These next steps for translating a recombinant manganese peroxidase for AMD and Stargardt’s disease overlay with the next steps of material expansion for Ichor

Therapeutics, my host institution. Additional capabilities to conduct large animal toxicology studies, particularly under Good Laboratory Practices (GLP), must be brought online to support an investigational new drug (IND) submission. The manufacturing of biological products is generally harder and more extensive than small molecule active pharmaceutical ingredients (APIs) and formations. For cost and the need for strict control over process, it is plausible, perhaps even likely, that clinical product manufacturing will need to be performed in-house. The scaled production of an enzyme therapy will require

significantly expanded manufacturing, purification, and analytical capabilities, with the

ability to meet Good Manufacturing Practices (GMP). Though daunting in scope and

complexity, there is motivation to be gleaned from the immense potential of this program

to treat the world’s greatest cause of age-associated vision loss.

WORKS CITED

1. Leading Causes of Death, 1900-1998. Diabetes mellitus 67 (1975).

2. Life Expectancy. Centers for Disease Control and Prevention (2018). Available at: https://www.cdc.gov/nchs/fastats/life-expectancy.htm. (Accessed: 22nd January 2019)

3. Kochanek, K. D. Mortality in the United States, 2016. 8 (2017).

4. Lipsky, M. S. & King, M. Biological theories of aging. Dis Mon 61, 460–466 (2015).

5. López-Otín, C., Blasco, M. A., Partridge, L., Serrano, M. & Kroemer, G. The hallmarks of aging. Cell 153, 1194–1217 (2013).

6. Grey, A. D. N. J. de et al. Is human aging still mysterious enough to be left only to scientists? BioEssays 24, 667–676 (2002).

7. de Grey, A. D. N. J. The foreseeability of real anti-aging medicine: focusing the debate. Exp. Gerontol. 38, 927–934 (2003).

8. Grey, A. D. N. J. D. et al. Time to talk SENS: critiquing the immutability of human aging. Annals of the New York Academy of Sciences 959, 452–462 (2002).

9. Shiloh, Y. & Lederman, H. M. Ataxia-telangiectasia (AT): an emerging dimension of premature ageing. Ageing research reviews 33, 76–88 (2017).

10. Oshima, J., Sidorova, J. M. & Monnat Jr, R. J. Werner syndrome: clinical features, pathogenesis and potential therapeutic interventions. Ageing research reviews 33, 105–114 (2017).

11. Karikkineth, A. C., Scheibye-Knudsen, M., Fivenson, E., Croteau, D. L. & Bohr, V. A. Cockayne syndrome: clinical features, model systems and pathways. Ageing research reviews 33, 3–17 (2017).

12. Faghri, S., Tamura, D., Kraemer, K. H. & DiGiovanna, J. J. Trichothiodystrophy: a systematic review of 112 published cases characterises a wide spectrum of clinical manifestations. Journal of medical genetics 45, 609–621 (2008).

13. Niedernhofer, L. J. et al. A new progeroid syndrome reveals that genotoxic stress suppresses the somatotroph axis. Nature 444, 1038 (2006).

14. Baple, E. L. et al. Hypomorphic PCNA mutation underlies a human DNA repair disorder. The Journal of clinical investigation 124, 3137–3146 (2014).

15. Defesche, J. C. et al. Familial hypercholesterolaemia. Nat Rev Dis Primers 3, 17093 (2017).

108

109

16. Altschwager, P., Ambrosio, L., Swanson, E. A., Moskowitz, A. & Fulton, A. B. Juvenile macular degenerations. Semin Pediatr Neurol 24, 104–109 (2017).

17. Kakkis, E. D., Kowalcyk, S. & Bronstein, M. G. Accessing the accelerated approval pathway for rare disease therapeutics. Nat. Biotechnol. 34, 380–383 (2016).

18. Hampton, T. Rare disease research gets boost. JAMA 295, 2836–2838 (2006).

19. Aitken, L., Gallagher, R. & Madronio, C. Principles of recruitment and retention in clinical trials. Int J Nurs Pract 9, 338–346 (2003).

20. Richesson, R. L. et al. An automated communication system in a contact registry for persons with rare diseases: scalable tools for identifying and recruiting clinical research participants. Contemp Clin Trials 30, 55–62 (2009).

21. Congdon, N. et al. Causes and prevalence of visual impairment among adults in the United States. Arch. Ophthalmol. 122, 477–485 (2004).

22. Singer, M. Advances in the management of macular degeneration. F1000Prime Rep 6, 29 (2014).

23. Gehrs, K. M., Anderson, D. H., Johnson, L. V. & Hageman, G. S. Age-related macular degeneration--emerging pathogenetic and therapeutic concepts. Ann. Med. 38, 450–471 (2006).

24. North, V., Gelman, R. & Tsang, S. H. Juvenile-onset macular degeneration and allied disorders. Dev Ophthalmol 53, 44–52 (2014).

25. Kiser, P. D., Golczak, M. & Palczewski, K. Chemistry of the retinoid (visual) cycle. Chem Rev 114, 194–232 (2014).

26. Molday, R. S. ATP-binding cassette transporter ABCA4: molecular properties and role in vision and macular degeneration. J. Bioenerg. Biomembr. 39, 507–517 (2007).

27. Feeney-Burns, L., Hilderbrand, E. S. & Eldridge, S. Aging human RPE: morphometric analysis of macular, equatorial, and peripheral cells. Invest. Ophthalmol. Vis. Sci. 25, 195–200 (1984).

28. Sparrow, J. R., Parish, C. A., Hashimoto, M. & Nakanishi, K. A2E, a lipofuscin fluorophore, in human retinal pigmented epithelial cells in culture. Invest. Ophthalmol. Vis. Sci. 40, 2988–2995 (1999).

29. Bergmann, M., Schütt, F., Holz, F. G. & Kopitz, J. Inhibition of the ATP-driven proton pump in RPE lysosomes by the major lipofuscin fluorophore A2-E may contribute to the pathogenesis of age-related macular degeneration. FASEB J. 18, 562–564 (2004).

110

30. The lipofuscin fluorophore A2E perturbs cholesterol metabolism in retinal pigment epithelial cells. Available at: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1904145/. (Accessed: 30th January 2019)

31. Wu, Y., Yanase, E., Feng, X., Siegel, M. M. & Sparrow, J. R. Structural characterization of bisretinoid A2E photocleavage products and implications for age-related macular degeneration. Proc. Natl. Acad. Sci. U.S.A. 107, 7275–7280 (2010).

32. Yoon, K. D., Yamamoto, K., Ueda, K., Zhou, J. & Sparrow, J. R. A novel source of methylglyoxal and glyoxal in retina: implications for age-related macular degeneration. PLoS ONE 7, e41309 (2012).

33. Zhou, J., Jang, Y. P., Kim, S. R. & Sparrow, J. R. Complement activation by photooxidation products of A2E, a lipofuscin constituent of the retinal pigment epithelium. Proc. Natl. Acad. Sci. U.S.A. 103, 16182–16187 (2006).

34. Schmidt-Erfurth, U. et al. Guidelines for the management of neovascular age- related macular degeneration by the European Society of Retina Specialists (EURETINA). Br J Ophthalmol 98, 1144–1167 (2014).

35. Tsybovsky, Y., Molday, R. S. & Palczewski, K. The ATP-binding cassette transporter ABCA4: structural and functional properties and role in retinal disease. Adv. Exp. Med. Biol. 703, 105–125 (2010).

36. Burke, T. R. & Tsang, S. H. Allelic and phenotypic heterogeneity in ABCA4 mutations. Ophthalmic Genet. 32, 165–174 (2011).

37. Chakravarthy, U. et al. Cigarette smoking and age-related macular degeneration in the EUREYE Study. Ophthalmology 114, 1157–1163 (2007).

38. Age-Related Eye Disease Study Research Group. A randomized, placebo- controlled, clinical trial of high-dose supplementation with vitamins C and E, beta carotene, and zinc for age-related macular degeneration and vision loss: AREDS report no. 8. Arch. Ophthalmol. 119, 1417–1436 (2001).

39. Age-Related Eye Disease Study 2 (AREDS2) Research Group et al. Secondary analyses of the effects of lutein/zeaxanthin on age-related macular degeneration progression: AREDS2 report No. 3. JAMA Ophthalmol 132, 142–149 (2014).

40. CATT Research Group et al. Ranibizumab and bevacizumab for neovascular age- related macular degeneration. N. Engl. J. Med. 364, 1897–1908 (2011).

41. Emerson, M. V. & Lauer, A. K. Current and emerging therapies for the treatment of age-related macular degeneration. Clin Ophthalmol 2, 377–388 (2008).

111

42. Garcia, J. M. et al. Stem cell therapy for retinal diseases. World J Stem Cells 7, 160– 164 (2015).

43. El Dib, R. P., Nascimento, P. & Pastores, G. M. Enzyme replacement therapy for Anderson-Fabry disease. Cochrane Database Syst Rev 2, CD006663 (2013).

44. Zimran, A. & Elstein, D. Management of Gaucher disease: enzyme replacement therapy. Pediatr Endocrinol Rev 12 Suppl 1, 82–87 (2014).

45. Yogalingam, G. et al. Cellular uptake and delivery of Myeloperoxidase to lysosomes promotes lipofuscin degradation and lysosomal stress in retinal cells. Journal of Biological Chemistry jbc–M116 (2017).

46. Schloendorn, J. et al. Medical bioremediation: a concept moving toward reality. Rejuvenation Res 12, 411–419 (2009).

47. Ratko, T. A., Marbella, A., Godfrey, S. & Aronson, N. Enzyme-Replacement Therapies for Lysosomal Storage Diseases. (Agency for Healthcare Research and Quality (US), 2013).

48. Boyle, D., Tien, L. F., Cooper, N. G., Shepherd, V. & McLaughlin, B. J. A mannose receptor is involved in retinal phagocytosis. Invest. Ophthalmol. Vis. Sci. 32, 1464– 1470 (1991).

49. Tarnowski, B. I., Shepherd, V. L. & McLaughlin, B. J. Mannose 6-phosphate receptors on the plasma membrane on rat retinal pigment epithelial cells. Invest. Ophthalmol. Vis. Sci. 29, 291–297 (1988).

50. Gong, B. et al. Characterization of N-linked glycosylation on recombinant glycoproteins produced in Pichia pastoris using ESI-MS and MALDI-TOF. Methods Mol. Biol. 534, 213–223 (2009).

51. Molday, R. S. & Moritz, O. L. Photoreceptors at a glance. J. Cell. Sci. 128, 4039– 4045 (2015).

52. Age-Related Macular Degeneration (AMD) | National Eye Institute. Available at: https://nei.nih.gov/eyedata/amd#5. (Accessed: 29th July 2018)

53. Wolf, G. Lipofuscin and macular degeneration. Nutr. Rev. 61, 342–346 (2003).

54. Tanna, P., Strauss, R. W., Fujinami, K. & Michaelides, M. Stargardt disease: clinical features, molecular genetics, animal models and therapeutic options. Br J Ophthalmol 101, 25–30 (2017).

55. Wang, J. et al. Photosensitization of A2E triggers telomere dysfunction and accelerates retinal pigment epithelium senescence. Cell Death & Disease 9, 178 (2018).

112

56. Kauppinen, A., Paterno, J. J., Blasiak, J., Salminen, A. & Kaarniranta, K. Inflammation and its role in age-related macular degeneration. Cell Mol Life Sci 73, 1765–1786 (2016).

57. Radu, R. A. et al. Complement system dysregulation and inflammation in the retinal pigment epithelium of a mouse model for Stargardt macular degeneration. J. Biol. Chem. 286, 18593–18601 (2011).

58. Zhou, J., Jang, Y. P., Kim, S. R. & Sparrow, J. R. Complement activation by photooxidation products of A2E, a lipofuscin constituent of the retinal pigment epithelium. Proc. Natl. Acad. Sci. U.S.A. 103, 16182–16187 (2006).

59. Perlee, L. T. et al. Inclusion of Genotype with Fundus Phenotype Improves Accuracy of Predicting Choroidal Neovascularization and Geographic Atrophy. Ophthalmology 120, 1880–1892 (2013).

60. Age-related macular degeneration: Treatment and prevention - UpToDate. Available at: https://www-uptodate-com.libproxy1.upstate.edu/contents/age-related- macular-degeneration-treatment-and- prevention?search=macular%20degeneration%20treatment&source=search_result& selectedTitle=1~63&usage_type=default&display_rank=1. (Accessed: 29th July 2018)

61. Charbel Issa, P., Barnard, A. R., Herrmann, P., Washington, I. & MacLaren, R. E. Rescue of the Stargardt phenotype in Abca4 knockout mice through inhibition of vitamin A dimerization. Proc Natl Acad Sci U S A 112, 8415–8420 (2015).

62. Ma, L., Kaufman, Y., Zhang, J. & Washington, I. C20-D3-vitamin A slows lipofuscin accumulation and electrophysiological retinal degeneration in a mouse model of Stargardt disease. J. Biol. Chem. 286, 7966–7974 (2011).

63. Cox, T. M. et al. Eliglustat compared with imiglucerase in patients with Gaucher’s disease type 1 stabilised on enzyme replacement therapy: a phase 3, randomised, open-label, non-inferiority trial. Lancet 385, 2355–2362 (2015).

64. Schloendorn, J. et al. Medical bioremediation: a concept moving toward reality. Rejuvenation Res 12, 411–419 (2009).

65. Wu, Y., Zhou, J., Fishkin, N., Rittmann, B. E. & Sparrow, J. R. Enzymatic degradation of A2E, a retinal pigment epithelial lipofuscin bisretinoid. J Am Chem Soc 133, 849–857 (2011).

66. Yogalingam, G., Beliakoff, G., Goldin, E. & Peto, M. Lysosomal aggregates. SENS Research Foundation Research Report 2013 3–4 (2018).

67. Yogalingam, G. et al. Cellular uptake and delivery of myeloperoxidase to lysosomes promote lipofuscin degradation and lysosomal stress in retinal cells. J. Biol. Chem. 292, 4255–4265 (2017).

113

68. Maiti, P. et al. Small molecule RPE65 antagonists limit the visual cycle and prevent lipofuscin formation. Biochemistry 45, 852–860 (2006).

69. Wright, C. B. & Ambati, J. Dry age-related macular degeneration pharmacology. Handb Exp Pharmacol 242, 321–336 (2017).

70. Tarnowski, B. I., Shepherd, V. L. & McLaughlin, B. J. Expression of mannose receptors for pinocytosis and phagocytosis on rat retinal pigment epithelium. Invest. Ophthalmol. Vis. Sci. 29, 742–748 (1988).

71. Dunn, K. C., Aotaki-Keen, A. E., Putkey, F. R. & Hjelmeland, L. M. ARPE-19, a human retinal pigment epithelial cell line with differentiated properties. Exp. Eye Res. 62, 155–169 (1996).

72. Morishima, S. et al. Expression and role of mannose receptor/terminal high- mannose type oligosaccharide on osteoclast precursors during osteoclast formation. J. Endocrinol. 176, 285–292 (2003).

73. Charbel Issa, P. et al. Fundus autofluorescence in the Abca4−/− mouse model of Stargardt disease—correlation with accumulation of A2E, retinal function, and histology. Invest Ophthalmol Vis Sci 54, 5602–5612 (2013).

74. Zhou, R. & Caspi, R. R. Ocular immune privilege. F1000 Biol Rep 2, (2010).

75. Kimura, M., Umemoto, Y. & Kawano, T. Hydrogen peroxide-independent generation of superoxide by plant peroxidase: hypotheses and supportive data employing ferrous ion as a model stimulus. Front Plant Sci 5, 285 (2014).

76. Greaton, C. J., Lane, K. B., Shepherd, V. L. & McLaughlin, B. J. Transcription of a single mannose receptor gene by macrophage and retinal pigment epithelium. Ophthalmic Res. 35, 42–47 (2003).

77. Boyle, D., Tien, L. F., Cooper, N. G., Shepherd, V. & McLaughlin, B. J. A mannose receptor is involved in retinal phagocytosis. Invest. Ophthalmol. Vis. Sci. 32, 1464– 1470 (1991).

78. Shepherd, V. L., Tarnowski, B. I. & McLaughlin, B. J. Isolation and characterization of a mannose receptor from human pigment epithelium. Invest. Ophthalmol. Vis. Sci. 32, 1779–1784 (1991).

79. Heth, C. A. & Bernstein, M. H. Mannose-sensitive HRP endocytosis by the retinal pigment epithelium. Experimental Eye Research 52, 75–82 (1991).

80. Park, S. J. et al. Intraocular pharmacokinetics of intravitreal vascular endothelial growth factor-Trap in a rabbit model. Eye (Lond) 29, 561–568 (2015).

81. Verdegem, P. J. E., Monnee, M. C. F., Mulder, P. P. J. & Lugtenburg, J. Condensation of all-E-retinal. Tetrahedron Letters 38, 5355–5358 (1997).

114

82. Parish, C. A., Hashimoto, M., Nakanishi, K., Dillon, J. & Sparrow, J. Isolation and one-step preparation of A2E and iso-A2E, fluorophores from human retinal pigment epithelium. PNAS 95, 14609–14613 (1998).

83. Liu, J., Itagaki, Y., Ben-Shabat, S., Nakanishi, K. & Sparrow, J. R. The biosynthesis of A2E, a fluorophore of aging retina, involves the formation of the precursor, A2- PE, in the photoreceptor outer segment membrane. J. Biol. Chem. 275, 29354– 29360 (2000).

84. Ahn, J., Wong, J. T. & Molday, R. S. The effect of lipid environment and retinoids on the ATPase activity of ABCR, the photoreceptor ABC transporter responsible for Stargardt macular dystrophy. J. Biol. Chem. 275, 20399–20405 (2000).

85. Gu, L., Lajoie, C. & Kelly, C. Expression of a phanerochaete chrysosporium manganese peroxidase gene in the yeast pichia pastoris. Biotechnology Progress 19, 1403–1409 (2008).

86. Reinboth, J. J., Gautschi, K., Munz, K., Eldred, G. E. & Remé, C. E. Lipofuscin in the retina: quantitative assay for an unprecedented autofluorescent compound (pyridinium bis-retinoid, A2-E) of ocular age pigment. Exp. Eye Res. 65, 639–643 (1997).

87. Gresh, J., Goletz, P. W., Crouch, R. K. & Rohrer, B. Structure-function analysis of rods and cones in juvenile, adult, and aged C57bl/6 and Balb/c mice. Vis. Neurosci. 20, 211–220 (2003).

88. Fishkin, N. E., Sparrow, J. R., Allikmets, R. & Nakanishi, K. Isolation and characterization of a retinal pigment epithelial cell fluorophore: An all-trans-retinal dimer conjugate. Proc Natl Acad Sci U S A 102, 7091–7096 (2005).

89. Med, S. C. T. Bottlenecks in deriving definitive hematopoietic stem cells from human pluripotent stem cells: A CIRM Mini-Symposium and Workshop Report. Stem Cells 3, 775–781 (2014).

90. Seita, J. & Weissman, I. L. Hematopoietic stem cell: self-renewal versus differentiation. Wiley Interdiscip Rev Syst Biol Med 2, 640–653 (2010).

91. Hornke, M. & Mandewirth, S. Merger & acquisitions (M & A) in the pharmaceutical industry: the wheel keeps on turning. Journal of Business Chemistry 7, 67–68 (2010).

92. Bonaccorsi, A., Lyon, T. P., Pammolli, F. & Turchetti, G. Auctions vs. bargaining: An empirical analysis of medical device procurement. Laboratory of Economics and Management, Sant’Anna School of Advanced Studies, Pisa2000 (2003).

93. Einav, L., Farronato, C., Levin, J. D. & Sundaresan, N. Sales Mechanisms in Online Markets: What Happened to Internet Auctions? (National Bureau of Economic Research, 2013).

115

94. Louit, D., Pascual, R., Banjevic, D. & Jardine, A. K. Optimization models for critical spare parts inventories—a reliability approach. Journal of the Operational Research Society 62, 992–1004 (2011).

95. Tersine, R. J. Principles of inventory and materials management. (North-Holland, 1988).

96. Drago, I., Bocchi, E., Mellia, M., Slatman, H. & Pras, A. Benchmarking personal cloud storage. in 205–212 (ACM Press, 2013). doi:10.1145/2504730.2504762

97. Walsh, E. & Cho, I. Using Evernote as an electronic lab notebook in a translational science laboratory. Journal of Laboratory Automation 18, 229–234 (2013).

98. Vanhecke, T. E. Zotero. Journal of the Medical Library Association: JMLA 96, 275 (2008).

99. Heinzman, J. M., Rice, S. D. & Corkan, L. A. Robotic liquid handlers and semiautomated cell quantification systems increase consistency and reproducibility in high-throughput, cell-based assay. Journal of the Association for Laboratory Automation 15, 7–14 (2010).

100. Moon, S.-H. et al. Optimizing human embryonic stem cells differentiation efficiency by screening size-tunable homogenous embryoid bodies. Biomaterials 35, 5987–5997 (2014).

101. Shevde, N. & Mael, A. Techniques in embryoid body formation from human pluripotent stem cells. in Basic Cell Culture Protocols (eds. Helgason, C. D. & Miller, C. L.) 946, 535–546 (Humana Press, 2013).

102. Bosma, M. J. & Carroll, A. M. The SCID mouse mutant: definition, characterization, and potential uses. Annu Rev Immunol 9, 323–350 (1991).

103. Studt, T. Economic Forces Continue to Drive Up Lab Construction Costs. R&D Magazine 50, 38 (2008).

104. Friedman, Y. Science Business: The Promise, the reality, and the future of biotech. By Gary P. Pisano. Biotechnology Journal 1, 1474–1474 (2006).

105. Ledford, H. Life hackers. Nature 467, 650–652 (2010).

106. Lalkaka, R. Technology business incubators to help build an innovation-based economy. Journal of Change Management 3, 167–176 (2002).

116

107. business model canvas crm - Google Search. Available at: https://www.google.com/search?hl=en&q=business+model+canvas+crm&tbm=isch &tbs=simg:CAQSlwEJw- JUvHVDf8UaiwELEKjU2AQaBAgUCAoMCxCwjKcIGmIKYAgDEijrB4QT7Ae DE4UTghPYE-4HgRPfE8Y2lT6sKJY- xzbSP403wTaXPr82GjCM2tDeYfzRN5pbe5lFmu8m6p11e5aNzfOhMQeXtk_1dC vDWvFG1- _1mI2nKrFeTL4kAgBAwLEI6u_1ggaCgoICAESBFBNjwAM&sa=X&ved=0ahU KEwjJoJmftZHgAhWvmOAKHf7-C34Qwg4IKygA&biw=1368&bih=771. (Accessed: 28th January 2019)

108. Wilgenburg, B. van, Browne, C., Vowles, J. & Cowley, S. A. Efficient, long term production of monocyte-ferived macrophages from human pluripotent stem cells under partly-defined and fully-defined conditions. PLoS ONE 8, e71098 (2013).