Background: Pompe disease (PD) is a rare neuromuscular disorder caused by deficiency of acid alpha-glucosidase (GAA), a lysosomal glycogen-catabolizing enzyme. Despite availability of a recombinant human GAA enzyme replacement therapy (rhGAA ERT), clinical unmet needs remain, including suboptimal responses in skeletal muscles caused in part by several key challenges: instability of ERT in circulation and inefficient uptake via the cation-independent mannose 6-phosphate receptor (CI-MPR) at low interstitial concentrations. Once inside cells, GAA requires processing to attain maximal activity for glycogen degradation; however, the relative contributions of proteolytic and N-glycan processing are poorly understood. AT-GAA—an investigational, 2-component therapy comprising cipaglucosidase alfa (a next-generation rhGAA enriched with bis-phosphorylated N-glycans for improved uptake) administered with miglustat (a small molecule stabilizer of cipaglucosidase alfa)—has been demonstrated to significantly improve the PD pathogenic cascade (eg, glycogen reduction, reversal of autophagic dysfunction, and muscle pathology) compared to alglucosidase alfa in Gaa knockout (KO) mice. We demonstrate that N-glycan processing is required for enzyme activation and further describe the relative impact of the 2 components of AT-GAA on observed efficacy in Gaa KO mice.

Objectives: To evaluate rhGAA and modified rhGAAs resistant to N-glycan trimming for processing and enzyme activation. To further characterize the relative effect of each of the individual components of AT-GAA (cipaglucosidase alfa and miglustat) on observed efficacy in Gaa KO mice.

Results: Cipaglucosidase alfa was fully processed, activated, and indistinguishable from mature, endogenous human GAA; GAA variants resistant to N-glycan trimming demonstrated lower activity. In Gaa KO mice, miglustat stabilized cipaglucosidase alfa and preserved its activity in the unfavorable physiological pH of blood following infusion.

Conclusion: Results highlight the importance of improving both rhGAA ERT uptake and preserving intracellular processing to maximize glycogen degradation. In Gaa KO mice, the impact of miglustat on cipaglucosidase alfa stability and activity is demonstrated, which has relevance toward developing an effective treatment for PD.

Enhancing Delivery of Acid Alpha-Glucosidase (GAA) to Skeletal Muscle in Pompe Disease (PD): Key Challenges and Attributes of AT-GAA

Nithya Selvan1, Suresh Venkateswaran1, Jessie Feng1 , Finn Hung1 , Matthew Madrid1 , Nickita Mehta1 , Matthew Graziano1, Nastry Brignol1, Yuliya McAnany1, Ritchie Khanna1 , Su Xu1, Steven Tuske1, Jon Brudvig2, Hung Do1*

1Amicus Therapeutics, Inc. Philadelphia, PA, USA.2Sanford Research, Sioux Falls, SD, USA.

*Corresponding author – hdo@amicusrx.com

Abstract

Dissecting the effect of protein and glycan processing on GAA activation

Schematic of the chemical modification of sialic acids on rhGAA

Glycan and protein processing are essential for optimal GAA activity

E. Pompe patient fibroblasts were incubated with 500 nM unmodified rhGAA, oxidized rhGAA or rhGAA-AOAA for 16 hours at 37°C. Cells were then incubated further in growth media lacking rhGAAs and harvested at indicated time points, lysed and analyzed by Western blotting. RhGAA and modified rhGAAs were differentially processed following internalization as indicated by banding pattern of intermediate forms of the protein.

F. To differentiate between proteolytic and glycan processing , PNGaseF digests were performed to remove all glycans on lysates harvested at 24-hours. Banding patterns indicated that modification of rhGAA blocks glycan processing of modified structures in cellulo though proteolytic processing still occurs.

Oxidized rhGAA and rhGAA-AOAA can still undergo proteolytic processing but not glycan trimming in cellulo.

Conclusions

Dissecting the effect of miglustat on AT-GAA efficacy

Acknowledgements, funding and conflicts of interest

We wish to thank Yi Lun, Adriane Schilling, Anju Nair, Rebecca Soska, M. Osman Sheikh, Nicholas Siano and Renee Krampetz fortheir contributions to the experiments described here. This work was performed at and funded by Amicus Therapeutics Inc. Allauthors, except JB, are current or former employees of Amicus Therapeutics Inc. and hold equity in the company.

Modified rhGAAs deficient in glycan trimming or proteolytic cleavage serve as tools to dissect the effect of the two forms of processing on the efficiency of glycogen degradation by GAA.

Blocking N-glycan trimming leads to only partial activation of GAA. Both proteolytic cleavage and glycan processing are essential for optimal glycogen degradation by GAA.

AT-GAA is a two-component investigational therapy comprised of cipaglucosidase alfa, which is a form of GAA enriched in natural bis-phosphorylated N-glycans for improved uptake into cells, and miglustat, a small molecule chaperone that stabilizes GAA. Like endogenous GAA, cipaglucosidase alfa undergoes full processing within cells.

Cipaglucosidase alfa is the primary contributor of AT-GAA efficacy. While co-administration of miglustat with cipaglucosidase alfa improves treatment, outcomes compared to administration of cipaglucosidase alone, miglustat alone is not sufficient to ameliorate cellular pathologies associated with PD in Gaa KO mice.

A. 4MU-α-Glc is a small synthetic substrate commonly used to assay for GAA activity. GAA 4MU-α-Glc hydrolytic activity was not affected by chemicalmodifications of N-glycans, removal of N-glycans or GAA proteolytic processing.

B. Glycogen hydrolytic activity of precursor rhGAA or fully processed rhGAA compared to the activity of partially processed rhGAA-AOAA (onlyproteolytically processed) or rhGAA-kif (only deglycosylated – site 5 close to active site is fully deglycosylated). Fully processed rhGAA, but notpartially processed forms, showed improved catalytic efficiency compared to precursor unmodified rhGAA. Error bars represent SEM.

C. Unlike partially processed modified rhGAAs, fully processed rhGAA had statistically significant ((p<0.05) one-way ANOVA with Fisher’s LSD post-hoctest) ~4-fold reduced Km for glycogen.

Blocking GAA glycan processing impairs glycogen hydrolytic activity

57 952

N-14

0

N-23

3

N-39

0

N-47

0

N-65

2

N-88

2

N-92

5

57 952

N-14

0

N-23

3

N-39

0

N-47

0

N-65

2

N-88

2

N-92

5

rhGAA

rhGAA - kif

57 952

N-14

0

N-23

3

N-39

0

N-47

0

N-65

2

N-88

2

N-92

5

rhGAA – kif + Endo H

Not Found

Not Found

Glycopeptide did not ionize sufficiently

Glycopeptide did not ionize sufficiently

40-50% site occupancy

40-50% site occupancy

40-50% site occupancy

rhGAA-kif can be deglycosylated without any proteolytic cleavage

rhGAA - kif

rhGAA – kif + Endo H Key: Hexose HexNAc

GlcNAcMannosePeptide

631ILQFNLLGVPLVGADVCGFLGNTSEE656

y2-H

2O y2y3 y4 b6

y6+H

exN

Ac

y7+H

exN

Acb9

y9+H

exN

Ac

b12

y12+

Hex

NAc

y14+

Hex

NAc

b62+ b11

y10+

Hex

NAc

b12

b6

b7

b9

b11

b12

Hex

NAc

1Hex

2

Hex

NAc

1Hex

3

Hex

NAc

1Hex

4

Hex

NAc

1Hex

5

Hex

2 Hex

NAc

1Hex

6

Hex

NAc

1Hex

7

Hex

NAc

1Hex

8

b12y1

4+H

exN

Ac

y17+

Hex

NAc

Modified rhGAAs undergo proteolytic cleavage but not glycan trimming

Glycopeptide analysis of unmodified rhGAA (C.) and modified rhGAA (D.) with representative MS1 and MS2 spectra of the highly abundant trypticglycopeptide, 882NNTIVNELVR891. Modified sialic acids on rhGAA - AOAA were insensitive to neuraminidase activity.

rhGAA - unmodified

Intermediate 95 kD

WB: GAA

WB: Actin

MW (kD) 15010075

5037

25

5037

Processed

rhGAA - AOAArhGAA - oxidized

Intermediate 76 kD

Processed

Precursor 110 kD

Processed

Precursor Processed – 24 h

unmod

ified

oxidi

zed

AOAAun

modifie

dox

idize

dAOAA

untre

ated

oxidi

zed

AOAA

unmod

ified

PNGaseF - - - - - - - + + + +MW (kD)

15010075

5037

5037

WB: GAA

WB: Actin

Intermediate 95 kD

Intermediate 76 kDDeglycosylated intermediate

Deglycosylated intermediate

Precursor 110 kD

Proteolytically cleaved peptide associated with mature 70kDDeglycosylated cleaved peptide

25201510

untre

ated

A Modified rhGAA that undergoes glycan trimming but not proteolytic cleavage

150100

MW (kD)150

100

Bacterial neuraminidase - + - + - + - +

EB : WGA

WB: GAA

N-glycan trimming as well as proteolytic cleavage are required for full GAA activation

Cipaglucosidase alfa undergoes full processing in cellulo

10075

15

50

37

50

37

250150

25

MW (kD)

Intermediate 95 kD

Intermediate 76 kD

Proteolytically cleaved peptide associated with mature 70 kD

WT fibroblasts

Pompefibroblasts

WB: GAA – higher exposure

WB: Actin

15

10075

250150

Proteolytically cleaved peptide associated with mature 70 kD

Intermediate 95 kD

Intermediate 76 kD

WB: GAA – low exposure

WB: GAA – higher exposure

Precursor 110 kD

Precursor 110 kD

Precursor Processed

WT or Pompe patient fibroblasts were incubated with 18.5 nM cipaglucosidase alfa or 500 nM alglucosidase for 16 hours at 37°C. Cells were harvested then lysed and analyzed by Western blotting.

Cipaglucosidase alfa is enriched in bis-phosphorylated N-glycans for improved uptake into cells. Only 18.5 nM cipaglucosidase alfa in the cell culture medium is required to match intracellular GAA levels attained by using 500 nM of the poorly phosphorylated alglucosidase alfa.

20 ng of cipaglucosidase alfa and alglucosidase alfa, as well as lysates of untreated cells were analyzed alongside those of treated cells.

Cipaglucosidase alfa and alglucosidase alfa are processed identically.

Cipaglucosidase alfa is internalized substantially better than the standard of care and is fully processed identical to endogenous

GAA

0 1 2 3 40

20

40

60

80

100

120

Time (Hours)

Nor

mal

ized

GA

A A

ctiv

ity(%

)

Stability in buffer (pH 7.4)

AT-GAA is a two-component investigational therapy comprising of cipaglucosidase alfa and miglustat.

A. Thermal stability of cipaglucosidase alfa alone at neutral (7.4) or acidic (5.2) pH, or in the presence of increasing concentrations of miglustat at pH 7.4. Unfolding of ATB200 was monitored by the increase in the fluorescence of SYPRO Orange as a function of temperature.

B. Time course for ATB200 inactivation (i.e., loss of activity) at 37°C in PBS, pH 7.4 (left) or in human blood ex vivo (right) with and without miglustat.

40 50 60 70 800

20

40

60

80

100

120

Temperature (°C)

Nor

mal

ized

(%)

Syp

ro O

rang

e Fl

uore

scen

ce pH 7.4

pH 7.4 + 10µM miglustat

pH 7.4 + 30µM miglustat

pH 7.4 + 100µM miglustat

pH 5.2

Thermal stability

Stabilizing cipaglucosidase alfa with miglustat – AT-GAA

Miglustat is a small molecule stabilizer of cipaglucosidase alfa

Miglustat improves the protein stability and reduces irreversible inactivation of cipaglucosidase

alfa at neutral/blood pH

Glycogen storage is efficiently reduced by AT-GAA; miglustat has no impact alone

Lysosomal and autophagosomal pathologies are reduced more efficiently by AT-GAA than by miglustat alone

Vehicle Vehicle Alglucosidase alfa Cipaglucosidase alfa + miglustat

Miglustat

Gaa WT Gaa KO

IHC : LAMP1

IHC : LC3 II

Hyperproliferation of lysosomes is a hallmark of PD; LAMP-1 is a lysosomal membrane marker. Lipidated LC3 (LC3 II) is found in the autophagosome membrane and LC3 II positive aggregates in vehicle treated Gaa KO

mice indicate increased autophagy. AT-GAA treatment of Gaa KO mice resulted in substantial reduction in LAMP1 and LC3 II signals in quadriceps compared to

alglucosidase alfa treatment and no reduction was seen in mice treated with miglustat alone. Data shown are representative of images from 7 or 8 animals analyzed per treatment group.

AT-GAA significantly reduces lysosomal build up and autophagy defects in Gaa KO mice

Approximately 12-week-old male Gaa KO mice received two biweekly IV administrations of vehicle, 20 mg/kg alglucosidase alfa, 20 mg/kg cipaglucosidase alfa with miglustat (10 mg/kg miglustat administered orally 30 minutes prior to cipaglucosidase alfa IV injection), or 10 mg/kg miglustat alone (orally).

Tissues were collected 14 days after the second administration, homogenized, and assayed biochemically for glycogen content. Tissue sections were also prepared for IHC (below). Glycogen storage was significantly reduced in quadriceps, heart, and triceps of AT-GAA-treated animals compared to vehicle

treated animals ((* = p<0.05, **=p<0.005, ***=p<0.0005 and ****=p<0.00005) one-way ANOVA with Fisher’s LSD post-hoc test). No reduction was seen in mice treated with miglustat alone. Data are average of measurements from 7 or 8 animals per treatment group and error bars represent SEM.

Co-administration of miglustat with cipaglucosidase alfa improves treatment outcomes but miglustat alone does not impact glycogen reduction in muscles of Gaa KO mice

A. Schematic for periodate oxidation of terminal sialic acid followed by aniline-catalyzed oximation with aminooxyacetic acid (AOAA) introduces anunnatural oxime bond in N-glycans. This could obstruct enzymatic hydrolysis of modified sialic acid by hampering substrate recognition bysialidases/neuraminidases, leading to a block in subsequent glycan processing of modified sialylated complex type N-glycans, as glycandegradation occurs sequentially from the outside-in.

B. RhGAA variants treated with bacterial neuraminidase were subjected to blotting analysis using WGA, which binds to terminal sialic acids.Oxidized rhGAA and rhGAA-AOAA were insensitive to treatment with bacterial neuraminidase in vitro.

0 1 2 3 40

20

40

60

80

100

120

Time (Hours)

Nor

mal

ized

GA

A A

ctiv

ity(%

)

Cipaglucosidase alfa

Cipaglucosidase alfa + 17 µM miglustatCipaglucosidase alfa + 170 µM miglustat

Stability in human blood

A. A stale cell line overexpressing rhGAA was cultured in the presence ofkifunensine to generate rhGAA-kif with mostly high-mannose-type N-glycans.Purified rhGAA-kif was incubated with Endo H for 10 days at 30°C to removehigh mannose type N-glycans while maintaining native structure and activity.LC-MS/MS analysis of rhGAA-kif in parallel with unmodified rhGAA led to thedetection of only high mannose type N-glycans at sites 1-5 but not sites 6 and7, which also had complex glycans. The most abundant glycan structuresobserved at each site

B. Representative Precursor MS1 and HCD MS2 spectra of the glycopeptide, 631ILQFNLLGVPLVGADVCGFLGNTSEE656, bearing the most abundant glycan

0 500 1000 1500 2000 25000.0

0.2

0.4

0.6

0.8

1.0

[4MU-Glc] ( µM)

4MU

rele

ased

(nm

ol /n

M G

AA

/ hr)

0 2000 4000 6000 80000

200

400

600

800

[Glycogen] (µg)

Glu

cose

rele

ased

(µM

/nM

GA

A/ h

r)

rhGAA precursor

rhGAA-AOAA precursor

rhGAA-kif precursor

rhGAA fully processed

rhGAA-AOAA -only proteolytically cleaved

rhGAA-kif + Endo H -only deglycosylated

0

5000

10000

15000

20000

25000

K m ( µ

g)

✱

4MU-Glc kinetics Glycogen kinetics Km glycogen kinetics

A. B.

C. rhGAA

rhGAA + Neuraminidase

D. rhGAA - AOAA

rhGAA - AOAA + Neuraminidase

E. F.

A. B.

are shown. GlcNAc was the only observed structure on sites 1-5, and the most abundant structure on sites 6 and 7 of rhGAA-

structure at Asn 652 (site 5 - in green) in rhGAA-kif rhGAA-kif treated with Endo H are shown. No proteolytic cleavage was observed in Endo H treatment rhGAA-kif . Deglycosylation at site 5 is hypothesized to be necessary for improved access of glycogen to the active site given its proximity to the active site.

A. B. C.

A.

B.

GaaWT

0

100

200

300

400

500

Gly

coge

n(

g/m

gto

talp

rote

in)

Gaa KO

******

****

****

Quadriceps

0

500

1000

Gly

coge

n(

g/m

gto

talp

rote

in)

Gaa KO

GaaWT

********

****

****

Heart

GaaWT

0

200

400

600

Gly

coge

n(

g/m

gto

talp

rote

in)

Gaa KO

******

****

****

**

Triceps

Miglustat

VehicleAlglucosidase alfaCipaglucosidase alfaCipaglucosidase alfa + miglustat

Reference: Xu et al., 2019. JCI Insight.

kif treated with Endo H.