REPORT

Absence of BiP Co-chaperone DNAJC3 CausesDiabetes Mellitus and Multisystemic Neurodegeneration

Matthis Synofzik,1,2,12,* Tobias B. Haack,3,4,12 Robert Kopajtich,3,4,12 Matteo Gorza,3,4 Doron Rapaport,5

Markus Greiner,6 Caroline Schonfeld,1,2,5 Clemens Freiberg,7 Stefan Schorr,6 Reinhard W. Holl,8

Michael A. Gonzalez,9 Andreas Fritsche,10 Petra Fallier-Becker,11 Richard Zimmermann,6

Tim M. Strom,3,4 Thomas Meitinger,3,4 Stephan Zuchner,9 Rebecca Schule,1,2,9 Ludger Schols,1,2,*and Holger Prokisch3,4

Diabetes mellitus and neurodegeneration are common diseases for which shared genetic factors are still only partly known. Here, we

show that loss of the BiP (immunoglobulin heavy-chain binding protein) co-chaperone DNAJC3 leads to diabetes mellitus and wide-

spread neurodegeneration. We investigated three siblings with juvenile-onset diabetes and central and peripheral neurodegeneration,

including ataxia, upper-motor-neuron damage, peripheral neuropathy, hearing loss, and cerebral atrophy. Exome sequencing identified

a homozygous stopmutation inDNAJC3. Screening of a diabetes database with 226,194 individuals yielded eight phenotypically similar

individuals and one family carrying a homozygousDNAJC3 deletion. DNAJC3was absent in fibroblasts from all affected subjects in both

families. To delineate the phenotypic and mutational spectrum and the genetic variability of DNAJC3, we analyzed 8,603 exomes,

including 506 from families affected by diabetes, ataxia, upper-motor-neuron damage, peripheral neuropathy, or hearing loss. This anal-

ysis revealed only one further loss-of-function allele in DNAJC3 and no further associations in subjects with only a subset of the features

of the main phenotype. Our findings demonstrate that loss-of-function DNAJC3 mutations lead to a monogenic, recessive form of

diabetes mellitus in humans. Moreover, they present a common denominator for diabetes and widespread neurodegeneration. This

complements findings from mice in which knockout of Dnajc3 leads to diabetes and modifies disease in a neurodegenerative model

of Marinesco-Sjogren syndrome.

Nonautoimmune diabetes mellitus and neurodegenera-

tion are common disorders for which shared genetic

factors are still only partly known. Monogenic forms of

diabetes include neonatal diabetes (MIM 606176) and

maturity-onset diabetes of the young (MIM 606391),

both of which arise from mutations that primarily

reduce pancreatic b cell function.1 Although monogenic

forms account for only about 1%–2% of diabetes cases,

they provide unique insights into the underlying basic

disease mechanisms, such as endoplasmic reticulum

(ER) stress.2,3 Given that reduced mitigation of ER stress

due to genetic mutations has been shown to cause

both diabetes mellitus and multisystemic neurodegenera-

tion (e.g., in Wolfram syndrome 1 [MIM 222300]3,4), it

might present a shared mechanism linking diabetes with

neurodegeneration.

Mutations leading to loss of the ER protein DNAJC3—

which serves to attenuate late phases of ER stress5—have

been shown to lead to pancreatic b cell failure and diabetes

in mice.6 However, it is still unknown whether mutations

in DNAJC3 (MIM 601184; RefSeq accession number

NM_006260.4) also cause disease in humans. Here, we

1Department of Neurodegenerative Diseases, Hertie-Institute for Clinical Brain

Zentrum fur Neurodegenerative Erkrankungen, 72076 Tubingen, Germany;

Munich, Germany; 4Institute of Human Genetics, Helmholtz Zentrum Munch

Germany; 5Interfaculty Institute of Biochemistry, University of Tubingen, 7207

versity of Saarland, 66421 Homburg, Germany; 7Department of Pediatrics and

University, 37075 Gottingen, Germany; 8Institute for Epidemiology andMedic

ald Foundation Department of Human Genetics and John P. Hussman Institu

Miami, FL 33136, USA; 10Division of Endocrinology, Diabetology, Angiology, N

versity of Tubingen, 72076 Tubingen, Germany; 11Institute of Pathology and12These authors contributed equally to this work

*Correspondence: matthis.synofzik@uni-tuebingen.de (M.S.), ludger.schoels@

http://dx.doi.org/10.1016/j.ajhg.2014.10.013. �2014 by The American Societ

The American

show in two index families that loss-of-functionmutations

in DNAJC3 and the resulting absence of ER protein

DNAJC3 lead to diabetes mellitus and multisystemic

neurodegeneration.

Whole-exome sequencing (WES) was performed in two

affected individuals (61691 and 61695, who are subjects

II.2 and II.4, respectively, in family 1 in Figure 1A) of a

consanguineous Turkish family affected by juvenile-onset

diabetes and central and peripheral neurodegeneration,

including early-onset ataxia, upper-motor-neuron damage,

demyelinating peripheral neuropathy, neuronal hearing

loss, and cerebral atrophy. This procedure and all following

procedures reported in this manuscript were approved by

the institutional review board of the University of Tubin-

gen in Germany (reference number 598/20118O1), and

proper informed consent was obtained from all subjects.

After in-solution capture of exonic sequences (SureSelect

Human All Exon 50 Mb Kit, Agilent Technologies), both

samples were sequenced to an average of 1263 and 1733

coverage on the Illumina platform (Genome Analyzer II3

System). For sequencing statistics, see Table S1 (available

online). We used the Burrows-Wheeler Aligner (version

Research, University of Tubingen, 72076 Tubingen, Germany; 2Deutsches3Institute of Human Genetics, Technische Universitat Munchen, 81675

en, German Research Center for Environmental Health, 85764 Neuherberg,

6 Tubingen, Germany; 6Medical Biochemistry and Molecular Biology, Uni-

Adolescent Medicine, University Medical Center Gottingen, Georg August

al Biometry, University of Ulm, 89081 Ulm, Germany; 9Dr. John T. Macdon-

te for Human Genomics, Miller School of Medicine, University of Miami,

ephrology, and Clinical Chemistry, Department of Internal Medicine, Uni-

Neuropathology, University of Tubingen, 72076 Tubingen, Germany

uni-tuebingen.de (L.S.)

y of Human Genetics. All rights reserved.

Journal of Human Genetics 95, 689–697, December 4, 2014 689

A

B

Figure 1. Pedigrees and DNJAC3 Mutations of the Index Families(A) Pedigrees of the index families and segregation of the mutations identified in DNAJC3. Family 1 includes one unaffected and threeaffected children of unrelated parents, and family 2 has two affected children of first-degree consanguineous parents. The identifiedDNAJC3 variants segregated with the disease in all affected members of the two families, and testing of parents demonstrated that theirrespective mutations are located in trans. Abbreviations are as follows: mut, mutation; and þ, wild-type.(B) Representation of identified DNAJC3 mutations. A schematic of the exon-intron arrangement of DNAJC3 (RefSeq NM_ 006260.4)includes the positions of the stop mutation identified in family 1 (top) and the deletion identified in family 2 (bottom). The electrophe-rogram of the deletion breakpoint was analyzed with primers F1 and R1. Coding regions are indicated as blue boxes.

0.5.8) for read alignment to the human reference assembly

(UCSC Genome Browser hg19) and SAMtools (version

0.1.7) for detection of single-nucleotide variants and small

insertions and deletions. Given the rare combination of

juvenile-onset diabetes and early-onset neurodegenera-

tion, we assumed the causative mutations to be rare and

to alter the protein sequence. We therefore excluded vari-

ants present in 4,517 exomes of control individuals with

unrelated phenotypes and searched for nonsynonymous

variants and splice-site mutations (Table 1). Given the

autosomal-recessive pattern of inheritance, we filtered

for genes harboring potential compound-heterozygous

or homozygous rare variants that were predicted to be

damaging by in silico prediction software tools (SIFT and

PolyPhen-2). This approach left only one gene with muta-

tions shared by both affected individuals: DNAJC3 (Table

1). The homozygous DNAJC3 mutation c.580C>T is pre-

dicted to cause premature truncation (p.Arg194*) of more

690 The American Journal of Human Genetics 95, 689–697, Decemb

than 60% of the protein sequence (Figure 1B). Sanger

sequencing of additional family members showed that all

three affected siblings harbor the c.580C>T mutation in

a homozygous state, whereas the healthy sister (subject

II.1 in Figure 1A) has two wild-type alleles. Testing of par-

ents demonstrated that bothmutations are located in trans

(for electropherograms of the Sanger sequences of all pedi-

gree members, see Figure S1).

To confirm the significance of DNAJC3 mutations in the

pathogenesis of diabetes-neurodegeneration syndromes,

we screened the German multicenter DPV (Diabetes Pa-

tienten Verlaufsdokumentation) registry, which contains

226,194 pediatric and adult individuals with diabetes

from 354 treatment centers. According to the key pheno-

typic characteristics of the affected members of index

family 1 (for a detailed description, see Table 2 and the

paragraph below), we used the following clinical screening

criteria: (1) diabetes onset at age 10–20 years, (2) absence of

er 4, 2014

Table 1. Variants Identified by Exome Sequencing in AffectedIndividuals of Index Family 1

II.2 (61691) II.4 (61695) Present in Both

NSVs not presentin 4,517 controlindividuals

216 238 119

At least twoNSVs per gene

11 14 5 (PDCD11,DNAJC3, KIF23,CPT1C, MAGEC1)

Two NSVspredicted to bedamaging andpresent in both

2 1 1 (DNAJC3)

At least two loss-of-function alleles

1 (DNAJC3) 1 (DNAJC3) 1 (DNAJC3)

Nonsynonymous variants (NSVs) include missense, nonsense, stop-loss, andsplice-site mutations and insertions and deletions. Variants were predicted tobe damaging by at least one out of three software predictions (SIFT, Poly-Phen-2, and MutationTaster).

b cell antibodies, (3) body mass index (BMI) below the

median for age and gender, and (4) ataxia and/or hearing

impairment. This filter left 35 individuals (31 with hearing

impairment, two with ataxia, and two with hearing im-

pairment and ataxia). Because the registry does not

collect biomaterials, these 35 individuals were approached

through their treatment centers (29 centers in total) by

their local clinicians and were asked to provide a DNA sam-

ple for DNAJC3 analysis. DNA could be obtained from

eight index subjects. Conventional Sanger sequencing

(oligonucleotide sequences are available upon request)

revealed no mutation in seven of eight subjects. In one

of the two index subjects with hearing impairment and

ataxia, we failed to amplify exons 6�12 by PCR, whereas

PCR amplification delivered the expected DNA fragments

for exons 1–5 in the same sample and for exons 1–12 in

control samples, suggesting a potential deletion affecting

exons 6–12. Breakpoints of the expected deletion were

characterized by PCR and a primer-walking approach and

were confirmed by Sanger sequencing (Figure 1B). This

72 kb homozygous deletion was also identified in the

affected sibling (subject II.2 of family 2, Figure 1A) and

was heterozygous in the parents (subjects I.1 and I.2 of

family 2, Figure 1A).

To delineate the phenotypic and mutational spectrum

and the genetic variability of DNAJC3, we screened a total

of 8,603 exomes (4,303 from the Hussman Institute for

Human Genomics [Miami] via the Genomes Management

Application7 and 4,300 from the Institute of Human

Genetics [Munich]) for rareDNAJC3 variants. Among these

8,603 exomes, 506 unrelated index subjects with a family

history consistent with recessive disease presented with

at least one of the main phenotypic features of the

DNAJC3 phenotypic cluster: diabetes (n ¼ 30), early-

onset ataxia (age of onset < 30 years; n ¼ 69), hereditary

spastic paraplegia (n ¼ 161), Charcot-Marie-Tooth disease

type 2 (n ¼ 153), or deafness (n ¼ 93). In a first step,

we screened all 8,603 exomes for potential homozygous

The American

or compound-heterozygous variants in DNAJC3 by using

the following filter criteria: low frequency in public data-

bases (minor allele frequency < 1% in the NHLBI Exome

Sequencing Project Exome Variant Server [ESP6500])

and genotype quality > 35. We identified two families

affected by homozygous DNAJC3 missense mutations

(family 1: c.207T>A [p.Asp69Glu]; family 2: c.1060G>C

[p.Glu354Gln]) and one family harboring compound-

heterozygous DNAJC3 missense mutations (c.641C>T

[p.Ala214Val] and c.1037G>A [p.Arg346Gln]), but none

of these mutations was considered to be disease causing

(for details, see Table S2 and Figure S2). Apart from our in-

dex family 1, we did not identify any other family carrying

two predicted loss-of-function variants in DNAJC3. Taken

together, these findings suggest the following notion:

although biallelic loss-of-function mutations in DNAJC3

cause diabetes and a multisystemic combination of ataxia,

peripheral neuropathy, upper-motor-neuron disease, and

hypacusis, they do not seem to be associated with only

one of these features or a small subset of them.

Next, we analyzed the overall occurrence ofDNAJC3 var-

iants to determine the genetic variability of DNAJC3 (same

filters as above, but no inheritance filter). In 8,603 exomes,

we identified 56 alleles with rare DNAJC3 variants, consist-

ing of 41 unique variants (one stop [c.580C>T from index

family 1], one frameshift, and 39 missense; see Table S3).

This observation demonstrates that truncating variants

in DNAJC3 are a very rare event (2 out of 17,206 alleles)

and that genetic variability of DNAJC3 is low overall (56

events out of 17,206 alleles). This notion receives further

independent support from the Residual Variation Intoler-

ance Score (RVIS), which is based on the 6,500 exomes

from NHLBI ESP6500.8 This score provides a measure of

the departure from the (genome-wide) average number of

common functional mutations found in genes with a

similar amount of mutational burden (RVIS ¼ 0 when

the gene has an average number of common functional

variants given its total mutational burden, RVIS <

0 when the gene has less common functional variation

than predicted, and RVIS > 0 when the gene has more).8

DNAJC3 yields a RVIS of �0.47, ranking it among the

23% most intolerant of human genes.8 Genes such as

DNAJC3, which are more intolerant of functional genetic

variation, have been shown to be significantly more likely

than other genes to harbor mutations that cause Mende-

lian diseases.8

DNAJC3 acts as co-chaperone of BiP (immunoglobulin

heavy-chain binding protein), a major ER-localized mem-

ber of the HSP70 family of molecular chaperones, which

reversibly bind to contiguous segments of hydrophobic

amino acids exposed in unfolded lumenal proteins

to impede aggregation and promote adequate folding.9

DNAJC3 is located in virtually all tissues in mice and hu-

mans and has especially high levels in the pancreas and

liver.10 In mice, loss of DNAJC3 leads to hyperglycemia

and glucosuria associated with increasing apoptosis of

pancreatic b cells and reduced insulin levels.6 To validate

Journal of Human Genetics 95, 689–697, December 4, 2014 691

Table 2. Characteristics of Subjects with Homozygous Loss-of-Function DNAJC3 Mutations and Healthy Sister 54829

Index Family 1 Index Family 2

II.2 (61691) II.3 (61693) II.4 (61695) II.1 (54829) II.1 (66050) II.2 (66051)

DNAJC3 variant c.580C>T(p.Arg194*)

c.580C>T(p.Arg194*)

c.580C>T(p.Arg194*)

NA deletion of exons6–12 (p.?)

deletion of exons6–12 (p.?)

Age at examination 39 years 34 years 20 years 41 years 20 years 14 years

Sex male female male female female female

Below-average bodyheight (percentilea)

þ, 152 cm (<3%) þ, 145 cm (<3%) þ, 156 cm (25%) þ, 155 cm (25%) þ, 136 cm (<3%) þ, 143 cm (3%)

Below-average bodyweight (percentilea)

þ, 45 kg (<3%) þ, 38 kg (<3%) þ, 49 kg (10%) �, 65 kg(75%–90%)

þ, 39 kg (<3%) þ, 39 kg (<3%)

Reduced BMI(percentile)

þ, 19.5(5%–10%)

þ, 18.1 (<3%) þ, 20.1(25%–50%)

�, 27.1(75%–90%)

þ, 21.1(25%–50%)

þ, 19.4(25%–50%)

Age at onset ofdiabetes

18 years 18 years 15 years no diabetes 14 years 11 years

Diabetes insulintreatment

þ þ þ � þ þ

HbA1c(ref: 4.3%–6.1%)

7.4% 6.9% 12.1% NA 7.5% 7.9%

IA-2 antibodies(ref: <0.9 U/ml)

0.1 U/ml 0.2 U/ml 0.2 U/ml NA 0.2 U/ml 0.4 U/ml

GAD2 antibodies(ref: <0.9 U/ml)

0.1 U/ml 0.4 U/ml 2.1 U/ml NA 0.1 U/ml 0.1 U/ml

Age at onset ofhypacusis

6 years 27 years 14 years NA 2 years NA

Age at onset of gaitdisturbance

6 years 34 years 19 years NA 2 years 11 years

Afferent ataxia þ þ þ � þ þ

Lower-limb areflexia � þ þ � þ �

Babinski sign þ/þ �/� �/� � þ �

MMSE score 27/30 27/30 25/30 30/30 ND ND

Backward calculation(MMSE subscore)

2/5 2/5 0/5 5/5 ND ND

SARA score 18.5/40 4/40 4/40 0/40 35/40 4/40

MEP not evoked to alllimbs

not evoked to alllimbs

not evoked to alllimbs

ND ND ND

Tibial SEP not evoked not evoked not evoked ND ND ND

Sensory NCSb

Sural(ref: >3.8 mV, >39 m/s)

4.3 mV,29 m/s (YY)

no SNAP (YY) 0.5 mV (YY),25 m/s (YY)

ND no SNAP (YY) no SNAP (YY)

Radial(ref: >16 mV, >50 m/s)

10.9 mV (Y),37 m/s (YY)

8.5 mV (Y),32 m/s (YY)

no SNAP (YY) ND no SNAP (YY) ND

Motor NCSc

Tibial(ref: >5 mV, >40 m/s)

12.8 mV,27 m/s (YY)

13.6 mV,23 m/s (YY)

6.6 mV,23 m/s (YY)

ND 0.2 mV (YY),16 m/s (YY)

9.5 mV,20 m/s (YY)

Ulnar(ref: >5 mV, >50 m/s)

10.5 mV,33 m/s (YY)

14.9 mV,32 m/s (YY)

13.4 mV,29 m/s (YY)

ND 11.6 mV,24 m/s (YY)

ND

Abbreviations are as follows: BMI, body mass index; MEP, motor evoked potential; MMSE, Mini-Mental State Examination16; NA, not applicable; NCS, nerve con-duction study; ND, not done; ref, reference value; SARA, Scale for the Assessment and Rating of Ataxia; SEP, sensory evoked potential; SNAP, sensory nerve actionpotential; Y, reduced value; and YY, severely reduced value.aPercentiles of body weight and body height were taken from reference values for German-born Turkish children and adults (18 years).bSNAPs are given in mV, and sensory nerve conduction velocities are given in m/s.cCompound muscle action potentials are given in mV, and motor nerve conduction velocities are given in m/s.

692 The American Journal of Human Genetics 95, 689–697, December 4, 2014

Figure 2. Immunoblot of DNAJC3The immunoblot shows the absence of DNAJC3 in all affected sub-jects (with blue IDs) of both families, whereas the two heterozy-gous parents (with black IDs) of the siblings in family 2 showednormal protein levels. Immunoblot analysis was performed asdescribed previously.11 DNAJC3 antibody (rabbit, Cell SignalingTechnology) was used at 1:1,000. b-actin antibody (Sigma-Aldrich)was used at 1:10,000.

the predicted loss-of-function character of the identified

mutations in both families, we analyzed DNAJC3 in fibro-

blasts by immunoblotting. Neither full-length nor trun-

cated DNAJC3 was detected in the affected individuals,

whereas the two heterozygous parents showed normal

protein levels (Figure 2). Given the well-established func-

tion of DNAJC3 in the downregulation of ER-associated

proteins involved in the initial ER stress response, as

well as the histopathological changes observed in some

DNAJC3-deficient cell types,6 we investigated the ER

morphology in fibroblasts from our individuals with

DNAJC3 loss of function. No changes in ER morphology

were observed by electron microscopy (Figure S3). This is

in line with the previous finding that histological changes

associated with the absence of DNAJC3 can be found only

in some tissues (e.g., pancreatic islets), but not in others

(e.g., pancreatic acini or other pancreatic parenchyma).6

We analyzed protein trafficking through the secretory

pathway in cell lines from subjects with homozygous

DNJAC3 mutations and healthy control individuals by

monitoring the secretion of Gaussia princeps luciferase

into the culture medium.12 Induction of ER stress by thap-

sigargin (an inhibitor of the ER Ca2þ ATPase) or tunicamy-

cin or glucose deprivation (both of which interfere with

N-linked protein glycosylation) resulted in impaired secre-

tion in mutant and control cells. No significant differences

were noted through induction or recovery from ER stress

(Figure S4), indicating intact protein secretion in fibro-

The American

blasts. In addition, we analyzed subjects’ fibroblasts for a

difference in ER calcium leakage. It has been shown for

HeLa cells that a loss of BiP function leads to increased

ER calcium leakage, as does replacement of BiP by an

altered BiP variant, which cannot productively interact

with its HSP40 co-chaperones.13 However, no increased

ER calcium leakage was observed (Figure S5). Taken

together, these experiments again point to the fact that fi-

broblasts might not be the most appropriate tissue for

testing for DNAJC3-related ER stress and calcium leakage,

most likely because of their low secretory activity and/or

the compensatory mechanisms that are active in the

affected subjects. However, no other tissue (e.g., pancreatic

b cells or neurons) was available for testing.

We aggregated clinical, electrophysiological, and imag-

ing data from all five affected subjects belonging to the

two index families affected by DNAJC3 loss-of-function

mutations. Before identification of the DNAJC3 muta-

tions, the phenotype in family 1 was clinically classified

as a mitochondriopathy, and the phenotype in family 2

was classified as an atypical Shwachman-Bodian-Dia-

mond syndrome (MIM 260400). All five affected subjects

presented with young-onset diabetes melllitus diagnosed

between 10 and 20 years of age, showed increased HbA1c

levels (6.9%–12.1%), and received insulin treatment

(Table 2). IA-2 antibodies were absent in all five subjects,

and GAD2 antibodies were present in four of five subjects

(Table 2). C-peptides—which reflect the degree of residual

pancreatic b cell function—were assessed in fasting serum

and upon response to standardized intravenous applica-

tion with 1 mg of glucagon13 (Table S4). Residual endog-

enous insulin secretion was present in all subjects, and

secretion could be stimulated by glucagon (both are char-

acteristic of most monogenic forms of diabetes14). How-

ever, all subjects showed an at least relative deficit in

insulin secretion, given that fasting C-peptide levels

were low in absolute levels and/or in relation to the

actual glucose levels and HbA1c levels (Table S4). Fasting

C-peptide levels were in the same low range as reported

for other monogenic forms of diabetes (between 100

and 700 pmol/l14). The presence of residual endogenous

insulin stimulation by glucagon in subjects with >3 years

of diabetes history and the absence of autoimmune

antibodies clearly argues against a diabetes mellitus

type 1.15

All subjects exhibited multisystemic central and periph-

eral neurodegeneration. This included early-onset ataxia

of combined cerebellar and afferent origin (age of onset

¼ 2–34 years) and sensorimotor peripheral neuropathy

predominantly of the demyelinating type (Table 2), which

was observed even in the subject with the most recent dis-

ease onset of 6 months earlier (subject II.2 in family 2).

Four of five (80%) individuals developed early-onset senso-

rineural hearing loss (range of onset¼ 2–27 years), and two

of five (40%) showed pyramidal tract signs (Table 2).

Cognitive screening by the Mini-Mental State Examina-

tion16 in family 1 revealed an isolated cognitive deficit in

Journal of Human Genetics 95, 689–697, December 4, 2014 693

Figure 3. MRI and Body Stature of Subjects with DNAJC3 Loss-of-Function MutationsMRI of subjects withDNAJC3 loss-of-functionmutations (A–G) and short stature and low BMI in an exemplary affected subject (H). MRIof index subjects II.2 (61691) from family 1 (A–D) and II.1 (66050) from family 2 (E–G) revealed generalized supra- and infratentorialatrophy pronounced in the cervical and thoracic cord (arrowheads; A, D, and E), cerebellar vermis (arrows; A and E), and crus cerebriand midbrain (arrows; C and F) in both subjects and in the pre- and postcentral gyrus in subject II.2 from family 1 (arrows; B). SubjectII.1 (66050) from family 2 also showed several small T2 hyperintense lesions without contrast enhancement bilaterally in the frontopar-ietal and periventricular regions (arrows; G). In index family 1, affected subject II.3 (61693; H, right) has a lower BMI and shorter staturethan her unaffected sister, II.1 (54829; H, left).

backward calculation of serial sevens, suggesting partial

cerebrocortical dysfunction. All five affected subjects

from the two index families had a below-average body

weight (four were below the third percentile, and one

was at the tenth percentile) and a below-normal BMI

(one was below the third percentile, one was below the

tenth percentile, and three were below the 50th percentile),

whereas both body weight and BMI were above average in

the healthy sister (II.1) from family 1 (each were between

the 75th and 90th percentiles; Table 2; Figure 3H). All five

affected subjects also showed a remarkably short body stat-

ure (four were below the third percentile, and one was at

the 25th percentile; for an example, see Figure 3H), suggest-

ing that it is highly likely that this is also part of the pheno-

type associated withDNAJC3mutations. However, because

body stature was also below average in the healthy sister

(II.1) from family 1, future studies are warranted to confirm

this feature. MRI available for index subjects II.2 (61691) of

family 1 and II.1 (66050) of family 2 revealed generalized

supra- and infratentorial atrophy pronounced in the cervi-

cal cord, cerebellar vermis, cerebellar hemispheres, and

midbrain in both subjects and in the pre- and postcentral

gyrus, crus cerebri, pons, and medulla in subject II.2

(Figure 3). Subject II.1 (66050) of family 2 also showed

several small T2 hyperintense lesions without contrast

enhancement bilaterally in frontoparietal and periventric-

ular regions (Figure 3G).

694 The American Journal of Human Genetics 95, 689–697, Decemb

Dnajc3 knockout causes gradual onset of hyperglycemia

and glucosuria associated with increasing apoptosis of

pancreatic b cells in mice, thus mimicking disease

processes in type 1 and late-stage type 2 diabetes.6 Our

findings demonstrate that—analogous to the mouse

model—homozygous DNAJC3 loss-of-function mutations

cause young-onset diabetes in humans. Moreover, like

Dnajc3-knockout mice,6 humans with DNAJC3 loss-of-

function mutations also show a lower body weight, prob-

ably because of less body fat.6

Our findings also reveal that the absence of DNAJC3 is

associated not only with diabetes mellitus but also with

widespread neurodegeneration (which was not studied in

detail in the Dnajc3-knockout mice6). It has already been

shown for mutations in other genes encoding ER-localized

BiP co-chaperones that they can lead tomultisystemic neu-

rodegeneration. For example, mutations in SIL1 lead to

Marinesco-Sjogren syndrome (MSS [MIM 248800]), a mul-

tisystemic syndrome including early-onset ataxia, cogni-

tive deficits, short stature, and pyramidal tract signs,17–19

thus mirroring main features of the phenotype associated

with DNAJC3 mutations. Both genes not only are linked

phenotypically but also interact on a molecular level.

Within the regulation of the BiP ATP-ADP cycle, DNAJC3

and SIL1 have opposing functions—loss of DNAJC3 has

been shown to ameliorate cerebellar Purkinje cell death

and ataxia in SIL1�/� mice.20

er 4, 2014

Figure 4. Role of DNAJC3 in the UPR and Its Interaction with Other Proteins Associated with Human DiseaseDNAJC3 (brown ovals) belongs to the DNAJ proteins localized in the lumen of the ER. It directly binds and recruits unfolded proteins(twisted black lines) in order to hand them over to the molecular chaperone BiP (green boxes). BiP is a central target of UPR signaling.Inter alia, it interacts with SIL1 (red circles), a nucleotide exchange factor, to restore adequate folding of misfolded proteins. An over-whelming load of misfolded proteins requires more of the available BiP to bind to the exposed hydrophobic regions of these misfoldedproteins. To meet this requirement, BiP dissociates from ER-transmembrane sensors, such as PERK (blue circles). The oligomerization ofPERK and the consequent autophosphorylation of the free luminal domain lead to the activation of PERK. The activated cytosolicdomain of PERK causes translational attenuation by directly phosphorylating the a subunit of the regulating initiator of themRNA trans-lation machinery, eIF-2a (blue boxes). This attenuation of translation further reduces the stress induced by misfolded proteins. Severalmutations in genes encoding proteins of these UPR pathways are well established in human diseases. Mutations in EIFKA3, encodingPERK, cause WRS, which is characterized by young-onset nonautoimmune insulin-requiring diabetes, skeletal dysplasia, and short stat-ure and thereby resembles several hallmarks of the DNAJC3-deficiency phenotype. Defects in SIL1 cause MSS.18,19 Clinical manifesta-tions of MSS include early-onset cerebellar ataxia and short stature, thus also mirroring hallmarks of DNAJC3 disease.

Dysfunctions in the regulation of ER stress might thus

serve as a common pathway linking diabetes mellitus

and neurodegeneration. It has already been shown for mu-

tations in other genes involved in the regulation of ER

stress that they can lead to monogenic diabetes mellitus

and multisystemic neurodegeneration, e.g., for mutations

in the gene encoding ER-localized transmembrane protein

WFS1.3,4 WFS1 neurodegeneration includes early-onset

ataxia, sensorineural hearing loss, and cognitive deficits

(Wolfram syndrome 121,22), thus mirroring several of the

systems also affected by DNAJC3-associated neurodegener-

ation. Likewise, mutations are well established for other

The American

genes encoding proteins directly involved in signaling

of the endoplasmic unfolded protein response (UPR),

which acts up- and downstream of DNAJC3 (Figure 4).

For example, mutations in EIFKA3 (MIM 604032), coding

for the UPR signaling protein PERK (PRKR-like endo-

plasmic reticulum kinase; Figure 4), causeWolcott-Rallison

syndrome (WRS [MIM 226980]), characterized by young-

onset nonautoimmune insulin-requiring diabetes, skeletal

dysplasia, and short body stature and thereby resembling

several hallmarks of the DNAJC3-deficiency phenotype.

Likewise, defects in SIL1 (MIM 608005), which encodes a

protein upstream of DNAJC3 in the UPR (Figure 4), cause

Journal of Human Genetics 95, 689–697, December 4, 2014 695

MSS.17–19 Clinical manifestations of MSS include early-

onset ataxia and short stature, thus again mirroring several

features of the DNAJC3-deficiency phenotype.

In summary, we have identified DNAJC3 mutations as a

cause ofmonogenic juvenile-onset diabetes combinedwith

early-onset multisystemic neurodegeneration. DNAJC3

plays a crucial role in ER protein folding and the UPR, thus

adding evidence that dysfunctions in regulation of ER stress

might serve as a commonpathway linkingdiabetesmellitus

and neurodegeneration. DNAJC3might be linked to neuro-

degeneration in other ER-stress-induced and UPR diseases

(e.g., Wolfram syndrome 1, WRS, and MSS) not only on a

phenotypic level but also on a molecular level.

Supplemental Data

Supplemental Data include five figures and four tables and can be

found with this article online at http://dx.doi.org/10.1016/j.ajhg.

2014.10.013.

Acknowledgments

This studywas supportedby theGermanFederalMinistry of Educa-

tion and Research (BMBF) (mitoNET grant 01GM0867 to H.P. and

grant 01GM1113E to L.S. and D.R.), the European Union (grant

F5-2012-305121 [NEUROMICS] to L.S. and grant PIOF-GA-2012-

326681 [HSP/CMT Genetics] to R.S.), E-Rare grants to GENOMIT

(01GM1207 toH.P.) and EUROSCAR (01GM1206 to L.S.), the Inter-

disciplinary Center for Clinical Research (IZKF) Tubingen (grants

2191-0-0 to M.S. and 1970-0-0 to R.S.), the BMBF Competence

Network for diabetes (FKZ 01GI1106 to R.H.), and the NIH (grants

5R01NS072248, 1R01NS075764, and 5R01NS054132 to S.Z.).

This study was also supported by a grant from the BMBF to the

German Center for Diabetes Research (DZD e.V.) in Munich. We

are grateful to ProfessorHartwigWolburg (Tubingen) for discussion

of the electron-microscopy findings.

Received: August 15, 2014

Accepted: October 28, 2014

Published: November 20, 2014

Web Resources

The URLs for data presented herein are as follows:

Diabetes Patienten Verlaufsdokumentation (DPV), http://www.

d-p-v.eu

Online Mendelian Inheritance in Man (OMIM), http://www.

omim.org

RefSeq, http://www.ncbi.nlm.nih.gov/RefSeq

References

1. Steck, A.K., and Winter, W.E. (2011). Review on monogenic

diabetes. Curr. Opin. Endocrinol. Diabetes Obes. 18, 252–258.

2. Petrova, K., Oyadomari, S., Hendershot, L.M., and Ron, D.

(2008). Regulated association of misfolded endoplasmic retic-

ulum lumenal proteins with P58/DNAJc3. EMBO J. 27, 2862–

2872.

3. Fonseca, S.G., Ishigaki, S., Oslowski, C.M., Lu, S., Lipson, K.L.,

Ghosh, R., Hayashi, E., Ishihara, H., Oka, Y., Permutt, M.A.,

696 The American Journal of Human Genetics 95, 689–697, Decemb

and Urano, F. (2010). Wolfram syndrome 1 gene negatively

regulates ER stress signaling in rodent and human cells.

J. Clin. Invest. 120, 744–755.

4. Fonseca, S.G., Urano, F., Weir, G.C., Gromada, J., and Burcin,

M. (2012). Wolfram syndrome 1 and adenylyl cyclase 8

interact at the plasma membrane to regulate insulin produc-

tion and secretion. Nat. Cell Biol. 14, 1105–1112.

5. Yan, W., Frank, C.L., Korth, M.J., Sopher, B.L., Novoa, I., Ron,

D., and Katze, M.G. (2002). Control of PERK eIF2alpha kinase

activity by the endoplasmic reticulum stress-induced molecu-

lar chaperone P58IPK. Proc. Natl. Acad. Sci. USA 99, 15920–

15925.

6. Ladiges, W.C., Knoblaugh, S.E., Morton, J.F., Korth, M.J.,

Sopher, B.L., Baskin, C.R., MacAuley, A., Goodman, A.G.,

LeBoeuf, R.C., and Katze, M.G. (2005). Pancreatic beta-cell

failure and diabetes in mice with a deletion mutation of the

endoplasmic reticulum molecular chaperone gene P58IPK.

Diabetes 54, 1074–1081.

7. Gonzalez, M.A., Lebrigio, R.F., Van Booven, D., Ulloa, R.H.,

Powell, E., Speziani, F., Tekin, M., Schule, R., and Zuchner, S.

(2013). GEnomes Management Application (GEM.app): a

new software tool for large-scale collaborative genome anal-

ysis. Hum. Mutat. 34, 842–846.

8. Petrovski, S., Wang, Q., Heinzen, E.L., Allen, A.S., and Gold-

stein, D.B. (2013). Genic intolerance to functional variation

and the interpretation of personal genomes. PLoS Genet. 9,

e1003709.

9. Gething, M.J. (1999). Role and regulation of the ER chaperone

BiP. Semin. Cell Dev. Biol. 10, 465–472.

10. Korth, M.J., Lyons, C.N., Wambach, M., and Katze, M.G.

(1996). Cloning, expression, and cellular localization of the

oncogenic 58-kDa inhibitor of the RNA-activated human

and mouse protein kinase. Gene 170, 181–188.

11. Haack, T.B., Gorza, M., Danhauser, K., Mayr, J.A., Haberberger,

B., Wieland, T., Kremer, L., Strecker, V., Graf, E., Memari, Y.,

et al. (2014). Phenotypic spectrum of eleven patients and

five novel MTFMTmutations identified by exome sequencing

and candidate gene screening. Mol. Genet. Metab. 111,

342–352.

12. Badr, C.E., Hewett, J.W., Breakefield, X.O., and Tannous, B.A.

(2007). A highly sensitive assay for monitoring the secretory

pathway and ER stress. PLoS ONE 2, e571.

13. Schauble, N., Lang, S., Jung, M., Cappel, S., Schorr, S., Ulucan,

O., Linxweiler, J., Dudek, J., Blum, R., Helms, V., et al. (2012).

BiP-mediated closing of the Sec61 channel limits Ca2þleakage from the ER. EMBO J. 31, 3282–3296.

14. Murphy, R., Ellard, S., and Hattersley, A.T. (2008). Clinical im-

plications of a molecular genetic classification of monogenic

beta-cell diabetes. Nat. Clin. Pract. Endocrinol. Metab. 4,

200–213.

15. Craig, M.E., Jefferies, C., Dabelea, D., Balde, N., Seth, A., and

Donaghue, K.C. (2014). Definition, epidemiology, and classifi-

cation of diabetes in children and adolescents. Pediatr. Dia-

betes 15 (Suppl 20 ), 4–17.

16. Folstein, M.F., Folstein, S.E., and McHugh, P.R. (1975). ‘‘Mini-

mental state’’. A practical method for grading the cognitive

state of patients for the clinician. J. Psychiatr. Res. 12, 189–198.

17. Krieger, M., Roos, A., Stendel, C., Claeys, K.G., Sonmez, F.M.,

Baudis, M., Bauer, P., Bornemann, A., de Goede, C., Dufke,

A., et al. (2013). SIL1 mutations and clinical spectrum in pa-

tients with Marinesco-Sjogren syndrome. Brain 136, 3634–

3644.

er 4, 2014

18. Senderek, J., Krieger, M., Stendel, C., Bergmann, C., Moser, M.,

Breitbach-Faller, N., Rudnik-Schoneborn, S., Blaschek, A.,

Wolf, N.I., Harting, I., et al. (2005). Mutations in SIL1 cause

Marinesco-Sjogren syndrome, a cerebellar ataxia with cataract

and myopathy. Nat. Genet. 37, 1312–1314.

19. Anttonen, A.K., Mahjneh, I., Hamalainen, R.H., Lagier-Tour-

enne, C., Kopra, O., Waris, L., Anttonen, M., Joensuu, T., Ka-

limo, H., Paetau, A., et al. (2005). The gene disrupted in

Marinesco-Sjogren syndrome encodes SIL1, an HSPA5 cocha-

perone. Nat. Genet. 37, 1309–1311.

The American

20. Zhao, L., Rosales, C., Seburn, K., Ron, D., and Ackerman, S.L.

(2010). Alteration of the unfolded protein response modifies

neurodegeneration in a mouse model of Marinesco-Sjogren

syndrome. Hum. Mol. Genet. 19, 25–35.

21. Synofzik, M., Weiss, D., Erharhaghen, J., Kruger, R., and

Schols, L. (2010). Severe orthostatic dysregulation associated

with Wolfram syndrome. J. Neurol. 257, 1751–1753.

22. Barrett, T.G., Bundey, S.E., and Macleod, A.F. (1995). Neurode-

generation and diabetes: UK nationwide study of Wolfram

(DIDMOAD) syndrome. Lancet 346, 1458–1463.

Journal of Human Genetics 95, 689–697, December 4, 2014 697