Proteomics 2013, 00, 1–8 1DOI 10.1002/pmic.201300162

Proteomics of CaCO3

biomineral-associated proteins: How to

properly address their analysis

Benjamin Marie1, Paula Ramos-Silva2,3, Frederic Marin2 and Arul Marie4

1 UMR 7245 CNRS/MNHN, Molecules de Communication et d’Adaptation des Micro-organismes, Museum Nationald’Histoire Naturelle, Paris, France

2 UMR 6282 CNRS/uB, Biogeosciences, Universite de Bourgogne, Dijon, France3 Section Computational Science, Informatics Institute, Universiteit van Amsterdam, Amsterdam, The Netherlands4 UMR 7245 CNRS/MNHN, Plateforme de Spectrometrie de Masse et de Proteomique, Museum National d’HistoireNaturelle, Paris, France

In a recent editorial (Proc. Natl. Acad. Sci., 2013 110, E2144–E2146) and elsewhere, questionshave been raised regarding the experimental practices in relation to the proteomic analysisof organic matrices associated to the biomineralized CaCO3 skeletons of metazoans such asmolluscan shells and coral skeletons. Indeed, although the use of new high sensitivity MStechnology potentially allows to identify a greater number of proteins, it is also equally (oreven more) sensitive to contamination of residual proteins from soft tissues, which are in closecontact with the biomineral. Based on our own past and present experimental know-how—observations that are reproducible and coherent with the current understanding of extracellularbiomineralization processes—we are convinced that a careful and appropriate cleaning ofbiominerals prior to any analysis is crucial for accurate proteomic investigations and subsequentpertinent interpretation of the results. Our goal is to alert the scientific community about theassociated bias that definitely should be avoided, and to provide critical recommendations onsample preparation and experimental design, in order to better take advantage of the aptness ofproteomic approaches aiming at improving our understanding of the molecular mechanismsin biomineralization.

Keywords:

Animal proteomics / Biomineralization / Bleaching treatment / Calcifying extracellularmatrix / Protein identification / Sample preparation

Received: April 29, 2013Revised: July 26, 2013

Accepted: August 5, 2013

1 Introduction

Most of metazoan skeletons are produced extracellularly viathe secretion of precursor ions, together with an acellu-lar organic matrix that remains embedded within the ma-ture biomineral structures once deposited. This extracellular

Correspondence: Dr. Benjamin Marie, UMR 7245 CNRS MCAM,Museum National d’Histoire Naturelle, 12 rue Buffon CP39, 75005Paris, FranceE-mail: bmarie@mnhn.frFax: +33-1-40-79-31-35

Abbreviations: ECM, extracellular matrix; NaOCl, sodiumhypochlorite; RLCD, repeat low complexity domain; SOMP, skele-ton organic matrix protein

matrix (ECM) comprises amalgamates of proteins, glycopro-teins, polysaccharides, and lipids, with the proteinaceous frac-tion being the dominant moiety. Although the ECM repre-sents only a small fraction of the biomineral weight (between0.1 and 5% w/w), it is thought to exquisitely regulate the min-eral deposition, and consequently, to play a central role in thewhole biomineralization process [1].

Since the 1990s, ECM proteins extracted from calcified ske-letons of few nonvertebrate metazoan models have been grad-ually characterized by “one-by-one” approaches (for review onechinoderms or mollusks see [2–4]). However, this strategydid not give a complete picture of these ECM protein reper-toires. More recently, thanks to advances in high-throughputgenomic and transcriptomic sequencing of an increasingnumber of nonmodel organisms, proteomic analyses of the

C© 2013 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com

2 B. Marie et al. Proteomics 2013, 00, 1–8

Correspondence concerning this andother Viewpoint articles can be accessedon the journals’ home page at:http://viewpoint.proteomics-journal.de

Correspondence for posting on thesepages is welcome and can also besubmitted at this site.

so-called skeleton organic matrix proteins (SOMPs) extractedafter dissolution of the mineral phase, combined with theinterrogation of nucleic acid datasets has resulted in thedescription of numerous novel proteins from various non-vertebrate metazoan species [5–14].

The combination of global SOMP MS-based proteomicsand transcriptomics/genomics performed by us [6, 7, 10, 12,13] has led to the identification of about 40–60 ECM-specificproteins (depending on the biological model), exhibitingunique primary structures with signal peptide, transmem-brane and repeat low-complexity domains, enzyme and/orECM signatures. In addition, a specific expression of theirtranscripts can be measured in skeleton-secreting tissues orby immunolocalization of the translated proteins [10,12], con-stituting a strong experimental evidence of their involvementin the biomineralization process. Surprisingly, few otherworks have published much larger lists of biomineral ECM-associated proteins that were identified employing similarapproaches (up to 200–300 proteins per model depending onthe taxa [9, 11]). But contrary to our findings, the latter listscontain, in addition to ECM-specific proteins, numerous in-tracellular proteins. In our experiments, these obvious cellconstituent proteins are not observed when biomineral struc-tures are adequately cleaned prior to ECM extraction. Hence,we assert that these proteins should be considered as con-taminants, and not assigned as true SOMPs without furtherinvestigation [15].

2 Contamination of extracellularcalcifying matrices extracted frombiomineral structure by cellularcomponents

During the editorial process of one of our previousmanuscripts on the proteomic investigation of the calcifiedshell layers from the gastropod Haliotis asinina [6], an anony-mous reviewer asked us to justify why we identified onlyECM-specific proteins and no intracellular ones, arguing thatbiomineralization does not take place in a clean room, andthat intracellular proteins are often part of the list of ECM’sproteins, as exemplified by the work pertaining to the calci-fied structures of the sea urchin Strongylocentrotus purpura-tus [5]. Although, intracellular proteins (such as actins andtubulins) can be sometimes detected within the extractableorganic components associated to mineralized tissues [14],we assume that their occurrence results from contamination

by cell constituting proteins of cellular remains [13, 15], andare not embedded or strongly associated to the mineral phase.Indeed, other proteomic analyses on the organic matrices ex-tracted from cleaned otolith (from fish) led to the identifica-tion of only few specific proteins (Table 1), all ECM-related,that appear to be directly involved in the formation of theseinner ear calcified structures [16].

To better illustrate this problem, we report on Table 1 themain recent proteomic approaches applied to study mineral-ized structures of CaCO3 in four metazoan phyla, describingthe cleaning procedure, demineralization and extraction stepstogether with the corresponding database search tools usedfor protein identification. By collecting the number of pro-teins identified in each study and inferring their predictedlocation based on sequence properties [17] it is clear that thecleaning step may, at least, partially, strongly influence thenumber of hits, in particular, it can increase the numberof identifications corresponding to intracellular constituentsand other ubiquitous proteins. We believe however that theseprotein hits can be avoided by an extensive and adapted clean-ing method of the biomineral structures, prior to the extrac-tion of the organic matrix.

In addition to the information described in Table 1, weexplain here the effect of the cleaning procedure typified bytwo examples: the first one deals with the investigation of theSOMP from freshwater gastropod Lymnaea stagnalis (Fig. 1).We demonstrate that some proteins, such as actins, tubu-lins, ATPases, and myosins, are intracellular contaminantsfrom cell fragments, that remain after a “simple bleaching”treatment, but can be removed by an additional drastic clean-ing of the biomineral fine powder with concentrated sodiumhypochlorite (NaOCl) (10%, 5 h) in addition to the umbilicusremoval. The second example refers to the SOMP analysisof the stony coral Acropora millepora [13, 15], for which—similarly to the first example—two bleaching treatments wererequired to remove cellular contaminants. We would like toinsist here on the fact that the publication of SOMP listscontaining such contaminants is misleading and detrimentalfor our understanding of biocalcification mechanisms andto the elaboration of molecular models, since there is cur-rently no evidence that intracellular proteins—no matter theirsubcellular localization—interact directly with the growingbiomineral. Moreover, because it blurs the picture of the di-versity of SOMPs, and the comprehension of ECM functionsin biomineralization processes, this problem aims at beingcarefully appreciated.

3 Why it is crucial to avoid cellularcomponent contamination andgenerate specific protein lists

One major point for understanding organic matrix-mediatedbiomineralization processes is to identify, for a given bio-logical model, all key proteins (the “minimal toolbox”) re-quired for mineralization and their respective functions.

C© 2013 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com

Proteomics 2013, 00, 1–8 3

Ta

ble

1.

Su

mm

ary

of

the

mai

nM

S-b

ased

pro

teo

mic

app

roac

hes

app

lied

inth

ere

cen

tye

ars

(200

5–20

13)

tom

iner

aliz

edst

ruct

ure

so

fca

lciu

mca

rbo

nat

efr

om

met

azo

ano

rig

in.

Sp

ecie

sC

alci

fied

Key

clea

nin

gD

emin

eral

izat

ion

Org

anic

Ext

ract

ion

LC-M

S/M

S/

Dat

aId

enti

fied

Pro

tein

tiss

ue

step

sfr

acti

on

sp

roce

du

rein

terr

oga

tio

nso

urc

ep

rote

ins

loca

lizat

ion

Cn

idar

iaA

cro

po

ram

illep

ora

[13]

Ske

leto

n(1

)W

ash

edfr

agm

ents

inN

aOC

l5%

v/v,

72h

.(2

)Po

wd

er(<

200

�m

)in

NaO

Cl1

%v/

v,5

h.

Ace

tic

acid

10%

v/v

over

nig

ht

at4�

Cu

nti

lpH

4

AS

MC

entr

ifu

gati

on

Ult

rafi

ltra

tio

nD

ialy

sis

LTQ

–FT

/MA

SC

OT

NC

BI

nu

cleo

tid

es(1

0138

0)E

ST

(15

389)

36E

CM

—15

EC

M/M

emb

ran

e—13

Mem

bra

ne—

3In

trac

ellu

lar—

0U

nkn

ow

n—

5A

IM6

×C

entr

ifu

gati

on

wit

hm

illiQ

wat

er

Acr

op

ora

mill

epo

ra[1

3,15

]S

kele

ton

(1)

Was

hed

frag

men

tsin

NaO

Cl5

%v/

v,72

h.

Ace

tic

acid

10%

v/v

over

nig

ht

at4�

Cu

nti

lpH

4

AS

MC

entr

ifu

gati

on

Ult

rafi

ltra

tio

nD

ialy

sis

LTQ

–FT

/MA

SC

OT

NC

BI

nu

cleo

tid

es(1

0138

0)E

ST

(15

389)

52E

CM

—12

EC

M/M

emb

ran

e—7

Mem

bra

ne—

5In

trac

ellu

lar—

14U

nkn

ow

n—

14

Sty

lop

ho

rap

isti

llata

[14]

Ske

leto

n(1

)W

ash

edfr

agm

ents

inN

aOC

l3%

wt/

v,4

h.

(2)

Pow

der

(<15

0�

m)

seco

nd

ble

ach

ing

.

1N

HC

lat

roo

mte

mp

erat

ure

pH

7

AS

MA

IMC

entr

ifu

gati

on

Ace

ton

e90

%C

entr

ifu

gati

on

LTQ

–FT

/X!

Tan

dem

Dra

ftg

eno

me

36E

CM

—7

EC

M/M

emb

ran

e—

0M

emb

ran

e—8

Intr

acel

lula

r—7

Un

kno

wn

—14

Ech

ino

der

mat

aS

tro

ng

ylo

cen

tro

tus

pu

rpu

ratu

s[5

]S

pic

ule

sIs

ola

tio

no

fsp

icu

les

follo

wed

by

cen

trif

uga

tio

nin

ters

per

sed

by

succ

essi

vere

susp

ensi

on

sin

:

Ace

tic

acid

50%

v/v

for

5h

at4�

CA

SM

Dia

lysi

sLT

Q–F

T/

Max

Qu

ant

Pred

icte

dan

no

tate

dp

rote

inm

od

els

(Gle

an3)

231

EC

M–7

2E

CM

/Mem

bra

ne—

40M

emb

ran

e—51

Intr

acel

lula

r—66

Un

kno

wn

—2

(1)

NaO

Cl4

.5%

v/v

for

1–2

min

.(2

)C

aCO

3-sa

tura

ted

wat

er.

(3)

Eth

ano

l,th

enac

eto

ne

(100

%).

Test

and

spin

e

(1)

Test

scu

tin

to2

hal

ves

and

was

hed

.(2

)3

×20

0m

LN

aOC

l(6

–14%

acti

vech

lori

ne)

,10

min

.

Ace

tic

acid

50%

v/v

over

nig

ht

at4�

CA

SM

AS

M

Dia

lysi

sLT

Q-F

T/M

AS

CO

TPr

edic

ted

ann

ota

ted

pro

tein

mo

del

s(G

lean

3)

110

EC

M—

35E

CM

/Mem

bra

ne—

23M

emb

ran

e—20

Intr

acel

lula

r—28

Un

kno

wn

—4

Too

th(1

)4

×20

0m

LN

aOC

l(6

–14%

acti

vech

lori

ne)

,1

h,w

ith

chan

ges

afte

r15

min

wit

ha

2-m

inso

nic

atio

nin

terv

alaf

ter

ever

ych

ang

e.(2

)R

edu

ced

top

ow

der

and

was

hed

agai

nas

in1.

Ace

tic

acid

50%

v/v

over

nig

ht

at4�

CA

SM

Dia

lysi

sLT

Q-F

T/M

AS

CO

TPr

edic

ted

ann

ota

ted

pro

tein

mo

del

s(G

lean

3)

138

EC

M—

49E

CM

/Mem

bra

ne—

24M

emb

ran

e—40

Intr

acel

lula

r—21

Un

kno

wn

—4

C© 2013 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com

4 B. Marie et al. Proteomics 2013, 00, 1–8Ta

ble

1.

Co

nti

nu

ed

Sp

ecie

sC

alci

fied

Key

clea

nin

gD

emin

eral

izat

ion

Org

anic

Ext

ract

ion

LC-M

S/M

S/

Dat

aId

enti

fied

Pro

tein

tiss

ue

step

sfr

acti

on

sp

roce

du

rein

terr

oga

tio

nso

urc

ep

rote

ins

loca

lizat

ion

Mo

llusc

aP

inct

ada

mar

gari

tife

raan

dP

inct

ada

max

ima

[10]

Sh

ell:

Nac

rePr

ism

s

(1)

Inta

ctsh

ells

inN

aOC

l1%

v/v

for

24h

.(2

)S

epar

ated

shel

llay

ers

tho

rou

gh

lyri

nse

dw

ith

wat

er,c

rush

edin

to∼1

-mm

2fr

agm

ents

and

sub

seq

uen

tly

into

fin

ep

ow

der

(>20

0�

m).

Ace

tic

acid

5%v/

vov

ern

igh

tat

4�C

un

tilp

H4.

2

AS

MC

entr

ifu

gati

on

Ult

rafi

ltra

tio

nD

ialy

sis

Q-T

OF/

MA

SC

OT

and

Pro

tein

-Pilo

t

NC

BI

ES

T(7

679

0)–

P.m

arga

riti

fera

ES

T+

nu

cleo

tid

e(7

272)

–P.

max

ima

80E

CM

—51

EC

M/M

emb

ran

e—4

Mem

bra

ne—

6In

trac

ellu

lar—

0U

nkn

ow

n—

19

AIM

6×

Cen

trif

uga

tio

nw

ith

mill

iQw

ater

Hal

ioti

sas

inin

a[6

]S

hel

l:N

acre

Pris

ms

(1)

Inta

ctsh

ells

inN

aOC

l1%

v/v

for

24h

.(2

)S

epar

ated

shel

llay

ers

tho

rou

gh

lyri

nse

dw

ith

wat

er,c

rush

edin

to∼1

-mm

2fr

agm

ents

and

sub

seq

uen

tly

into

fin

ep

ow

der

(>20

0�

m).

Ace

tic

acid

5%v/

vov

ern

igh

tat

4�C

un

tilp

H4.

2

AS

M

AIM

Cen

trif

uga

tio

nU

ltra

filt

rati

on

Dia

lysi

s

6×

Cen

trif

uga

tio

nw

ith

mill

iQw

ater

Q-T

OF/

MA

SC

OT

NC

BI

Nu

cleo

tid

es+

ES

T(9

.167

)

14E

CM

—11

EC

M/M

emb

ran

e—0

TM

—0

Intr

acel

lula

r—0

Un

kno

wn

—3

Cra

sso

stre

ag

igas

[11]

Sh

ell:

Nac

rePr

ism

s

(1)

Inta

ctsh

ells

inN

aOC

l,24

h.

30m

Lo

fac

etic

acid

solu

tio

n5%

un

til

pH

4.0,

stir

red

over

nig

ht.

No

tsp

ecifi

ed

TC

A20

%,2

hC

entr

ifu

gati

on

(3×)

Ace

ton

ean

dce

ntr

ifu

gati

on

LTQ

-FT

/MA

SC

OT

NC

BI

An

no

tate

dp

rote

inm

od

els

(26

086)

259

EC

M—

75E

CM

/Mem

bra

ne—

10M

emb

ran

e—30

Intr

acel

lula

r—13

5U

nkn

ow

n—

9

Lott

iag

igan

tea

[9]

Sh

ell:

Sp

her

ulit

icPr

ism

atic

Cro

ss-

lam

ella

r

(1)

Inta

ctsh

ells

inN

aOC

l(6

–14%

acti

vech

lori

ne)

for

(A)

2h

atR

T,(B

)2

hw

ith

ult

raso

un

d,(

C)

24h

wit

hu

ltra

sou

nd

Ace

tic

acid

50%

v/v

over

nig

ht

at4–

6�C

AS

MA

IM2×

Dia

lysi

sLT

Q-F

T/

Max

Qu

ant

gen

om

e.j

gi-

psf.

org

/

An

no

tate

dp

rote

inm

od

els

(23

851)

311

EC

M—

141

EC

M/M

emb

ran

e—20

TM

—31

Intr

acel

lula

r—89

Un

kno

wn

—30

Lott

iag

igan

tea

[12]

Sh

ell:

Pris

mat

icC

ross

-la

mel

lar

(1)

M+

2,M

+1,

Man

dM

−1

laye

rsw

ere

cru

shed

into

app

roxi

mat

ely

1-m

m2

frag

men

ts(2

)S

hel

lfra

gm

ents

inN

aOC

l1%

v/v,

24h

.

Ace

tic

acid

5%v/

vov

ern

igh

tat

4�C

un

tilp

H4.

2

AS

MC

entr

ifu

gati

on

Ult

rafi

ltra

tio

nD

ialy

sis

Q-T

OF/

MA

SC

OT

NC

BI

Nu

cleo

tid

es+E

ST

(252

091)

gen

om

e.j

gi-

psf.

org

/

An

no

tate

dp

rote

inm

od

els

(23

851)

39E

CM

—31

EC

M/M

emb

ran

e—3

Mem

bra

ne—

3In

trac

ellu

lar—

0U

nkn

ow

n—

2A

IM6

×C

entr

ifu

gati

on

wit

hm

illiQ

wat

er

Ch

ord

ata

Gal

lus

gallu

s[1

8]E

gg

shel

l(1

)Is

ola

ted

egg

shel

lsin

5%E

DTA

and

then

was

hed

wit

hw

ater

.

Ace

tic

acid

10%

v/v

AS

MC

entr

ifu

gati

on

Dia

lysi

sLT

Q-F

T/M

AS

CO

TE

BI

chic

ken

IPIp

rote

inse

qu

ence

dat

abas

e(∼

2577

2)

520

EC

M—

226

EC

M/M

emb

ran

e—43

Mem

bra

ne—

75In

trac

ellu

lar—

140

Un

kno

wn

—36

Dan

iore

rio

and

On

corh

ynch

us

myk

iss

[16]

Oto

lith

(1)

Iso

late

do

tolit

hin

0,65

%so

diu

mhy

po

chlo

rite

,th

enso

nic

ated

and

was

hed

wit

hw

ater

.

ED

TAex

cess

ES

Man

dE

ISM

Cen

trif

uga

tio

nU

ltra

filt

rati

on

Q-T

OF/

MA

SC

OT

NC

BI,

En

se

mb

l

Fish

gen

om

es(∼

210

662

6)

8E

CM

—7

EC

M/M

emb

ran

e—0

Mem

bra

ne—

1In

trac

ellu

lar—

0U

nkn

ow

n—

0

Acc

essi

on

nu

mb

ers

of

iden

tifi

edp

rote

ins

wer

eco

llect

edfr

om

the

liter

atu

rean

dco

rres

po

nd

ing

seq

uen

ced

ata

reso

urc

es.

Th

esu

bce

llula

rlo

cati

on

was

pre

dic

ted

by

mea

ns

of

bio

info

rmat

ics

too

ls(h

ttp

://w

ww

.cb

s.d

tu.d

k/se

rvic

es)a

cco

rdin

gto

the

pro

toco

ldes

crib

edin

[17]

.In

bri

ef,p

rote

inse

qu

ence

sw

ere

anal

yzed

wit

hT

arg

etP

,TM

HM

Mfo

rtr

ansm

emb

ran

ed

om

ain

s,S

ign

alP

for

the

pre

sen

ceo

fp

epti

de

sig

nal

s,G

PIf

or

the

pre

sen

ceo

fG

PIa

nch

ors

and

Sec

reto

meP

for

no

ncl

assi

cals

ecre

tio

n.C

om

ple

tep

rote

inse

qu

ence

sw

ith

ou

tp

ote

nti

alse

cret

ory

and

/or

mem

bra

ne

sig

nat

ure

sw

ere

con

sid

ered

intr

acel

lula

rw

ith

ou

tfu

rth

erch

arac

teri

zati

on

of

thei

rp

red

icte

dlo

cati

on

insi

de

the

cell.

As

for

pro

tein

frag

men

ts,l

oca

lizat

ion

was

pre

dic

ted

by

gen

eo

nto

log

yw

hen

avai

lab

leo

rco

nsi

der

edu

nkn

ow

no

ther

wis

e.

C© 2013 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com

Proteomics 2013, 00, 1–8 5

Figure 1. Removal of organic contaminants from biomineralizedshell structures of Lymnaea stagnalis. Comparison of the proteinsidentified by proteomics on the skeletal organic matrix under twodifferent conditions. “Simple bleaching” consisted of treating theskeletal fragments with NaOCl solution once (5% v/v, 24 h), whilein “extended bleaching” the “simple bleaching” was followedby complete removal of the entire umbilicus shell portion anda subsequent treatment with NaOCl solution (10% v/v, 5 h) onthe skeletal sieved powder (<200 �m). The acetic acid-solubleand -insoluble matrices were digested with trypsin and were in-jected into a Q-Star XL nano-electrospray quadrupole/TOF tan-dem mass spectrometer then protein identifications were per-formed with the Mascot search engine against specific transcrip-tomic datasets.

The proteomic investigations of SOMPs constitute a valu-able approach to reach this goal, and gradually providefundamental insights on the molecular basis of biomineral-ization [6,10,12]. In practice, only the investigation of genuineSOMPs—directly associated with the mineral phaseformation—such as the example of caspartin, an intracrys-talline protein from Pinna nobilis calcitic prisms, can con-tribute to important fundamental discoveries [19]. Moreover,understanding how these proteins have evolved, by a compar-ative approaches on different models, is also likely to providenew insights into the events that supported the organismevolution [12]. In this specific context, the presence of con-taminating cytoskeletal proteins (e.g. actins, tubulin, myosin,. . . ) and more generally of intracellular components (e.g. hi-stones, ribosomal proteins, . . . ) is particularly misleading.The case of actin, one of the most important cytoskeletal pro-teins, is representative of the problem, since it was includedin the so-called SOMP list from different species [14,20], sug-gesting homologous mechanisms associated with biominer-alization processes without further substantial evidence.

For the proteins presenting obvious transmembrane do-mains, it appears that some of them may also participatein the biomineralization process (Table 1), by means of thedomains that are targeted outside the cell. These extracel-lular regions may potentially act in the mineralizing space

being subsequently cleaved by proteases and occluded in thegrowing biomineral. We present evidences supporting thishypothesis in the works of Ramos-Silva et al. [13, 15]. Thesame scenario is conveniently suggested by Mann et al. [5],when performing a global proteomics analysis on the spiculeorganic matrix of the sea urchin.

Until now, we have considered that the data about ECMcontamination did not deserve to be included in our publica-tions that concerns the identification of biomineral-associatedproteins by proteomics [6, 7, 10, 12, 13], however the mislead-ing interpretation of some potential contaminants describedas SOMPS that are being published in an increasing numberof research articles [9, 11, 14, 20, 21], urge us to inform thescientific community about this issue.

4 Our recommendations on how toanalyze proteins from biomineralizedstructures by proteomics

According to the most commonly accepted view, the forma-tion of the metazoan external calcified structures results fromthe secretion of an acellular matrix that remains occludedwithin the biomineral phase once precipitated. Alternatively,some other calcified structures, such as the sclerites producedby soft corals [20], are formed, in a first step, inside intracel-lular vesicles, then being externalized by exocytosis. Duringthese extracellular and/or confined intracellular processes,cellular contaminants can remain entrapped in the miner-alized structures, such as the microcavities present in somemollusk shell layers (Fig. 2A and B). Besides the specimencellular remains, another noticeable source of contaminationis exogenous and is mostly represented by the microorgan-isms that can grow within the exoskeletal cavities, such as thevarious specific endolith algae of stony corals [22] or the bor-ing groups (bacteria, sponge, or algae) that can live in somemollusk shells.

As a first cleaning step, we recommend to mechanicallyremove (by abrasion or sawing) such “potentially rich incontaminant” zones, in order to minimize contamination,or to critically appreciate the data obtained, when primarilyanalyzing biominerals by proteomics. Considering themicrostructural diversity of bioconstructions producedby living organisms, we recommend the development ofoptimized cleaning methods for the analysis of all newbiomaterials, based on the detailed characterization ofthese microstructures. A bleaching treatment in two stepsmay assure the quality of the proteomic data: first one forremoving most of the superficial contaminants, a secondone for “flushing out” the most occluded ones. For instance,we observed that for some biominerals, only a thoroughincubation of skeleton fine powder (<200 �M)—and notsimply of skeletal fragments—in concentrated NaOCl (10%,5 h), before extraction, was fully efficient at removing con-taminants, leaving intact the skeleton associated proteins,i.e. true SOMPs that are part of the “biomineralization

C© 2013 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com

6 B. Marie et al. Proteomics 2013, 00, 1–8

Figure 2. Details of two mollusk shell layers containing void areas that can entrap organic contaminants, as exemplified for the chalkylayer of the Pacific oyster Crassostrea gigas (A) or the bioperforated external prismatic layer of the giant limpet Lottia gigantea (B).These structures potentially entrap endogenous (e.g. cellular debris from the mantle cells) or exogenous (e.g. bacteria or other parasiticmicro-organisms) contaminant organic material.

toolkit” [10,12,13]. We recommend the users to adopt similarcleaning procedures. Rather than using sodium hydroxide orhydrogen peroxide, we recommend NaOCl, which offers thecompromise between being a very effective bleaching agentwithout dissolving the mineral phase [23], but this should betested according to the different biominerals investigated.

Generally, considering the diversity of CaCO3 biominer-als, one should consider that there is not one universal clean-ing protocol for biomineralized structures. Protocols needto be optimized by adjusting the concentration of the clean-ing agent, the duration of the treatment, and the grain sizeof the powder, for proteomic approaches on new biomin-erals. One disadvantage of sample cleaning is that it canalso potentially discard proteins of interest that are “weakly”bound to the mineral phase. However, our previous analy-sis indicates that cytoskeletal and intracellular componentsare the ones that are predominantly removed by NaOCltreatment [13, 15]. Comparative approaches of the samplecleaning procedures would also lead to discriminate betweenbiomineral-associated proteins that are weakly bound aroundthe biomineral without being occluded, and “true” SOMPs,that are specifically integrated inside the mineral phase. Be-side the cleaning procedure, attention should be accorded tothe subsequent analytical steps:

(i) Many known SOMPs present remarkable bias in theiramino acid composition, presence of multiple repeatedlow-complexity domains (RLCDs), higher insolubility,lack of trypsin cleavage sites. This requires the develop-ment of novel protein digestion strategies (e.g. enzyme-,microwave-assisted digestions, or acid hydrolysis) thatallow generating peptides of optimal length for MS/MSanalysis [24].

(ii) It is also worth to note that the specific amino acidcomposition of many SOMP influences the results ofthe peptide-spectrum matches according to chosen frag-mentation technique (CID, higher energy dissociation,electron transfer dissociation), and as well as the perfor-mance of the search engines [25].

(iii) Often, SOMPs present PTMs (e.g. glycosylations, phos-phorylations, . . . ) which can limit the access to the pro-tein cleavage sites, and induce poorer ionization, result-ing in lesser protein identifications. As the amount ofsaccharidic moieties can significantly differ accordingto the investigated species, chemical or enzymatic deg-lycosylation step of biomineral-extracted matrix may berecommended in some cases.

(iv) The completeness of the database for SOMP encodingtranscripts, either from genome or transcriptome as-sembly, is especially critical for the identification of agreater number of SOMPs. The availability of mineral-izing tissue-specific transcriptome datasets representsalso a great advantage. However few examples highlightthat these resources can sometimes be incomplete andnot always sufficient to identify all SOMPs [6, 12].

In addition to the above points, critical analysis of the char-acteristics of the identified proteins is required, which in-cludes the following criteria: the specific expression of theirtranscript in mineralizing tissues, the presence of predictedsignal peptide, and transmembrane- or ecto-domains, theirsequence similarity with other biomineralization proteins,and functional characterization according to experimental ev-idence or, alternatively, to gene ontology determination usingsequence comparison approach.

C© 2013 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com

Proteomics 2013, 00, 1–8 7

5 Conclusion

The purpose of this viewpoint is to alert the scientific commu-nity about the proper approach for analyzing skeletal organicmatrix proteins by proteomics, and to get better benefit ofthis approach. Indeed, incorrect interpretation of contami-nants as genuine SOMPs can potentially blur the biologicalpatterns that may emerge from the analysis, and may maskfascinating evolutionary and/or functional trends. From ourviewpoint, contaminations are technically possible to limit,if not to avoid, but require scrupulous and accurate sampletreatments.

Moreover, it is obvious that the use of more and moresensitive MS technologies (e.g. Orbitrap), leads to greaterprotein scoring but also increases the susceptibility to detectassociated contaminants present even in very small amounts,then the limit between true SOMPs present in small amount,and contaminants, would be even more difficult to draw.

The authors would like to thank Dr. Isabelle Zanella-Cleon(Institut de Biologie et de Chimie des Proteines, Lyon France)for the mass spectrometry analysis, Jerome Thomas (Universitede Bourgogne, Dijon, France) for picture processing. Lymnaeastagnalis shells were provided by Dr. Daniel John Jackson fromGottingen University, Germany.

The authors have declared no conflict of interest.

6 References

[1] Baeuerlein, E., in: Baeuerlein, E. (Ed.), Handbook ofBiomineralization—Biological Aspects and Structure For-mation, Wiley-VCH Verlag GmbH & Co KGaA, Weinheim,Germany 2007, pp. 1–19.

[2] Killian, C. E., Wilt, F. H., Molecular aspects of biomineraliza-tion of the echinoderm endoskeleton. Chem. Rev. 2008, 108,4463–4474.

[3] Marin, F., Luquet, G., Marie, B., Medakovic, D., Molluscanshell protein: primary structure, origin and evolution. Curr.Top. Dev. Biol. 2008, 80, 209–276.

[4] Marin, F., Marie, B., Benhamada, S., Silva, P., et al., ‘Shel-lome’: proteins involved in mollusk shell biomineralization– diversity, functions, in: Watabe, S., Maeyama, K., Naga-sawa, H. (Eds.), Recent Advances in Pearl Research, Terra-pub, Tokyo, Japan 2013, pp. 149–166.

[5] Mann, K., Wilt, F. H., Poustka, A. J., Proteomic analysis ofsea urchin (Strongylocentrotus purpuratus) spicule matrix.Proteome Sci. 2010, 8, 33.

[6] Marie, B., Marie, A., Jackson, D., Dubost, L. et al., Proteomicanalysis of the organic matrix of the abalone Haliotis asininacalcified shell. Proteome Sci. 2010, 8, 54.

[7] Joubert, C., Piquemal, D., Marie, B., Manchon, L. et al., Tran-scriptome and proteome analysis of Pinctada margaritiferacalcifying mantle and shell: focus on biomineralization. BMCGenomics 2010, 11, 613.

[8] Berland, S., Marie, A., Duplat, D., Milet, C. et al., Cou-pling proteomics and transcriptomics for the identifica-tion of novel and variant forms of mollusc shell pro-teins: a study with P. margaritifera. Chembiochem. 2011, 12,950–961.

[9] Mann, K., Edsinger-Gonzales, E., Mann, M., In-depth pro-teomic analysis of mollusc shell: acid-soluble and acid-insoluble matrix of the limpet Lottia gigantea. Proteome Sci.2012, 10, 28.

[10] Marie, B., Joubert, C., Tayale, A., Zanella-Cleon, I. et al.,Different secretory repertoires control the biomineraliza-tion processes of prism and nacre deposition of thepearl oyster shell. Proc. Natl. Acad. Sci. U.S.A. 2012, 109,20986–20991.

[11] Zhang, G., Fang, X., Guo, X., Li, L. et al., The oyster genomereveals stress adaptation and complexity of shell formation.Nature 2012, 490, 49–54.

[12] Marie, B., Jackson, D. J., Ramos-Silva, P., Zanella-Cleon, I.et al., The shell-forming proteome of Lottia gigantea re-veals both deep conservations and lineage-specific novel-ties. FEBS J. 2013, 280, 214–232.

[13] Ramos-Silva, P., Kaandorp, J., Huisman, L., Marie, B. et al.,The skeletal organic matrix of the coral Acropora millepora:the evolution of calcification by cooption and domain shuf-fling. Mol. Biol. Evol. 2013, 30, 2099–2112.

[14] Drake, J. L., Mass, T., Haramaty, L., Zelzion, E. et al., Pro-teomic analysis of skeletal organic matrix from the stonycoral Stylophora pistillata. Proc. Natl. Acad. Sci. U.S.A. 2013,110, 3788–3793.

[15] Ramos-Silva, P., Marin, F., Kaandorp, J., Marie, B., “Biomin-eralization toolkit”: the importance of sample cleaning priorto the characterization of biomineral proteomes. Proc. Natl.Acad. Sci. U.S.A. 2013, 110, E2144–E2146.

[16] Kang, Y-J., Stevenson, A. K., Yau, P. M., Kollmar, R., Sparcprotein is required for normal growth of zebrafish otoliths.JARO 2007, 9, 436–451.

[17] Emanuelsson, O., Brunak, S., von Heijne, G., Nielsen, H., Lo-cating proteins in the cell using TargetP, SignalP and relatedtools. Nat. Protoc. 2007, 2, 953–971.

[18] Hincke, M. T., Nys, Y., Gautron, J., Mann, K. et al., Theeggshell: structure, composition and mineralization. Front.Biosci. 2012, 17, 1266–1280.

[19] Pokroy, B., Fitch, A.N., Marin, F., Kapon, M. et al., Anisotropiclattice distorsions in biogenic calcite induced by intra-crystalline organic molecules. J. Struct. Biol. 2006, 155,96–103.

[20] Rahman, M. A., Shinjo, R., Oomori, T., Worheide, G., Analysisof the proteinaceous components of the organic matrix ofcalcitic sclerites from the soft coral Sinularia sp. PLoS One2013, 8, e58781.

[21] Nemoto, M., Wang, Q., Li, D., Pan, S. et al., Proteomic anal-ysis from the mineralized radular teeth of the giant Pacificchiton, Cryptochiton stelleri (Mollusca). Proteomics 2012, 12,2890–2894.

[22] Le Campion-Alsumard, T., Golubic, S., Hutchings, P., Micro-bial endoliths in skeleton of live and dead corals: Porites

C© 2013 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com

8 B. Marie et al. Proteomics 2013, 00, 1–8

lobata (Moorea, French Polynesia). Mar. Ecol. Prog. Ser.1995, 117, 149–157.

[23] Gaffey, S. J., Bronniman, C. E., Effects of bleaching on or-ganic and mineral phases in biogenic carbonates. J. Sed.Petrol. 1993, 63, 752–754.

[24] Bedouet, L., Marie, A., Berland, S., Marie, B. et al., Proteomicstrategy for identifying mollusc shell proteins using mildchemical degradation and trypsin digestion of insoluble or-

ganic shell matrix: a pilot study on Haliotis tuberculata. Mar.Biotechnol. 2011, 14, 446–458.

[25] Marie, A., Alves, S., Marie, B., Dubost, L. et al., Analysis oflow complex region peptides derived from mollusc shell ma-trix proteins using CID, high-energy collisional dissociation,and electron transfer dissociation on an LTQ-Orbitrap: impli-cations for peptide to spectrum match. Proteomics 2012, 12,3069–3075.

C© 2013 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com