cell settings, insulin acutely stimulated an increasein labeling of intermediates in pyrimidine syn-thesis in G9C cells expressing wild-type CAD,and this was blocked by the S6K1 inhibitor (Fig.4E and fig. S7, A and B). Although the synthesisof N-carbamoyl-aspartate increased in responseto insulin in CAD-S1859A mutant–expressingG9C cells, the labeling of pyrimidine intermedi-ates downstream of CAD was not stimulated byinsulin, with thesemetabolites detected at amountssimilar to that in wild-type cells treated with theS6K1 inhibitor (Fig. 4E). These data suggest thatthe S6K1-mediated phosphorylation of S1859enhances the in vivo dihydroorotase (E3) activityof CAD, with additional points of regulation frominsulin and S6K1 possibly affecting upstream stepsin the pathway (fig. S7C). The cells expressingCAD-S1859A were no longer acutely sensitiveto insulin for the stimulated incorporation of denovo–synthesized pyrimidines into RNA andDNA (fig. S7D).

This study demonstrates that mTORC1 servesas a molecular link between growth signals andacute control over pyrimidine synthesis. It is worthemphasizing that mTORC1 and S6K1 are notessential for de novo pyrimidine synthesis per sebut are required to increase flux through this path-way in response to growth-promoting signals,

such as insulin and nutrients. The direct regula-tion of CAD by S6K1 serves as a mechanism toincrease the pool of nucleotides available for theRNA and DNA synthesis that accompanies cellgrowth. In addition to protein and lipid synthesis,pyrimidine synthesis represents another majoranabolic process that is responsive to changes incellular growth conditions through mTORC1signaling.

References and Notes1. M. Laplante, D. M. Sabatini, Cell 149, 274 (2012).2. V. Iadevaia, Y. Huo, Z. Zhang, L. J. Foster, C. G. Proud,

Biochem. Soc. Trans. 40, 168 (2012).3. T. Porstmann et al., Cell Metab. 8, 224 (2008).4. K. Düvel et al., Mol. Cell 39, 171 (2010).5. J. Huang, B. D. Manning, Biochem. J. 412, 179

(2008).6. S. Menon, B. D. Manning, Oncogene 27, (Suppl 2),

S43 (2008).7. A. Radu, V. Neubauer, T. Akagi, H. Hanafusa,

M. M. Georgescu, Mol. Cell. Biol. 23, 6139 (2003).8. M. Mori, M. Tatibana, Methods Enzymol. 51, 111

(1978).9. R. A. Williamson et al., Transplant. Proc. 28, 3088

(1996).10. E. A. Carrey, D. G. Campbell, D. G. Hardie, EMBO J. 4,

3735 (1985).11. P. P. Hsu et al., Science 332, 1317 (2011).12. Y. Yu et al., Science 332, 1322 (2011).13. H. Flotow, G. Thomas, J. Biol. Chem. 267, 3074

(1992).

14. D. R. Alessi, F. B. Caudwell, M. Andjelkovic, B. A. Hemmings,P. Cohen, FEBS Lett. 399, 333 (1996).

15. L. R. Pearce et al., Biochem. J. 431, 245 (2010).16. C. C. Dibble, J. M. Asara, B. D. Manning, Mol. Cell. Biol.

29, 5657 (2009).17. L. M. Graves et al., Nature 403, 328 (2000).18. L. A. Musmanno, R. S. Jamison, R. S. Barnett, E. Buford,

J. N. Davidson, Somat. Cell Mol. Genet. 18, 309 (1992).

Acknowledgements: We thank M. Yuan and S. Breitkopffor technical assistance with MS/MS and D. Patterson,D. Kwiatkowski, and M. Zbinden for cell lines. This work wassupported in part by a grant from the LAM Foundation(I.B.-S.); NIH grants F32-DK095508 (J.J.H.), P01-CA120964(B.D.M. and J.M.A.), and R01-CA122617 (B.D.M.); DanaFarber/Harvard Cancer Center grant P30-CA006516 (J.M.A.);and a Sanofi Innovation Award (B.D.M.). I.B.-S. and J.J.H.conceived and performed all experiments, analyzed allexperimental data, and prepared the manuscript. J.M.A.performed and analyzed the MS/MS experiments. B.D.M.directed the research, reviewed all experimental data, andprepared the manuscript. All authors have reviewed themanuscript and declare no competing financial interests.

Supplementary Materialswww.sciencemag.org/cgi/content/full/science.1228792/DC1Materials and MethodsFigs. S1 to S7Tables S1 to S4References (19–23)

13 August 2012; accepted 25 January 2013Published online 21 February 2013;10.1126/science.1228792

Proteomic Mapping of Mitochondriain Living Cells via SpatiallyRestricted Enzymatic TaggingHyun-Woo Rhee,1*† Peng Zou,1* Namrata D. Udeshi,2 Jeffrey D. Martell,1 Vamsi K. Mootha,2,3,4

Steven A. Carr,2 Alice Y. Ting1,2‡

Microscopy and mass spectrometry (MS) are complementary techniques: The former providesspatiotemporal information in living cells, but only for a handful of recombinant proteins ata time, whereas the latter can detect thousands of endogenous proteins simultaneously, but onlyin lysed samples. Here, we introduce technology that combines these strengths by offeringspatially and temporally resolved proteomic maps of endogenous proteins within living cells.Our method relies on a genetically targetable peroxidase enzyme that biotinylates nearby proteins,which are subsequently purified and identified by MS. We used this approach to identify495 proteins within the human mitochondrial matrix, including 31 not previously linked tomitochondria. The labeling was exceptionally specific and distinguished between inner membraneproteins facing the matrix versus the intermembrane space (IMS). Several proteins previouslythought to reside in the IMS or outer membrane, including protoporphyrinogen oxidase,were reassigned to the matrix by our proteomic data and confirmed by electron microscopy.The specificity of peroxidase-mediated proteomic mapping in live cells, combined with itsease of use, offers biologists a powerful tool for understanding the molecular composition ofliving cells.

We sought to develop a method thatcircumvents the limited specificityand loss of material associated with

organelle purification in traditional mass spec-trometry (MS)–based proteomics. Our approachinvolves tagging the proteome of interest witha chemical handle such as biotin while the cell

is still alive, with all membranes, complexes,and spatial relationships preserved. Thus, werequired a genetically targetable labeling en-zyme that covalently tags its neighbors, butnot more distant proteins, in living cells. Onecandidate is promiscuous biotin ligase (1–3),but its labeling is extremely slow (requiring 24

hours) (fig. S1) (1, 2), and the proposed mech-anism proceeds through a biotin-adenylateester, which has a half-life of minutes, imply-ing a large labeling radius. Horseradish per-oxidase (HRP)–catalyzed nitrene generation isanother possibility (4), but we were unable todetect this labeling (fig. S2), and HRP is in-active when expressed in the mammalian cy-tosol (5).

We recently introduced engineered ascorbateperoxidase (APEX) as a genetic tag for electronmicroscopy (EM) (5). Unlike HRP, APEX isactive within all cellular compartments. In ad-dition to catalyzing the H2O2-dependent polym-erization of diaminobenzidine for EM contrast,APEX also oxidizes numerous phenol deriv-atives to phenoxyl radicals. Such radicals areshort lived (<1 ms) (6, 7), have a small label-ing radius (<20 nm) (8, 9), and can covalentlyreact with electron-rich amino acids such asTyr, Trp, His, and Cys (10–13). This chemis-try forms the basis of tyramide signal ampli-fication (14), but it has not been extended toliving cells.

1Department of Chemistry, Massachusetts Institute of Technol-ogy (MIT), Cambridge, MA 02139, USA. 2Broad Institute ofMIT and Harvard, Cambridge, MA 02142, USA. 3Departmentof Molecular Biology, Massachusetts General Hospital, Boston,MA 02115, USA. 4Department of Systems Biology, HarvardMedical School, Boston, MA 02115, USA.

*These authors contributed equally to this work.†Present address: Ulsan National Institute of Science andTechnology, Ulsan, Korea.‡To whom correspondence should be addressed. E-mail:ating@mit.edu

15 MARCH 2013 VOL 339 SCIENCE www.sciencemag.org1328

REPORTS

on

July

22,

201

3w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

To examine whether APEX could be em-ployed for proteomic labeling (Fig. 1A), we tar-geted APEX to the mitochondrial matrix ofhuman embryonic kidney (HEK) cells and ini-tiated labeling by adding biotin-phenol and 1 mMH2O2 to the cell medium. Labeling was terminatedafter 1 min by cell fixation or lysis. Imaging byconfocal microscopy (Fig. 1B) or stochastic op-tical reconstruction microscopy (STORM) (Fig.1C) (15) showed that biotinylated proteins over-lapped tightly with the mito-APEX construct.Streptavidin blot analysis of cell lysates showed thatnumerous endogenous proteins were biotinylatedin an APEX- and H2O2-dependent manner (Fig.1D and fig. S3).

To test the generality of our approach, we alsoanalyzed other constructs that target APEX todifferent cellular regions (figs. S4 and S5). Sevendifferent cytosol-facing APEX fusions gave dis-tinct “fingerprints” in a streptavidin blot analysis,suggesting that targeted APEX biotinylates onlya subset of cytosolic proteins, probably those inits close vicinity. We performed additional ex-periments to characterize the small-moleculespecificity of APEX (fig. S2), the membrane per-meability of the phenoxyl radical (fig. S6), and

the covalent adducts formed with amino acidsin vitro (fig. S7; see also supplementary mate-rials and methods).

We used mitochondrial matrix-targeted APEXto perform a proteomic experiment. Thoughmitochondria have been extensively character-ized by MS proteomics, all previous studies haveused mitochondrial purification, which is asso-ciated with sample loss and contamination. Con-sequently, the most comprehensive inventory ofmitochondrial proteins (16) integrates MS pro-teomic data with green fluorescent protein imag-ing and computational analysis. Furthermore,proteome-scale maps of the matrix subcompart-ment in mammalian cells contain only a smallnumber of proteins (17), representing very lowcoverage, probably because of the challenge ofenriching for this subcompartment.

Endogenous proteins biotinylated by mito-APEX for 1 min in live HEK cells (as in Fig. 1)were purified using streptavidin beads, digestedto peptides, and identified by tandemMS. We usedstable isotope labeling (18) of experimental andcontrol samples to distinguish between biotinylatedproteins and nonspecific binders (fig. S8). Weperformed two independent replicates, each of

which produced a bimodal distribution of proteinsbased on isotope ratio (fig. S8C). The high-ratiodistributions were strongly enriched for mito-chondrial proteins, so we separated these hits andintersected the results from both replicates toobtain a list of 495 proteins (table S1), which wecall our “matrix proteome.” This list is expectedto contain both soluble matrix proteins and innermitochondrial membrane (IMM) proteins thatcontact the matrix space.

Crossing our matrix proteome with earlierliterature revealed that it was highly enriched forboth mitochondrial and mitochondrial matrixproteins (Fig. 2A). Ninety-four percent (464 pro-teins) had prior mitochondrial annotation, leaving31 “mitochondrial orphans” without any previous-ly known connection to mitochondria (table S2).To further quantify the specificity of our matrixproteome, we examined the components of theelectron-transport chain (Fig. 2C) and the TOM/TIM/PAM protein-import pathway (Fig. 2D), be-cause they are structurally and/or topologicallywell characterized. In our matrix proteome, wedetected only those subunits with exposure tothe matrix space, illustrating the specificity andmembrane-impermeability of our tagging.

Fig. 1. Labeling the mito-chondrial matrix proteomein living cells. (A) Labelingscheme. The APEX peroxidasewas genetically targeted tothe mitochondrial matrixvia fusion to a 24–amino acidtargeting peptide (5). Label-ing was initiated by the ad-dition of biotin-phenol andH2O2 to live cells for 1 min.Cells were then lysed, andbiotinylated proteins wererecovered with streptavidin-coated beads, eluted, separatedon a gel, and identified by MS.The peroxidase-generated phe-noxyl radical is short-lived andmembrane-impermeant and,hence, covalently tags onlyneighboring and not distantendogenous proteins. B, biotin.(B) Confocal fluorescence im-aging of biotinylated proteins(stained with neutravidin) af-ter live labeling of HEK cellsexpressing mito-APEX as in(A). Controls were performedwith either biotin-phenol orH2O2 omitted. DIC, differen-tial interference contrast. (C)Superresolution STORM im-ages showing streptavidinand APEX (AF405/AF647) localization patterns at 22-nm resolution in U2OScells. Samples were reacted as in (B). (D) Gel analysis of biotinylated mito-chondrial matrix proteins, before (lanes 1 to 3) and after (lanes 4 to 6)streptavidin bead enrichment. Samples were labeled as in (B). Substratesare biotin-phenol and H2O2. Mammalian cells have four endogenouslybiotinylated proteins, three of which were observed in the negative con-

trol lanes (2 and 3) of the streptavidin blot. (E) Electron microscopy ofHEK cells expressing mito-APEX. EM contrast was generated by treating fixedcells with H2O2 and diaminobenzidine. APEX catalyzes the polymerization ofdiaminobenzidine into a local precipitate, which is subsequently stained withelectron-dense OsO4 (5). Dark contrast is apparent in the mitochondrial matrix,but not the intermembrane space.

Atransfect

with

mito-APEX

B OH

biotin-phenol

H2O2

1 min

lyse cells streptavidin beads mass spectrometry

B OH

B O

endogenous proteins

B B

B B

neutravidin

(AF647)

mito-APEX (anti-V5)

+DIC

omit biotin-phenol omit H2O2

20 µm

D

C streptavidin (Cy3/Cy5)

mito-APEX (anti-V5) merge

500 nm

mito-APEX: streptavidin bead-enriched

bead only

whole cell lysate

30-

40-

50-

60-

80-

100-

230-

ponceau coomassie streptavidin- HRP

kD

+ +

+

+

+ +

+

+ substrates:

+ +

+

+

+ +

E 100 nm 100 nm

mito-APEX mito-APEX

1 2 3 1 2 3 1 4 5 6

www.sciencemag.org SCIENCE VOL 339 15 MARCH 2013 1329

REPORTS

on

July

22,

201

3w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

To analyze depth of coverage, we checkedour matrix proteome for well-established groupsof soluble matrix proteins (Fig. 2B). We detectedmembers of each group at a rate of 80 to 90%and found nearly identical subsets of proteins ineach of the two replicates, suggesting that cov-erage was high, but for only ~85% of proteins.The proteins we consistently did not detect werenot low-abundance proteins (fig. S8F), nor didthey lack surface-exposed tyrosine residues. Wehypothesize that these proteins were stericallyburied in macromolecular complexes, makingthem inaccessible to the phenoxyl radical.

For a subset of proteins in our proteome, wedetected directly biotinylated peptides (fig. S9and table S4). Tandem MS sequencing showedthat biotin-phenol was conjugated to tyrosine sidechains. In nearly all cases, the biotinylated tyro-sine residue mapped to a surface-exposed siteon a soluble protein or a matrix-exposed site ona transmembrane protein.

Our matrix proteome of 495 proteins pro-vides a number of interesting insights. First, the31 mitochondrial orphans may be newly dis-covered mitochondrial proteins. We selected and

imaged six of these at random and found com-plete or partial mitochondrial localization for allof them (fig. S10). Second, 240 proteins withunknown submitochondrial localization can nowbe assigned by our data to the matrix compart-ment (table S3). Third, we detected six proteinspreviously assigned to the intermembrane space(IMS) or outer mitochondrial membrane (OMM):PPOX, CPOX, PNPT1, CHCHD3, COASY, andSAMM50. To determine if our detection of theseproteins in the matrix was accurate, we performedEM imaging, taking advantage of APEX’s ad-ditional functionality as an EM tag (5). APEXfusions to five of the six proteins showed matrixstaining by EM (Fig. 3 and fig. S11). We wereunable to examine the final protein, SAMM50,because APEX insertion at four different sitesabolished mitochondrial targeting.

The proteins PPOX and CPOX are particu-larly interesting in this group because they cat-alyze two of the later steps in heme biosynthesis(Fig. 3A). Previous studies on purified mitochon-dria or mitoplasts treated with proteases ormembrane-impermeant inhibitors have localizedboth enzymes to the IMS (19–22). Structural anal-

ysis and modeling have indicated that PPOXdocks to ferrochelatase (FECH), the final iron-inserting enzyme of heme biosynthesis, throughthe IMM (23) (Fig. 3F). This model is incon-sistent with our EM data, because we foundthat both the C terminus and amino acid 205 ofPPOX localize to the matrix (Fig. 3C). Our EMdata on CPOX, on the other hand, are consistentwith previous literature, because we found thatresidue 70 localizes to the matrix (explaining thedetection of CPOX in our matrix proteome),whereas the C terminus and residue 120 flankingthe active site localize to the IMS (Fig. 3D). Ourreassignment of PPOX from the IMS to thematrix has implications for the nature of its inter-actions with CPOX and FECH and the mecha-nisms by which its substrates are transported acrossthe IMM.

We have developed a method for mappingthe proteomic composition of cellular organ-elles, using a genetically targetable peroxidasethat catalyzes the generation of short-lived, highlyreactive, and membrane-impermeant radicals inlive cells. With a temporal resolution of 1 min,labeled proteins are harvested and identified by

A B100%

80%

60%

40%

20%

0% i

4

16

2

20

13

66

3

22

4

23

ii iii iv v

i. ii. iii. iv. v.

aminoacyl-tRNA synthetases nucleoid-associated proteins mitoribosomal proteins tricarboxylic acid cycle enzymes amino acid metabolism enzymes

not detected detected

human proteome proteome

1526 464

31 17337

289 224

32

103 4

2 100%

80%

60%

40%

20%

0%

matrix specificity mito specificity

matrix mito proteome proteome

matrix

non-mito mito

OMM IMS matrix

D

PAM

TIM23

TOM

mtHsp70

44 16

23 17

21

50

40

20

22 70

5 6 7

18

Mge1 MPP

matrix

IMS

cytosol

9 8 10

not detected detected

CIMS

Complex II Complex III Complex IV Complex I (core subunits)

ATP synthase

1

2 3

9 5

6

8 7, 10, 11

14 13 12

15

4 matrix

not detected detected

A

B

C D

A B

C

D E+I

F G

H

J

K

A B

C

D E+I

F G

H K

J A

B

C

E F

G

M L

D H I K

A

B

C

E F

K

M

L

D H I

J J

F6

a

b

c

OSCP

d

13

Fig. 2. Specificity and depth of coverage of the mitochondrial matrix proteome.(A) Analysis of specificity. The left two columns show the fraction of proteinswith prior mitochondrial annotation in the entire human proteome (column 1)and in our matrix proteome (column 2). The right two columns show thedistribution of proteins with prior submitochondrial localization information,for all mitochondrial proteins (column 3) and for our matrix proteome (col-umn 4). See table S6 for details. (B) Analysis of depth of coverage. Five groupsof well-established mitochondrial matrix proteins (i to v) were crossed with ourproteomic list. For each group, 80 to 91% of proteins were detected in ourmatrix proteome. See table S7 for details. (C) Analysis of labeling specificity

for protein complexes of the IMM. The subunits of complexes I to IV and F0-F1adenosine triphosphate (ATP) synthase, for which structural information isavailable, are illustrated. Subunits detected in our matrix proteome areshaded red; those not detected are shaded gray. Note that becausestructural information is not available for all 45 subunits of complex I, somesubunits that appear exposed here may not be exposed in the completecomplex. OSCP, oligomycin sensitivity conferral protein. (D) Same analysisas in (C), for proteins of the TOM/TIM/PAM protein-import machinery thatspan the OMM and IMM. All proteins depicted in (C) and (D) are listed,with additional information, in table S8.

15 MARCH 2013 VOL 339 SCIENCE www.sciencemag.org1330

REPORTS

on

July

22,

201

3w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

MS with the use of well-established techniques.In addition to its simplicity, our method has nonoticeable toxicity, requires far less material thanconventional organellar proteomics, and takeshours to implement rather than days (as for sub-cellular fractionation). Our initial demonstrationon the human mitochondrial matrix proteomeshows that specificity is exceptionally high be-cause labeling is performed in living cells whilemembranes and other structures are still intact.Depth of coverage is also high for the majorityof proteins—most likely those that are stericallyaccessible to the phenoxyl radical. Notably, ourmethod provides insight into the topology ofidentified proteins. Finally, the same peroxidase,APEX, can be used for both proteomic mappingand EM visualization (5).

References and Notes1. K. J. Roux, D. I. Kim, M. Raida, B. Burke, J. Cell Biol. 196,

801 (2012).2. B. Morriswood et al., Eukaryot. Cell 12, 356 (2013).3. E. Choi-Rhee, H. Schulman, J. E. Cronan, Protein Sci. 13,

3043 (2004).4. N. Kotani et al., Proc. Natl. Acad. Sci. U.S.A. 105, 7405

(2008).5. J. D. Martell et al., Nat. Biotechnol. 30, 1143 (2012).

6. J. F. Wishart, B. S. Madhava Rao, in Recent Trendsin Radiation Chemistry (World Scientific, Singapore,2010), pp. 577–596.

7. A. Mortensen, L. H. Skibsted, J. Agric. Food Chem. 45,2970 (1997).

8. M. Bendayan, Science 291, 1363 (2001).9. G. Mayer, M. Bendayan, J. Histochem. Cytochem. 45,

1449 (1997).10. K. Minamihata, M. Goto, N. Kamiya, Bioconjug. Chem.

22, 2332 (2011).11. M. S. Rogers et al., Biochemistry 47, 10428 (2008).12. B. Bhaskar et al., J. Mol. Biol. 328, 157 (2003).13. F. Amini, T. Kodadek, K. C. Brown, Angew. Chem. Int.

Ed. Engl. 41, 356 (2002).14. W. Schönhuber, B. Fuchs, S. Juretschko, R. Amann,

Appl. Environ. Microbiol. 63, 3268 (1997).15. B. Huang, M. Bates, X. Zhuang, Annu. Rev. Biochem.

78, 993 (2009).16. D. J. Pagliarini et al., Cell 134, 112 (2008).17. F. Forner, L. J. Foster, S. Campanaro, G. Valle, M. Mann,

Mol. Cell. Proteomics 5, 608 (2006).18. S. E. Ong et al., Mol. Cell. Proteomics 1, 376 (2002).19. G. C. Ferreira, T. L. Andrew, S. W. Karr, H. A. Dailey,

J. Biol. Chem. 263, 3835 (1988).20. J.-C. Deybach, V. da Silva, B. Grandchamp, Y. Nordmann,

Eur. J. Biochem. 149, 431 (1985).21. G. H. Elder, J. O. Evans, Biochem. J. 172, 345 (1978).22. B. Grandchamp, N. Phung, Y. Nordmann, Biochem. J.

176, 97 (1978).23. M. Koch et al., EMBO J. 23, 1720 (2004).24. I. Hamza, H. A. Dailey, Biochim. Biophys. Acta Mol.

Cell Res. 1823, 1617 (2012).

25. R. Nilsson et al., Cell Metab. 10, 119 (2009).26. X. Qin et al., FASEB J. 25, 653 (2011).

Acknowledgments: We thank N. Watson (WhiteheadInstitute Keck Microscopy Facility) and E. Vasile (Koch InstituteMicroscopy Core Facility) for performing EM imaging,H. B. Fraser for assistance with data analysis and manuscriptediting, C. Uttamapinant for picoyl azide-AF647, X. Zhuangand her lab for advice on STORM, H. A. Dailey for advice onCPOX and PPOX, and S. Calvo for assistance with data analysis.Funding was provided by the NIH (grants DP1 OD003961 toA.Y.T. and R01 GM077465 to V.K.M.), the Dreyfus Foundation(A.Y.T.), the American Chemical Society (A.Y.T.), and the BroadInstitute of MIT and Harvard (S.A.C.). The authors have noconflicting financial interests. A patent application relating tothe use of enzymes for proteomic mapping in live cells has beenfiled by MIT. Proteomic data can be found in supplementarytables S1 to S9. The authors will make the genetic constructsused in this work widely available to the academic communitythrough Addgene (www.addgene.org/).

Supplementary Materialswww.sciencemag.org/cgi/content/full/science.1230593/DC1Materials and MethodsFigs. S1 to S11Tables S1 to S9References (27–53)

24 September 2012; accepted 11 January 2013Published online 31 January 2013;10.1126/science.1230593

F

C

matrix

C

N N

FECH

PPOX

previous model

IMS

N

C

23

110

CPOX

80

90

E

N C

205 209

PPOX

matrix

IMS PPOX (2.49) succinyl-CoA

CPOX (1.91)

ALAS1 (3.15)

Fe heme FECH

(2.31)

ALAD HMBS UROS UROD

A B

IMS matrix

PPOX1-477 APEX

CPOX1-453 APEX

C CPOX-APEX APEX205-PPOX APEX120-CPOX PPOX-APEX

200 nm

APEX70-CPOX D

APEX PPOX1-205 PPOX206-477

APEX CPOX121-453 1-120

APEX CPOX71-453 1-70

APEX205-PPOX

PPOX-APEX

APEX70-CPOX

APEX120-CPOX

CPOX-APEX

200 nm

Fig. 3. Submitochondrial localization of the heme biosynthesis enzymesCPOX and PPOX. (A) Model showing the submitochondrial localizations of theeight core enzymes that catalyze heme biosynthesis, according to previousliterature (24). Four of these enzymes are detected in our matrix proteomeand are shown in red [with log2(heavy/light) ratios from replicate 1. CoA,coenzyme A. Drawing adapted from (25). (B) Domain structures of PPOX andCPOX fusions to APEX, imaged by EM in (C) and (D), respectively. AdditionalEM images of PPOX-APEX are shown in fig. S11B. Scale bars in (C) and (D),

200 nm. (E) Our model for PPOX and CPOX localization, based on our EMdata and previous literature (19–22). The predicted membrane-bindingregion of PPOX (residues 92 to 209) is shown in yellow (26). Hollowarrowheads point to predicted cleavage sites in CPOX. (F) Previous modelshowing the docking of a PPOX dimer and a FECH dimer through the IMM(23). N- and C-terminal ends of PPOX are labeled. Our data contradict thismodel, because the EM images in (C) show that the C terminus and residue205 of PPOX are in the matrix, not the IMS.

www.sciencemag.org SCIENCE VOL 339 15 MARCH 2013 1331

REPORTS

on

July

22,

201

3w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

Originally published 31 January 2013; corrected 14 March 2013

www.sciencemag.org/cgi/content/full/science.1230593/DC1

Supplementary Materials for

Proteomic Mapping of Mitochondria in Living Cells via Spatially-

Restricted Enzymatic Tagging

Hyun-Woo Rhee, Peng Zou, Namrata D. Udeshi, Jeffrey D. Martell, Vamsi K. Mootha,

Steven A. Carr, Alice Y. Ting*

*To whom correspondence should be addressed. E-mail: ating@mit.edu

Published 31 January 2013 on Science Express

DOI: 10.1126/science.1230593

This PDF file includes:

Materials and Methods

Figs. S1 to S11

Full Reference List

Other Supplementary Material for this manuscript includes the following:

(available at www.sciencemag.org/cgi/content/full/science.1230593/DC1)

Tables S1 to S9 (in 2 Excel files)

Correction: In the main PDF, “Optimization of proteomic labeling conditions” was

separated into its own section, and a new method for “Localization of biotin-phenol

labeling sites” was added. In Table S4 (contained in an Excel file), biotinylated amino

acid residue numbers were added under the heading “Site Number.” Some abbreviations

were also spelled out for clarity.

Materials and Methods

Characterization of APEX-mediated biotinylation by fluorescence imaging (related to

Figs. 1B, S1, S2, S4, and S7)

HEK293T or COS7 cells were cultured in MEM supplemented with 10% fetal bovine

serum at 37 °C under 5% CO2. Cells were transfected with the APEX fusion construct of

interest using Lipofectamine 2000 (Invitrogen). Typically, 0.2 µg plasmid and 0.5 µL (for

COS7 cells) or 1.0 µL (for HEK293T cells) Lipofectamine in MEM (without serum) was

used for a 7×7 mm coverslip of cells 60-80% confluence. 4-6 hours after transfection, the

cell culture medium was changed back to MEM supplemented with 10% fetal bovine

serum. 18-24 hours after media change, biotin-phenol (final concentration Cf 500 µM)

was added in the cell culture media and the cells were incubated for 30 minutes at 37 °C.

Then H2O2 (Cf 1 mM) was added for 1 minute to initiate biotinylation. To halt the

reaction, the cells were washed three times with a “quencher solution” consisting of 10

mM sodium azide, 10 mM sodium ascorbate, and 5 mM Trolox in DPBS (Dulbecco’s

Phosphate Buffered Saline). Thereafter, the cells were fixed in 4% paraformaldehyde in

PBS at room temperature for 15 minutes. Cells were then washed three times with PBS

and permeabilized with pre-chilled methanol at -20 ºC for 5 minutes.

To detect biotinylated proteins, the fixed HEK293T or COS7 cells were stained

with 50 nM neutravidin-AF (AlexaFluor) or streptavidin-AF, after blocking overnight in

1% casein in PBS (“blocking buffer”) at 4 ºC. Neutravidin and streptavidin fluorophore

conjugates, which could be used interchangeably without a difference in performance,

were prepared by coupling neutravidin protein from Invitrogen or homemade streptavidin

protein (27) with AF647-NHS or AF568-NHS esters (Invitrogen) following the

manufacturer’s instructions.

To detect APEX expression, we also stained the cells with anti-Flag (Sigma;

1:300 dilution), anti-V5 (Invitrogen; 1:2000 dilution), or anti-myc (Millipore; 1:1000

dilution) mouse antibodies in blocking buffer for 1 hour at room temperature. Samples

were then washed 4 × 5 minutes with PBS and incubated with secondary goat anti-

mouse-AF antibody (Invitrogen; 1:1000 dilution) in blocking buffer for 1 hour at room

temperature. Samples were washed 4 × 5 minutes with PBS, then maintained in DPBS at

room temperature for imaging.

Imaging was performed in confocal mode using a Zeiss AxioObserver.Z1

microscope equipped with a Yokogawa spinning disk confocal head and a Cascade II:512

camera. The confocal head contained a Quad-band notch dichroic mirror

(405/488/568/647 nm). Samples were excited by solid state 491 (~20 mW), 561 (~20

mW), or 640 nm (~5 mW) lasers. Images were acquired using Slidebook 5.0 software

(Intelligent Imaging Innovations), through a 63× oil-immersion objective for YFP/AF488

(528/38 emission filter), AF568 (617/73 emission filter), AF647 (700/75 emission filter),

and differential interference contrast (DIC) channels. Acquisition times ranged from 20 to

500 milliseconds. Imaging conditions and intensity scales were matched for each dataset

presented together.

In Fig. S1, mitochondria were stained by incubating cells with 500 nM

Mitotracker Deep Red (Invitrogen) in media for 30 minutes at 37 ºC, prior to cell

fixation.

Characterization of APEX-mediated biotinylation by streptavidin blotting of whole cell

lysate (related to Figs. 1D, S1, S3, and S5)

HEK293T cells transfected and labeled as above were washed 3 times with quencher

solution, then scraped and pelleted by centrifugation at 200 rpm for 5 minutes. The pellet

was stored at -80 °C, then lysed with RIPA lysis buffer (50 mM Tris, 150 mM NaCl,

0.1% SDS, 0.5% sodium deoxycholate, 1% Triton X-100, 1× protease cocktail (Sigma

Aldrich, catalog no. P8849), 1 mM PMSF (phenylmethylsulfonyl fluoride), 10 mM

sodium azide, 10 mM sodium ascorbate, and 5 mM Trolox) for 1 minute at 4 °C. The cell

pellet was resuspended by gentle pipetting. Lysates were clarified by centrifugation at

13,000 rpm for 10 minutes at 4 °C before separation on a 10% SDS-PAGE gel.

For blotting analysis, gels were transferred to nitrocellulose membrane, stained by

Ponceau S (10 minutes in 0.1% w/v Ponceau S in 5% acetic acid/water), and blocked

with “blot blocking buffer” (3% w/v BSA and 0.1% Tween-20 in Tris-buffered saline) at

4 °C overnight. The blots were immersed in 0.3 μg/mL streptavidin-HRP (Thermo

Scientific) at room temperature for 60 minutes, then rinsed with blot blocking buffer 4 ×

5 minutes before development with SuperSignal West Pico reagent (Thermo Scientific)

and imaging on an Alpha Innotech gel imaging system.

Electron microscopy of APEX fusions (related to Figs. 1E, 3, and S11)

HEK293T or COS7 cells were transfected with the APEX fusion construct of interest

using Lipofectamine as described above. Fixation was performed with 2% glutaraldehyde

in PBS as described previously (5). A solution of 3,3’-diaminobenzidine (DAB) was

overlaid, APEX-catalyzed polymerization was allowed to proceed for 5 - 25 minutes, and

DAB polymers were subsequently stained with OsO4 as described previously (5). Cells

were then rinsed 5 × 2 minutes in chilled distilled water and placed in chilled 2% aqueous

uranyl acetate (Electron Microscopy Sciences) for 1 hour. Cells were brought to room

temperature, washed in distilled water, then carefully scraped off the plastic, resuspended,

and centrifuged at 700 g for 1 minute to generate a cell pellet. The supernatant was

removed, and the pellet was dehydrated in a graded ethanol series (70%, 95%, 100%,

100%), for 10 minutes each time, then infiltrated in Eponate resin (Ted Pella) or

EMBED-812 (Electron Microscopy Sciences) using 1:1 (v/v) resin and anhydrous

ethanol for 1 hour, then 2:1 resin:ethanol overnight, then 100% resin for 2 hours. Finally,

the sample was transferred to fresh resin and polymerized at 60 °C for 48 hours.

Embedded cell pellets were cut with a diamond knife into 400 nm sections and imaged on

a FEI-Tecnai™ G² Spirit BioTWIN transmission electron microscope operated at 80 kV.

For the model presented in Fig. 3E, human PPOX (residues 1-477) and human

CPOX (residues 119-453) structures from PDB IDs 3NKS (26) and 2AEX (28) were

used. For the model in Fig. 3F, tobacco PPOX and human FECH structures from PDB

IDs 1SEZ (23) and 2QD1 (29) were used.

Two-color STORM super-resolution imaging (Fig. 1C)

U2OS cells were plated on Lab-Tek™ II Chambered Coverglass (Thermo Scientific) pre-

coated with 10 μg/mL human fibronectin for 30 minutes. At 60-70% confluence, cells

were transfected with mito-APEX, labeled with biotin-phenol and H2O2, and fixed and

permeabilized as described above under “Characterization of APEX-mediated

biotinylation by fluorescence imaging.”

For two-color STORM, mito-APEX was detected by staining with mouse anti-V5

antibody (Invitrogen; 1:2000 dilution) in blocking buffer for 1 hour at room temperature.

Cells were then washed 4 × 5 minutes with PBS and incubated with donkey anti-mouse-

AF405/AF647 conjugate (1:2000 dilution). To detect biotinylated proteins, streptavidin-

Cy3/Cy5 conjugate was added together with the secondary antibody, and the sample was

incubated in blocking buffer for 5 minutes at room temperature. The antibody conjugate

was prepared by coupling donkey anti-mouse antibody (Jackson Immunoresearch) with

AF405-NHS and AF647-NHS esters (Invitrogen). The streptavidin conjugate was

prepared by coupling homemade streptavidin protein (27) with Cy3-NHS and Cy5-NHS

esters (Invitrogen). Labeled cells were washed 4 × 5 minutes with PBS.

Stained samples were imaged in STORM imaging buffer (10% glucose (w/v), 50

mM Tris pH 8.2, 10 mM NaCl, 0.56 mg/mL glucose oxidase, 0.8 mg/mL catalase, and

7.7 mg/mL cysteamine). STORM images were acquired on a home-built setup

constructed on an Olympus IX-81 stand. A polychroic mirror (Di01-R405/488/561/635,

Semrock) was used to reflect lasers onto the sample. A 642 nm laser line operating at

maximum power (140 mW) was used to image AF647 and Cy5; 405 nm and 561 nm

laser lines were used alternately to activate fluorophore pairs, and their intensities (from

0.1 mW to ~10 mW) were adjusted during the course of imaging experiment to maintain

a steady level of fluorophore activation. Samples were illuminated by lasers in the

pseudo-TIRF mode, and images were collected on an EMCCD camera (Evolve,

Photometrics). Super-resolution images were typically reconstructed from more than

10,000 frames of localizations for each channel. Data were analyzed using Insight3

software from Xiaowei Zhuang (Harvard University).

Due to the use of overlapping reporter dyes (AF647 and Cy5) on the antibody and

the streptavidin, we performed controls to check the degree of cross-talk between the two

STORM channels. Control samples were prepared with only antibody labeling, or only

streptavidin labeling, and imaged under identical conditions. By measuring fluorophore

localization numbers we found that AF405/AF647 is activated by the 561 nm laser by

5.2% relative to activation by the 405 nm laser. Conversely, Cy3/Cy5 is activated by the

405 nm laser by 14.6% relative to activation by the 561 nm laser. These values represent

the cross-talk between streptavidin and anti-V5 channels in Fig. 1C.

Proteomic mapping of the mitochondrial matrix

We followed the experimental configuration shown in Fig. S8A. First, to metabolically

label the proteomes of HEK293T cultures with heavy or light isotopes of lysine and

arginine (18), we split early passage HEK293T cells on day 0 into two T25 flasks at 10–

15% confluence. One flask was cultured in light SILAC media: Dulbecco's Modified

Eagle’s Medium (DMEM) deficient in L-arginine and L-lysine (custom preparation from

Caisson Laboratories) that we supplemented with L-arginine (Arg0) and L-lysine (Lys0)

(Sigma-Aldrich) at concentrations of 84 mg/L and 146 mg/L, respectively, 10% dialyzed

fetal bovine serum (Sigma-Aldrich), penicillin, streptomycin, glutamine (Gibco), and 4.5

g/L glucose. The second flask was cultured in heavy SILAC media with the same

composition as above except Arg0 and Lys0 were replaced by L-arginine-13

C6,15

N4

(Arg10) and L-lysine-13

C6,15

N2 (Lys8) (Sigma Isotech). Every two days, before cells

reached confluency, the heavy and light SILAC cultures were split into fresh SILAC

heavy and light media, respectively. After three passages in T25 flasks, on day 6, light

and heavy SILAC cultures were expanded into T75 flasks at 20% confluence and

cultured for two more days. We prepared two T75 flasks (H1, H2) for heavy SILAC

cultures and three T75 flasks (L1, L2, L3) for light SILAC cultures.

On day 9, the H1, H2, L2, and L3 cultures containing 7-8 million cells each were

transfected with 15 µg mito-APEX plasmid using 75 µL Lipofectamine in 30 mL light or

heavy DMEM without serum or antibiotics. 4 hours after transfection, the media was

replaced with fresh light or heavy SILAC media, and the cells were incubated overnight.

On day 10, 18 hours after transfection, 500 µM biotin-phenol in 20 mL SILAC media

was added to the H1, H2, L1, and L3 cultures. Cells were incubated for 30 minutes at 37

°C before addition of H2O2 (final concentration Cf 1 mM) for 1 minute at room

temperature. Cells were then immediately quenched by aspirating off the biotin-

phenol/H2O2 solution and replacing with a “quencher solution” consisting of 10 mM

sodium azide, 10 mM sodium ascorbate, and 5 mM Trolox in 20 mL DPBS. Cells were

washed twice with the quencher solution and twice with DPBS to remove excess biotin-

phenol. For the L2 culture (negative control with biotin-phenol and H2O2 omitted), cells

were washed three times with DPBS. After the final wash, 7 mL of quencher solution

was added to each flask, and cells were collected by gentle pipetting followed by

centrifugation at 200 rpm for 5 minutes at room temperature. Supernatant was discarded,

and the cell pellet was stored at -80 °C.

Cell pellets were lysed by thawing on ice, then adding 300 µL of freshly-prepared

RIPA lysis buffer (50 mM Tris, 150 mM NaCl, 0.1% SDS, 0.5% sodium deoxycholate,

1% Triton X-100, 1× protease cocktail (Sigma Aldrich, catalog no. P8849), 1 mM PMSF,

10 mM sodium azide, 10 mM sodium ascorbate, and 5 mM Trolox) and incubating for 5

minutes at 4 °C. Lysates were clarified by centrifugation at 13,000 rpm for 10 minutes at

4 °C. Protein concentration in the supernatant was measured using the Pierce 660 nm

Protein Assay kit, with bovine serum albumin as the reference standard.

Optimization of proteomic labeling conditions

The optimized proteomic labeling conditions described above were determined after

performing the following tests. To optimize the H2O2 concentration, we first decided to

restrict the H2O2 exposure time to 1 minute, to minimize cellular toxicity and maximize

the temporal resolution of our method. Given this temporal restriction, we found that

lowering the H2O2 concentration to <1 mM significantly reduced the biotinylation signal,

assessed by imaging. We suspect this is the case despite the low Km of APX for H2O2

(~20 µM (30)) and the high membrane permeability of H2O2 because endogenous

catalase activity in mammalian cells rapidly consumes exogenously added H2O2.

To optimize biotin-phenol delivery, we tested loading for 1 minute, 3 minutes, 10

minutes, and 30 minutes and found much stronger labeling with the 30 minute pre-

incubation prior to H2O2 addition. This may be because the biotin-phenol probe has

limited membrane permeability. We also performed an empirical titration and found that

concentrations below 500 µM significantly reduced biotinylation signal, assessed by

imaging.

Finally, we found that it was critical to terminate the labeling after 1 minute with

a quencher solution containing radical quenchers and peroxidase inhibitor. With ordinary

cell lysis buffer, we observed biotinylation of outer mitochondrial membrane proteins

such as TOM20 by matrix-localized APEX, probably because peroxidase-mediated

labeling continues even after dissolution of mitochondrial membranes. Upon addition of

quenchers to the wash and lysis buffers, labeling of TOM20 and other contaminants was

eliminated.

Enrichment of biotinylated proteins with streptavidin beads

The supernatants from lysed SILAC cultures above were combined as shown in Fig.

S8A. H1 and L1 were mixed in a 1:1 ratio (by protein concentration) to give the

Replicate 1 sample, and H2 and L2 were mixed in a 1:1 ratio to give the Replicate 2

sample. In addition, L3 supernatant was spiked in at 5% (by protein concentration) to

each sample to facilitate quantitation of isotope ratios (to avoid infinite H/L values due to

absence of light peptides).

Streptavidin-coated magnetic beads (Pierce catalog no. 88817) were prepared by

washing twice with RIPA lysis buffer (formulation given above). 4 mg of total protein (at

4-5 mg/mL protein concentration) from each replicate sample was mixed with 500 µL of

streptavidin bead slurry. The suspensions were incubated at room temperature for 1 hour

with gentle rotation. Streptavidin beads were then washed with 2 × 1 mL RIPA lysis

buffer, once with 1 mL of 2M urea in 10 mM Tris-HCl pH 8.0, and again with 2 × 1 mL

RIPA lysis buffer. Biotinylated proteins were eluted by incubating the beads with 60 µL

1x NuPAGE LDS Sample Buffer (Invitrogen) supplemented with 20 mM DTT and 2 mM

biotin and heating to 95 °C for 5 minutes.

In-gel protein digestion and extraction

Biotinylated proteins eluted from magnetic streptavidin beads were separated on a

NuPAGE Novex Bis-Tris 4-12% gel run for 1 hour at 130V. The gel was stained

overnight with Coomassie G-250 (Invitrogen) and subsequently destained with water.

Lanes for each replicate were manually cut into 16 gel bands. Gel bands were destained

in 1:1 acetonitrile:100 mM ammonium bicarbonate pH 8.0 for several hours followed by

dehydration with acetonitrile. Dehydrated gel bands were swelled with 100 µL of 10 mM

DTT in 100 mM ammonium bicarbonate for one hour while shaking. The DTT solution

was subsequently removed and 100 µL of 55 mM iodoacetamide in 100 mM ammonium

bicarbonate was added to each gel band and allowed to react for 45 minutes in the dark.

The iodoacetamide solution was removed and bands were dehydrated with acetonitrile.

Approximately 10-50 µL of 20 ng/µL Sequencing Grade Trypsin (Promega) was added

to each sample and digestion was completed overnight with shaking at room temperature.

After the overnight digestion, excess trypsin solution was removed from each

sample and bands were incubated with 20 µL of extraction solution (60%

acetonitrile/0.1% TFA). The extraction solution was collected after 10 minutes and this

process was repeated two additional times. The final peptide extraction was completed by

incubating gel bands with 100% acetonitrile for 1-2 minutes. Extracted peptides were

dried to completeness in a vacuum concentrator, then reconstituted in 100 µL of 0.1%

formic acid and loaded onto C18 StageTips (31). The tips were washed twice with 50 µL

of 0.1% formic acid, and peptides were eluted with 50% acetonitrile/0.1% formic acid,

before drying in a vacuum concentrator.

Liquid chromatography-mass spectrometry

Dried peptides were reconstituted in 0.1% formic acid/3% acetonitrile and analyzed on a

Q Exactive mass spectrometer (Thermo Scientific) coupled online to an Easy-nLC 1000

UPLC (Proxeon). Online chromatographic separation was completed using a capillary

column (360 µm o.d. × 75 µm i.d.) with a 10 µm integrated emitter tip and self-packed

with 23 cm of 1.9 µm ReproSil-Pur C18-AQ resin (Dr. Maisch GmbH). The UPLC

mobile phase A was 0.1% formic acid/3% acetonitrile and the mobile phase B was 0.1%

formic acid/90% acetonitrile. A linear gradient flowing at 200 nL/min and starting at 6%

B and increasing to 30% B over 82 minutes was used to separate peptides with a total run

time of 120 minutes.

The Q Exactive mass spectrometer was operated in the data-dependent mode such

that after each MS1 scan, consecutive high energy collision dissociation MS2 scans were

recorded on the 12 most abundant precursors. The resolution used to acquire MS1 and

MS2 scans was 70,000 and 17,500, respectively. The MS1 advance gain control (AGC)

target was set to 3×106 ions and the maximum ion time was set to 10 msec. MS2 scans

were recorded with an AGC target of 5×104 ions with a maximum ion time of 120 msec,

isolation width of 2.5 m/z, normalized collision energy of 25, dynamic exclusion time of

20 sec, underfill ratio of 5.0%, and peptide match and isotope exclusion enabled.

Proteins and peptides were identified and quantified using the MaxQuant software

package (version 1.2.2.5) (32, 33). Data were searched against the UniProt human

database which contains 81,470 entries. A list of 248 common laboratory contaminants as

provided by the MaxQuant framework was also added to the database. A fixed

modification of carbamidomethylation of cysteine and variable modifications of N-

terminal protein acetylation, oxidation of methionine, and biotin-phenol on tyrosine were

searched. The enzyme specificity was set to trypsin allowing cleavages N-terminal to

proline and a maximum of two missed cleavages was utilized for searching. The

maximum precursor ion charge state was set to 6. The precursor mass tolerance was set to

20 ppm for the first search where initial mass recalibration is completed and 6 ppm for

the main search. The MS/MS tolerance was set to 50 ppm and the top MS/MS peaks per

100 Da was set to 10. The peptide and protein false discovery rates were set to 0.01 and

the minimum peptide length was set to 6. Unmodified, N-terminally acetylated peptides,

and peptides with oxidized methionines were used for protein quantification.

Cut-off analysis to determine the mitochondrial matrix proteome (related to Fig. S8)

Protein information was obtained from the MaxQuant Protein Groups table and is

presented in Tables S1-S3. Proteins identified as reverse hits or contaminants were

removed from the dataset. For each replicate, only proteins identified by three or more

unique peptides with quantified SILAC ratios were retained for further analysis. Protein

SILAC ratios were calculated by MaxQuant as the median of all corresponding peptide

SILAC ratios (32, 33). Protein SILAC ratios from a given replicate were normalized such

that the median H/L value was 1.00.

Proteins identified in each replicate experiment were sorted into one of three

classes: (1) “mito” for those with previous mitochondrial annotation in MitoCarta (16),

the Gene Ontology Cell Component database, or literature; (2) “false-positive” for those

found in a hand-curated list of 2410 non-mitochondrial proteins (16); and (3) neither.

Separate histograms were prepared for class (1) and (2) proteins, binned by log2(H/L)

SILAC ratio, as in Fig. S8C. We calculated the false positive rate (FPR) as a function of

protein SILAC ratio using the equation:

The numerator is the conditional probability of finding a false positive protein in a

particular SILAC ratio range and is calculated as the number of false positive proteins

detected in that SILAC ratio range divided by the total number of false positive proteins

detected in the entire dataset. Likewise, the denominator is the conditional probability of

finding a known mitochondrial protein in a particular SILAC ratio range and is calculated

similarly. The FPR is the ratio of the two conditional probabilities, and is plotted in Fig.

S8D for Replicate 1.

We selected as our cut-off point for each replicate the SILAC ratio at which the

FPR reaches 0.1. All proteins above this cut-off are >10 times more likely to be

mitochondrial than non-mitochondrial. The cut-offs (in log2(H/L) units) for Replicates 1

and 2 were 1.05 and 1.25, respectively. This gave 527 enriched proteins for Replicate 1

and 579 enriched proteins for Replicate 2. We intersected these two lists to obtain the 495

proteins that we define as our mitochondrial matrix proteome.

Localization of biotin-phenol labeling sites

Biotin-phenol site information was obtained from the MaxQuant evidence table and is

presented in Table S4. Only modified peptides with a score >50 were retained for further

analysis. SILAC ratios of modified peptides were not used in this analysis. For a given

biotin-phenol-modified peptide, there can be multiple non-distinct peptides detected (due

to spread in liquid chromatography retention time and charge state); we present the

highest-scoring version across all non-distinct species in Table S4.

Synthesis of biotin substrates (related to Fig. S2)

Biotin conjugate 5 (biotin-XX-tyramide) was purchased from Invitrogen. Biotin

conjugate 8 (EZ-link photoactivatable biotin) was purchased from Pierce. The other

biotin conjugates were synthesized according to this general scheme:

All starting materials were purchased from Sigma Aldrich or Alfa Aesar except 3-

nitrotyramine (PIN Chemicals, UK). Biotin (100 mg) was dissolved in 2 mL DMSO

(dimethyl sulfoxide) at room temperature. Into this solution was slowly added 1.1

equivalents of HATU (2-(7-aza-1H-benzotriazole-1-yl)-1,1,3,3-tetramethyluronium

hexafluorophosphate) and 3.0 equivalents of DIPEA (N,N-diisopropylethylamine). The

mixture was stirred for 10 minutes at room temperature. Then 1.0 equivalent of the

primary amine was added slowly, and the reaction was incubated overnight at room

temperature with stirring. Products were purified by semi-preparative HPLC (Varian

Prostar model 210) using a C18 reverse phase column (Agilent Dynamax 250 × 21.4 mm,

microsorb 300-5 C18) with a 5% to 95% gradient of acetonitrile in water over 30

minutes. Biotin conjugates 1, 2, 3, 4, and 6 eluted at 18-20 minutes. Overall yields were

40-60%.

Biotin conjugate 7 was synthesized by hydrolysis of biotin conjugate 6. Ten

equivalents of NaOH were added to 100 mg of 6 in 1 mL methanol, and the mixture was

allowed to stir overnight at room temperature. Methanol was then removed in vacuo and

7 was purified by HPLC as described above.

MS characterization of purified probes on an Applied Biosystems Q-TRAP

instrument showed:

biotin conjugate 1: calculated for C18H26N3O3S [M+H]+: 364.1695; found: 364.3

biotin conjugate 2: calculated for C18H25N4O5S [M+H]+: 409.1546; found: 409.3

biotin conjugate 3: calculated for C18H26N3O4S [M+H]+: 380.1644; found: 380.4

biotin conjugate 4: calculated for C19H28N3O4S [M+H]+: 394.1801; found: 394.4

biotin conjugate 6: calculated for C20H28N3O5S [M+H]+: 422.1750; found: 422.2

biotin conjugate 7: calculated for C19H26N3O5S [M+H]+: 408.1593; found: 408.4

1H-NMR for biotin conjugate 1 (500 MHz, d6-DMSO): δ 1.27 (2H, m), 1.42-1.61

(4H, m), 2.04 (2H, t, J = 7.5 Hz ), 2.57 (3H, m), 2.79 (1H, dd, J1 = 7.5 Hz, J2= 5 Hz),

3.05 (1H, dd, J1 = 6 Hz, J2= 2 Hz), 3.18 (2H, dd, J1= 6.5 Hz, J2= 6.5 Hz), 4.12 (1H, m),

4.30 (1H, m), 6.46 (1H, s), 6.56 (1H, s), 6.67 (2H, d, J = 8 Hz), 6.96 (2H, d, J = 8 Hz),

7.85 (1H, t, J = 5.5 Hz). 13

C-NMR for biotin conjugate 1 (125 MHz, d6-DMSO): δ

25.502, 28.175, 28.358, 34.557, 35.396, 40.062, 55.625, 59.456, 61.283, 115.283,

129.618, 140.697, 155.921, 163.161, 172.288, 192.483.

Imaging assay to determine the membrane-permeability of the phenoxyl radical (Fig. S6)

HEK293T cells were transfected with W41A

APX targeted to either the cytosol (NES) or

inner leaflet of the plasma membrane (CAAX). 24 hours after transfection, cells were

incubated with 500 µM alkyne-phenol and 1 mM H2O2 for 1 minute at room temperature.

After washing cells three times with DPBS, click chemistry was performed on the cell

surface by incubating with 10 µM AF647-picolyl azide conjugate and a pre-formed

mixture of 200 µM CuSO4, 1 mM THPTA ligand, and 2.5 mM sodium ascorbate in

DPBS for 10 minutes at room temperature (34). Cells were quickly washed with DPBS

three times and immediately imaged live by confocal microscopy as described above.

For comparison, click chemistry was also performed on fixed and permeabilized

cells, in order to visualize the total population of alkyne-phenol-labeled molecules. For

this labeling, cells were first blocked after fixation using 1% bovine serum albumin + 1%

casein in PBS for 30 minutes at room temperature. Then cells were treated with 2.5 µM

AF647-picolyl azide conjugate and a pre-formed mixture of 1 mM CuSO4, 100 µM

TBTA ligand, and 2.5 mM sodium ascorbate in the same blocking buffer for 1 hour at

room temperature.

Gel filtration chromatography (Fig. S7B)

APEX and wtAPX were expressed in BL21-DE3 E. coli and purified by nickel affinity

chromatography as previously described (5). Protein concentrations were measured using

the BCA assay (Pierce). We found that purified APEX had an Abs405/Abs280 ratio > 2.0,

indicating high heme occupancy (5). Purified wtAPX had a lower ratio than this, so we

reconstituted the enzyme with purified heme as previously described (5) to obtain a final

Abs405/Abs280 ratio of 2.3. Gel filtration chromatography was performed on a Waters

HPLC system using a Superdex 75 10/300 column (GE Healthcare) as described

previously (5).

In vitro analysis of biotin-phenol reaction with purified amino acids (Fig. S7C)

Reactions were assembled as follows: 500 µM amino acid (all 20 were purchased from

Sigma), 500 µM biotin-phenol, 1 mM H2O2, and 10 nM HRP (Sigma catalog no. P6782)

in PBS pH 7.4. HRP was added last to initiate the reaction. After 10 minutes at room

temperature, reactions were quenched by addition of sodium azide (Cf 1 mM). Product

mixtures were analyzed by LC-MS. The LC was a Shimadzu UFLC-XR equipped with a

4.6 × 50 mm C18 column (Shimadzu catalog no. 220-91394-00) running a gradient of

5% to 95% acetonitrile in 0.1% trifluoroacetic acid/water over 20 minutes. The MS was

an Applied Biosystems Q-TRAP instrument.

Imaging analysis of mitochondrial orphans (Fig. S10)

Six mitochondrial orphan genes (TRMT61B, TRUB2, RPUSD3, PREPL, PNPO,

TRMT11) were obtained with C-terminal V5 tags in the pLX304 mammalian expression

vector from the Broad hORFeome library. COS7 cells were transfected with these

expression plasmids using Lipofectamine 2000. 24 hours later, cells were fixed and

stained with anti-V5 antibody as described above under “Characterization of APEX-

mediated biotinylation by fluorescence imaging”. To visualize mitochondria, cells were

also stained with rabbit anti-Tom20 antibody (Santa Cruz Biotechnology; 1:400 dilution)

in blocking buffer for 1 hour at room temperature, followed by secondary goat anti-

rabbit-AF647 conjugate (Invitrogen; 1:1000 dilution) in blocking buffer for 1 hour at

room temperature. Labeled cells were washed 4 × 5 minutes with PBS before imaging.

Confocal microscopy was performed in the same way as described above.

To quantify the extent of mitochondrial localization for each of the orphan

proteins, the mitochondrial regions were first identified by automated image

segmentation of the anti-Tom20 (AF647) channel. These regions were examined visually

to ensure good overlap with the mitochondrial marker. The mean fluorescence intensity

of the orphan protein (AF568) channel was then calculated for both regions overlapping

with mitochondria (“mito”) and regions not overlapping with mitochondria (“non-mito”).

After subtracting background, the intensity ratios of “mito” over “non-mito” were

calculated. For each orphan protein, 3-5 fields of view each containing 1-3 cells were

analyzed in this way, and the mito/non-mito ratios were averaged.

Generation of cells stably expressing PPOX-APEX (related Fig. S11B).

HEK293T cells were transfected at ~70% confluence with PPOX-APEX (in pLX304)

and two lentiviral packaging plasmids (dR8.91 and pVSVG (16)) using Lipofectamine

2000. 48 hours after transfection, the cell culture media containing lentivirus was

collected and filtered through a 0.45 µm syringe filter. To a fresh HEK293T culture at

~70% confluence with ~200,000 cells was added 500 µL of the filtered lentivirus-

containing media. Two days after infection, cells were transferred to complete FBS media

containing 2 µg/mL blasticidin for selection. Infected cells were maintained in 2 µg/mL

blasticidin-containing media for 7 days (three passages), before replating for EM imaging.

Cloning and mutagenesis

A variety of strategies were used to clone the APEX fusion constructs. In some cases, we

PCR-amplified the APEX gene, digested with restriction enzymes, and ligated

directionally into similarly-cut vectors (e.g., pcDNA3, pDisplay, pEGFP, or pTRC). In

other cases, we used PCR overlap extension to fuse or insert the APEX gene to the gene

of interest. The resulting PCR product was digested with restriction enzymes and pasted

into cut vectors. This strategy was used to clone APEX120-CPOX, APEX70-CPOX,

APEX205-PPOX and APEX14-CHCHD3. Lastly, we used Gateway technology

(Invitrogen) to clone some constructs, such as PPOX-APEX.

In some APEX fusion constructs, a short linker was added between APEX and the

gene of interest. This was achieved by designing longer primers for PCR amplification of

the APEX gene. In all cases, the CMV promoter was used to drive expression in

mammalian cells. The pDisplay vector was used for constructs targeted to the secretory

pathway (e.g., ER lumen and cell surface); pDisplay contains an Igk chain signal

sequence and the transmembrane domain of the PDGF receptor (which is deleted for ER

lumen-targeted constructs). Occasionally, the pEGFP vector was used because it

originally contained the gene of interest (as in the case of actin-APEX). In all other cases,

pcDNA3 was used as a generic vector for expression of proteins targeted to the

mammalian cell cytosol or organelles other than the secretory pathway.

Point mutants were generated using standard QuikChange mutagenesis

(Stratagene). Primers 30-40 base pairs in length were designed to have melting

temperatures between 78-80 °C, with one or two mismatched base pairs.

We avoided using the HA epitope tag in this work because it is tyrosine-rich, and

hence likely to be damaged by the peroxidase/biotin-phenol reaction.

Genetic constructs

The two tables below summarize the plasmids used for peroxidase expression in

mammalian and bacterial cells. Human CPOX, PPOX, CHCHD3, Sam50, and COASY

genes were obtained from the Broad hORFeome library. Vimentin-APEX has been

described (5).

Mammalian expression plasmids

Name Features Promoter/

Vector

Variants Details

mito-APEX NotI-mito-BamHI-V5-

APEX-Stop-XhoI

CMV/

pCDNA3

Soybean APEX has 3 mutations

relative to wtAPX: K14D, W41F, and

E112K.

wtAPX gene from soybean was a gift

from Emma Raven (University of

Leicester).

mito matrix targeting sequence:

MLATRVFSLVGKRAISTSVCVRAH

(5)

V5: GKPIPNPLLGLDST

PPOX-APEX NotI-PPOX-BamHI-V5-

APEX-Stop-XhoI

CMV/

pCDNA3

PPOX-APEX

for stable cells

attB1-PPOX-BamHI-V5-

APEX-Stop-attB2

CMV/

pLX304

APEX205-PPOX NotI-PPOX(1-205)-NheI-V5-

APEX-AflII-10aa linker-

PPOX(206-477)-Stop-XhoI

CMV/

pCDNA3

10aa linker: GGSGGSGGSR

CPOX-APEX HindIII-CPOX-NotI-APEX-

Flag-Stop-AflII

CMV/

pCDNA3

Flag: DYKDDDDK

APEX120-CPOX NotI-CPOX(1-120)-NheI-V5-

APEX-AflII-10aa linker-

CPOX(121-453)-Stop-XhoI

CMV/

pCDNA3

10aa linker: GGSGGSGGSR

APEX70-CPOX NotI-CPOX(1-70)-NheI-V5-

APEX-AflII-10aa linker-

CPOX(71-453)-Stop-XhoI

CMV/

pCDNA3

10aa linker: GGSGGSGGSR

APEX-NES NotI-Flag-APEX-NES-Stop-

XhoI

CMV/

pCDNA3

W41X NES: LQLPPLERLTLD(35)

IMS-APEX NotI-LACTB(1-68)-BamHI-

V5-APEX-Stop-XhoI

CMV/

pCDNA3

LACTB(1-68):

MYRLLSSVTARAAATAGPAWDGG

RRGAHRRPGLPVLGLGWAGGLGL

GLGLALGAKLVVGLRGAVPIQS

(36).

The LACTB gene was a gift from

Timothy A. Brown and David A.

Clayton (HHMI Janelia Farm)

APEX-NLS NotI-Flag-APEX-EcoRI-

NLS-Stop-XhoI

CMV/

pCDNA3

NLS:

PKKKRKVDPKKKRKVDPKKKRKV

(37)

APEX-TM EcoRI-ss-HA-ApaI-APEX-

SacII-myc-TM-Stop-NotI

CMV/

pDisplay

ss: Igk chain signal sequence in

pDisplay (Invitrogen).

HA: YPYDVPDYA

myc: EQKLISEEDL

TM: transmembrane domain of PDGF

receptor from pDisplay (Invitrogen)

APEX-KDEL EcoRI-ss-BglII-APEX-V5-

KDEL-Stop-NotI

CMV/

pDisplay

KDEL: ER retention motif

C1(1-29)-APEX HindIII-C1(1-29)-NotI-APEX-

V5-Stop-XhoI

CMV/

pCDNA3

C1(1-29):

MDPVVVLGLCLSCLLLLSLW

KQSYGGGKL (endoplasmic reticulum

membrane anchor (38))

Connexin43-

APEX

NotI-Cx43-12aa linker-

BamHI-V5-APEX-Stop-

XhoI

CMV/

pCDNA3

12aa linker: GSKGSGSTSGSG

Actin-APEX NheI-Flag-APEX-4aa linker-

XhoI-actin-Stop-BamHI

CMV/

pEGFP

APEX is derived from pea (5) rather

than soybean

4aa linker: SGLR

actin is human β-actin

W41FAPX-NES NotI-Flag-W41FAPX-NES-

Stop-XhoI

CMV/

pCDNA3

W41A

W41AAPX-CAAX NotI-Flag-W41AAPX-EcoRI-

CAAX-Stop-XhoI

CMV/

pCDNA3

CAAX:

RSKLNPPDESGPGCMSCKCVLS

(39)

HRP-TM EcoRI-ss-HA-ApaI-HRP-

SacII-myc-TM-Stop-NotI

CMV/

pDisplay

The HRP gene was a gift from Frances

Arnold (Caltech)

HRP-KDEL EcoRI-ss-BglII-HRP-V5-

KDEL-Stop-NotI

CMV/

pDisplay

PNPT1-APEX NotI-PNPT1-BamHI-V5-

APEX-XhoI

CMV/

pCDNA3

The human PNPT1 gene was a gift

from Tsutomu Suzuki (University of

Tokyo)

APEX14-

CHCHD3

NotI-CHCHD3(1-14)-NheI-

V5-APEX-AflII-10aa linker-

CHCHD3(15-227)-Stop-XhoI

CMV/

pCDNA3

10aa linker: GGSGGSGGSR

COASY-APEX NotI-COASY-BamHI-V5-

APEX-Stop-XhoI

CMV/

pCDNA3

mito-pBirA NotI-mito-BamHI-V5-

pBirA-Stop-XhoI

CMV/

pCDNA3

pBirA: R118G mutant of BirA

(promiscuous biotin ligase)

pBirA-KDEL EcoRI-ss-BglII-pBirA-

KDEL-Stop-NotI

CMV/

pDisplay

pBirA-NLS NotI-Flag-pBirA-EcoRI-

NLS-Stop-XhoI

CMV/

pCDNA3

Bacterial expression plasmids

Name Features Promoter/Vector

His6-APEX NcoI-His6-APEX-Stop-XbaI lacO/pTRC99

His6-wtAPX NcoI-His6-wtAPX-Stop-XbaI lacO/pTRC99

Discussion of labeling radius

A future challenge lies in carefully defining the radius of peroxidase-catalyzed labeling

when it is not membrane-enclosed. An imaging readout following live cell labeling can

be misleading because biotinylated proteins can themselves diffuse during the 1 minute

labeling window. Previous EM studies on fixed cells suggest that the labeling radius of

the biotin phenoxyl radical can be as small as 20 nm (8, 9), and in the live cell context the

radius may be even smaller due to the presence of endogenous radical quenchers such as

glutathione. It should also be possible to modulate labeling radius using electron

withdrawing substituents or by co-application of radical scavengers to cells.

Supplementary Figures

Fig. S1. Comparison of promiscuous biotinylation by APEX and pBirA. (A)

Streptavidin blot analysis of whole cell lysate after live biotinylation by APEX or

promiscuous biotin ligase (pBirA = R118G BirA (3)) in three different HEK293T cell

compartments. The same samples were visualized by Ponceau stain on the left. APEX

labeling was performed for 1 minute (1m) with 500 µM biotin-phenol and 1 mM H2O2.

pBirA labeling was performed for 1 minute or 24 hours (24h) with 50 µM biotin, as

previously described (1). NLS = nuclear localization signal. KDEL = endoplasmic

reticulum retention sequence. mito = targeting sequence for the mitochondrial matrix.

pBirA-catalyzed biotinylation is undetectable after 1 minute in the nucleus and

mitochondrial matrix, and is very low in the lumen of the endoplasmic reticulum even

after 24 hours, in contrast to APEX-catalyzed biotinylation. (B) Imaging analysis of

biotinylation in the mitochondrial matrix. The top row shows 1 minute biotinylation by

mito-APEX, performed as in (A). In the bottom two rows, HEK293T cells were

transfected with mitochondrial matrix-localized pBirA. 24 hours after transfection, 50

μM biotin was added to the culture medium to permit promiscuous biotinylation. After 10

minutes or 1 hour, cells were fixed and stained with neutravidin to detect biotinylated

proteins and anti-V5 antibody (AF488 visualization) to detect pBirA expression.

Mitotracker Deep Red staining was also used to visualize mitochondria. The neutravidin

channel is shown at higher contrast in the rightmost column. The staining seen in rows 2

and 3 reflects endogenously biotinylated proteins (located in mitochondria (40)) because

the staining is equivalent between transfected (V5-positive) and untransfected cells. Thus

biotinylation of the mitochondrial matrix proteome by pBirA is insignificant after 1 hour,

whereas APEX-catalyzed biotinylation in the same compartment is strong after 1 minute.

Fig. S2. Reactivity of eight different biotin conjugates towards APX and HRP. Top:

Structures of biotin conjugates tested. These differ in electronic properties, linker length,

sterics, and charge. Aryl azide 8 was used in a previous study with HRP and proposed to

form a nitrene (4). Bottom: Imaging results with HEK293T cells expressing cytosolic

Flag-W41F

APX-NES (top row) or cell surface HRP-myc-TM (bottom row). NES = nuclear

export signal. TM = transmembrane domain of the PDGF receptor. For APX, labeling

was performed by incubating cells with 500 µM of the indicated biotin substrate for 30

minutes, then adding 1 mM H2O2 for 1 minute to initiate biotinylation. Cells were then

fixed and stained. For cell surface labeling with HRP, 100 μM of the indicated substrate

was added for 10 minutes, then 1 mM H2O2 was added for 1 minute to initiate labeling.

Cells were then fixed and stained. Anti-Flag/myc, streptavidin, and DIC images are

overlaid. Labeling was not detected with HRP-TM and probe 8, in contrast to a previous

report (4). The simplest conjugate, biotin-phenol 1, gave the strongest labeling of

intracellular proteins, so we used this probe for all peroxidase-mediated biotinylation

experiments in this work. Scale bars, 10 μm.

Fig. S3. APEX-catalyzed labeling requires addition of exogenous H2O2. HEK293T

cells expressing APEX in the mitochondrial matrix (lanes 4-6) or the intermembrane

space (IMS; lanes 7-9) were treated with biotin-phenol for 30 minutes and 1 mM H2O2

for 1 minute, as in the proteomic experiments, or with biotin-phenol alone for 30 minutes,

or with neither reagent. Cells were then quenched and lysed, and whole cell lysate was

analyzed by streptavidin-HRP blotting (two different exposures shown in middle and

right blots). Labeling of many endogenous proteins was observed in lanes 6 and 9. No

biotinylation above background was detected when biotin-phenol was added alone

without exogenous H2O2 (lanes 5 and 8), indicating that endogenous H2O2 is not

sufficient to start the reaction in either the matrix or the IMS. Untransfected HEK293T

cells are included for comparison in lanes 1-3. Arrowheads point to endogenously

biotinylated proteins (40).

Fig. S4. Imaging analysis of APEX- and HRP-catalyzed biotinylation in multiple

cellular compartments. (A) HEK293T cells were transfected with APEX or HRP

constructs targeted to various compartments. Labeling was performed with biotin-phenol

and H2O2 for 1 minute. Cells were then fixed and stained with streptavidin-AF568 to

visualize biotinylated proteins, and antibody to visualize Flag, V5, or myc tags on the

APEX/HRP enzymes. HRP-NES and mito-HRP constructs did not give detectable

streptavidin staining, as expected (data not shown). NLS = nuclear localization signal;

NES = nuclear export signal; mito = mitochondrial matrix signal; KDEL = endoplasmic

reticulum retention motif; TM = transmembrane helix of PDGF receptor. Specific

targeting sequences are given in the Materials & Methods. (B) Additional imaging

analysis of endogenous proteins biotinylated by mito-APEX and HRP-KDEL, with

respect to organelle markers. mito-GFP is GFP appended to a mitochondrial matrix signal.

ER-GFP is GFP targeted to the endoplasmic reticulum membrane via fusion to the first

29 amino acids of cytochrome P450 (38) (this plasmid was a gift from Erik Snapp, Albert

Einstein College of Medicine). COS7 cells were labeled and stained as in (A). Merged

images are shown in the right two columns. Tight overlap is observed between proteins

biotinylated by APEX/HRP and the respective organelle markers. All scale bars, 10 µm.

Fig. S5. Streptavidin blot analysis of APEX- and HRP-catalyzed biotinylation in

many cellular compartments. (A) The same constructs as in Fig. S4A were introduced

into HEK293T cells, which were labeled with biotin-phenol and H2O2 as in Fig. S4, then

lysed. Whole cell lysates were analyzed by 10% SDS-PAGE, with Ponceau S staining

(left) and streptavidin-HRP visualization of biotinylated proteins (right). Arrowheads

point to endogenously biotinylated proteins (40). (B) Same analysis as in (A), for fusion

constructs targeted to different regions of the cytosol. NES = nuclear export signal; IMS

= mitochondrial intermembrane space signal; C1(1-29) = endoplasmic reticulum membrane

anchor (APEX faces the cytosol) (38); CAAX = plasma membrane anchor (APEX faces

the cytosol); Cx43 = connexin 43. Note that the IMS freely exchanges components <5 kD

(including the biotin-phenoxyl radical) with the cytosol via porins in the outer

mitochondrial membrane (41). Pea APEX (5) instead of soybean APEX was used for the

vimentin and actin fusions. Soybean W41A

APX was used for the CAAX fusion. For each

of the organelle-targeted constructs in (A), and the cytosol-facing constructs in (B), a

unique “fingerprint” of biotinylated endogenous proteins is observed, illustrating the

spatial specificity of the labeling.

Fig. S6. Peroxidase-generated phenoxyl radicals do not cross the plasma membrane.

(A) Assay scheme. HEK293T cells were transfected with W41A

APX targeted to the

cytosol (NES) or inner leaflet of the plasma membrane (CAAX; green pacman). Labeling

was performed with alkyne-phenol and H2O2 for 1 minute. Thereafter click chemistry

was performed at the cell surface with membrane-impermeant AF647-picolyl azide (34).

Only if the phenoxyl radical crosses the plasma membrane will alkyne be present at the

cell surface and detectable by the AF647 reagent. Dimeric W41A

APX was used in this

experiment because it has higher activity in the mammalian cytosol than monomeric

APEX. (B) Images with W41A

APX-CAAX (left) and W41A

APX-NES (right). Nuclear-

localized YFP was a transfection marker. As a control (bottom rows labeled “total”), cells

were fixed and permeabilized prior to click chemistry to detect intracellular alkyne-

phenol labeling. Insets show the same fields of view with 50-fold greater contrast.

Quantitation shows that the signal at the cell surface (top rows) represents <2% of the

total signal (bottom rows). The same results were obtained when we used biotin-phenol

instead of alkyne-phenol and detected labeling at the cell surface with membrane-

impermeant streptavidin-AF568 (data not shown). The alkyne-phenol experiment is

shown here because it is more rigorous; detection of cell surface labeling by biotin-

phenol, especially near the lipid bilayer, could be impaired by the large size of

streptavidin.

Fig. S7. Peroxidase engineering and characterization of monomericity and amino

acid specificity in vitro. (A) The APEX construct previously introduced as a genetic tag

for electron microscopy was derived from pea ascorbate peroxidase (APX) (5). This

construct had three mutations relative to wild-type pea APX: two to make it monomeric

(K14D, E112K), and one to boost activity towards aromatic substrates (W41F). In this

work, we discovered that K14D and E112K had a more strongly monomerizing effect on

APX from soybean than from pea (both wild-type enzymes are constitutive dimers (42,

43)). We therefore switched to the soybean enzyme. In addition, we performed

comprehensive mutagenesis at the W41 active site residue to maximize intracellular

activity towards biotin-phenol. Confocal images are shown for HEK293T cells

expressing all possible W41x mutants (fused to nuclear export sequences, and on a

monomeric soybean APX background). After labeling for 1 minute with biotin-phenol

and H2O2, cells were fixed and stained with streptavidin-AF568 to detect biotinylated

proteins (red) and anti-Flag to detect peroxidase expression (green). The same experiment

with dimeric W41F

APX (soybean) is shown for comparison. Anti-Flag, streptavidin, and

DIC images are overlaid. From this experiment, we concluded that the W41F mutation

was best, and combined it with the monomerizing mutations to give soybean APEX

(K14D, W41F, E112K). Note that soybean-derived APEX was used for all EM and

proteomic labeling experiments in this work, except where indicated otherwise. Scale

bars, 10 μm. (B) Gel filtration chromatography was used to analyze the monomericity of

soybean-derived APEX, pea-derived APEX from our previous study (5), and wild-type

soybean APX (calculated molecular weight 29 kD). As expected (43), wild-type APX

runs as a dimer at all protein concentrations tested (250 nM–68 μM; apparent MW ~58

kD). Soybean APEX runs as a monomer (apparent MW ~30 kD), even at 270 μM, while

pea APEX runs as a monomer at lower concentrations but shows signs of dimerization at

higher concentrations, as previously observed (5). (C) HRP, biotin-phenol (BP), and

H2O2 were combined in 20 separate in vitro reactions with each of the amino acids (at

500 µM). Reactions were quenched after 10 minutes at room temperature and analyzed

by HPLC and MS. HRP was used because it has higher in vitro activity than APEX (5).

In all reactions, BP dimer was detected. In the reaction with tyrosine, a BP-tyrosine

adduct and tyrosine dimer with the masses and predicted structures shown were also

detected. Negative controls with HRP or H2O2 omitted eliminated the formation of these

products. In the reaction with cysteine, the amino acid was consumed but we did not

detect the expected cysteine-BP adduct or cysteine dimer. No reaction was detected for

the 18 other samples (apart from BP dimerization). We conclude that the reaction of BP

with tyrosine is the easiest to detect (see also Fig. S9), but reactions at other electron-rich

sidechains are also possible and have literature precedent (10-12). We note that based on

analysis of available protein structures, >90% of proteins are expected to have one or

more surface-exposed tyrosines1

. Thus, even if phenoxyl radicals have a strong

preference for labeling at tyrosine, this specificity should not limit by much the potential

coverage of this technique.

1. This value was calculated by determining the number of Tyr residues for each protein in the

human proteome, and applying the statistic that 80% of Tyr are surface-exposed (14). Hence, for

the 677 (3.3%) proteins with 1 Tyr, 80% of them will have a surface Tyr. For the 813 (4%)

proteins with 2 Tyr, 96% of them will have one or more surface Tyr, and so on.

Fig. S8. Determination of the mitochondrial matrix proteome. (A) Experimental setup.

We established “heavy” and “light” isotopically labeled HEK293T cell cultures using the

SILAC approach (18). Two replicate heavy SILAC cultures were labeled by pre-loading

biotin-phenol for 30 minutes, then adding 1 mM H2O2 for 1 minute to initiate

biotinylation. Two light SILAC cultures were processed in parallel as negative controls,

one with APEX omitted, and one with biotin-phenol and H2O2 omitted. After lysis, heavy

and light lysates were combined, enriched with streptavidin beads, eluted, separated by

gel, digested to peptides with trypsin, and analyzed by tandem MS. (B) Table showing

numbers of detected peptides, detected proteins, and enriched proteins for each replicate,

as well as the overlap between the two replicates. We required three or more unique

quantified peptides to consider a protein “detected”. (C) Distribution of SILAC ratios

(log2(H/L)) from the Replicate 1 experiment. The top panel shows “true positive”

analysis: the number of proteins within each SILAC ratio range (bin size 0.1) with prior

mitochondrial annotation. The bottom panel shows “false positive” analysis: the number

of proteins within each SILAC ratio range appearing in a hand-curated list of 2410 non-

mitochondrial proteins (“false positive list”) (16). (D) False positive rate (FPR) as a

function of SILAC ratio. At a given SILAC ratio, the FPR measures the probability of

finding a false positive versus a true mitochondrial protein. We selected an FPR of 0.1 as

our cut-off point, i.e. 10 times more likely for a protein to be mitochondrial than a false

positive. Replicate 1 had 527 proteins above this cut-off, and Replicate 2 had 579

proteins. The overlap between these groups was 495 proteins, which we defined as our

matrix proteome (Table S1). (E) Correlation between SILAC ratios for Replicates 1 and 2.

R values are shown for all detected proteins, and for proteins in our matrix proteome

(proteins above and to the right of the dotted lines). (F) Correlation between SILAC ratio

(from Replicate 1) and protein abundance in human U2OS cells (44), for proteins in our

matrix proteome. The bottom strip, in red, shows protein copy number for matrix proteins

from Fig. 2B that were not detected in our matrix proteome.

Fig. S9. Mapping biotinylated peptides to transmembrane and soluble matrix

proteins. (A) Occurrence of surface-exposed tyrosines in soluble proteins with known

structure. (i) is from reference (45). (iii) is based on analysis of the 60 biotinylated

peptides we detect that map to soluble proteins with known structure (39 proteins). (ii) is

based on all the tyrosines (biotinylated or not, 401 in total) within these 39 proteins. The

criterion for “surface-exposed” is having >10 Å2 solvent-accessible area, as calculated by

PyMOL (45). This analysis shows that peroxidase-catalyzed biotinylation favors surface

exposed tyrosine sites on soluble proteins. Protein structures were obtained by searching

UniProt IDs in the Swiss Model Repository. Structures with >85% sequence homology to

our detected proteins were used for this analysis. (B) Protein structures showing the

positions of five of the biotinylated tyrosines (colored red, bp = biotin-phenol

modification) analyzed in (A). Non-labeled tyrosines within each protein are colored blue.

Sequences of detected biotinylated peptides are shown beneath each protein structure. (C)

Analysis of biotinylated peptides that map to transmembrane proteins with known or

predicted topologies (according to the UniProt database). Nearly all of the biotinylation

sites (14 out of 15) map to the matrix-exposed portion of these proteins. (D) Specific

examples from (C). Biotinylation sites are depicted by the red circle. Sequences of

detected biotinylated peptides are given underneath. (E) MS/MS spectrum recorded on

the SILAC heavy-labeled [M + 2H]2+

ion (m/z 1086.54) corresponding to the pyruvate

dehydrogenase E1β peptide VFLLGEEVAQbp

YDGAYK shown in (B). The measured

precursor mass is within 1 p.p.m. of the calculated mass. Observed singly charged

fragment ions are labeled in the spectrum and indicate that Y63 is modified with biotin-

phenol. Details for all 167 biotinylated peptides are provided in Table S4.

Fig. S10. Fluorescence imaging shows mitochondrial localization of six

mitochondrial orphans. Six proteins selected at random from the 31 mitochondrial

orphans (Table S2) identified in our matrix proteome were V5-tagged and expressed in

COS7 cells. Protein localization was detected by anti-V5 staining (AF568). Mitochondria

were visualized by anti-Tom20 staining (AF647). The ratio of V5 signal intensity

overlapping with mitochondria versus not overlapping with mitochondria was quantified

(bottom). For reference, the mito/non-mito ratio was 5.4±2.0 for mito-APEX (a positive

control) and 1.4±0.16 for a cytosolic fluorescent protein (mCherry) fusion (a negative

control). Four of these six orphan proteins are related to tRNA editing: TRMT61B and

TRMT11 are tRNA methyltransferases, and TRUB2 and RPUSD3 are pseudouridylate

synthases. TRUB2 has a yeast homolog, PUS4, with mitochondrial annotation (46).

TRUB2 and RPUSD3 both have predicted mitochondrial targeting sequences. PREPL is

prolyl endopeptidase-like protein. Though annotated as a cytosolic protein, a previous

study found 35% of prolyl endopeptidase activity to be associated with crude

mitochondria (47). PNPO is pyridoxamine-phosphate oxidase, which catalyzes the last

and rate-determining step of vitamin B6 synthesis, thought to occur in the cytosol. PNPO

has a yeast homolog, PDX3, which was identified in an IMS proteomic study (48), but

not validated biochemically. Scale bars, 15 μm.

Fig. S11. Electron microscopy of APEX fusions to PNPT1, CHCHD3, COASY, and

PPOX. (A) PNPT1 and CHCHD3 were previously localized to the mitochondrial

intermembrane space (IMS) (49, 50), and COASY was previously localized to the outer

mitochondrial membrane (OMM) (51), yet we detected all three proteins in our

mitochondrial matrix proteome. APEX fusions to PNPT1 (C-terminal fusion) and

CHCHD3 (insertion between amino acids 14 and 15) gave mitochondrial matrix staining

by EM in HEK293T and COS7 cells, respectively. PNPT1 (polyribonucleotide

nucleotidyltransferase 1) is an RNA binding and processing protein. In addition to our

EM data which localizes the C-terminus to the matrix, we detected two PNPT1-derived

biotinylated peptides (Table S4) that indicate that Tyr302 and Tyr356 are both matrix-

exposed. CHCHD3 (coiled-coil-helix-coiled-coil-helix domain-containing protein 3)

plays a role in bridging the outer and inner mitochondrial membranes (52), so perhaps it

is not surprising that a portion of it (residue 14, according to our EM data) is matrix-

exposed. A C-terminal APEX fusion to COASY (bifunctional coenzyme A synthase) in

HEK293T cells gave matrix staining in some fields of view (top two images), and IMS

staining in other fields of view (bottom). (B) Same experiment as in Fig. 3C, except that

PPOX-APEX was stably rather than transiently expressed. These stable HEK293T cells

were prepared by lentiviral infection of the PPOX-APEX gene and selection in 2 µg/mL

blasticidin. All scale bars, 200 nm.

Supplementary Spreadsheet 1. Mitochondrial matrix proteome.

Table S1: Our mitochondrial matrix proteome (495 proteins), ranked from most enriched

to least enriched, according to log2(H/L) SILAC ratio from replicate 1. 93 additional

proteins below our cut-off point are also included but shaded grey.

Table S2: Mitochondrial orphans (31 proteins), ranked from most enriched to least

enriched.

Table S3: Mitochondrial matrix orphans (240 proteins), ranked from most enriched to

least enriched. These proteins have prior mitochondrial annotation but were not

previously known to be localized to the matrix.

Table S4: Biotinylated peptides detected (167 peptides derived from 103 proteins). The

peptides are grouped by protein and ranked by protein SILAC ratio.

Table S5: Column definitions.

Supplementary Spreadsheet 2. Analysis of protein groups and membrane complexes.

Table S6: Sub-mitochondrial localization of detected proteins. Related to Fig. 2A.

Table S7: Mitochondrial matrix protein groups detected. Related to Fig. 2B (analysis of

depth of coverage). Eight proteins in these groups that were not detected by RNA-Seq in

HEK293T cells (53) were not included in the analysis.

Table S8: Inner mitochondrial membrane protein complexes detected. Related to Figs.

2C-D.

Table S9: Column definitions.

References and Notes

1. K. J. Roux, D. I. Kim, M. Raida, B. Burke, A promiscuous biotin ligase fusion protein

identifies proximal and interacting proteins in mammalian cells. J. Cell Biol. 196, 801

(2012). doi:10.1083/jcb.201112098 Medline

2. B. Morriswood et al., Novel bilobe components in Trypanosoma brucei identified using

proximity-dependent biotinylation. Eukaryot. Cell (2012). Medline

3. E. Choi-Rhee, H. Schulman, J. E. Cronan, Promiscuous protein biotinylation by Escherichia

coli biotin protein ligase. Protein Sci. 13, 3043 (2004). doi:10.1110/ps.04911804 Medline

4. N. Kotani et al., Biochemical visualization of cell surface molecular clustering in living cells.

Proc. Natl. Acad. Sci. U.S.A. 105, 7405 (2008). doi:10.1073/pnas.0710346105 Medline

5. J. D. Martell et al., Engineered ascorbate peroxidase as a genetically encoded reporter for

electron microscopy. Nat. Biotechnol. 30, 1143 (2012). doi:10.1038/nbt.2375 Medline

6. J. F. Wishart, B. S. Madhava Rao, in Recent Trends in Radiation Chemistry (World Scientific,

Singapore, 2010), pp. 577–596.

7. A. Mortensen, L. H. Skibsted, Importance of carotenoid structure in radical-scavenging

reactions. J. Agric. Food Chem. 45, 2970 (1997). doi:10.1021/jf970010s

8. M. Bendayan, Worth its weight in gold. Science 291, 1363 (2001).

doi:10.1126/science.291.5507.1363

9. G. Mayer, M. Bendayan, Biotinyl-tyramide: A novel approach for electron microscopic

immunocytochemistry. J. Histochem. Cytochem. 45, 1449 (1997).

doi:10.1177/002215549704501101 Medline

10. K. Minamihata, M. Goto, N. Kamiya, Protein heteroconjugation by the peroxidase-catalyzed

tyrosine coupling reaction. Bioconjug. Chem. 22, 2332 (2011). doi:10.1021/bc200420v

Medline

11. M. S. Rogers et al., Cross-link formation of the cysteine 228-tyrosine 272 catalytic cofactor

of galactose oxidase does not require dioxygen. Biochemistry 47, 10428 (2008).

doi:10.1021/bi8010835 Medline

12. B. Bhaskar et al., A novel heme and peroxide-dependent tryptophan-tyrosine cross-link in a

mutant of cytochrome c peroxidase. J. Mol. Biol. 328, 157 (2003). doi:10.1016/S0022-

2836(03)00179-7 Medline

13. F. Amini, T. Kodadek, K. C. Brown, Protein affinity labeling mediated by genetically

encoded peptide tags. Angew. Chem. Int. Ed. Engl. 41, 356 (2002). doi:10.1002/1521-

3773(20020118)41:2<356::AID-ANIE356>3.0.CO;2-M Medline

14. W. Schönhuber, B. Fuchs, S. Juretschko, R. Amann, Improved sensitivity of whole-cell

hybridization by the combination of horseradish peroxidase-labeled oligonucleotides and

tyramide signal amplification. Appl. Environ. Microbiol. 63, 3268 (1997). Medline

15. B. Huang, M. Bates, X. Zhuang, Super-resolution fluorescence microscopy. Annu. Rev.

Biochem. 78, 993 (2009). doi:10.1146/annurev.biochem.77.061906.092014 Medline

16. D. J. Pagliarini et al., A mitochondrial protein compendium elucidates complex I disease

biology. Cell 134, 112 (2008). doi:10.1016/j.cell.2008.06.016 Medline

17. F. Forner, L. J. Foster, S. Campanaro, G. Valle, M. Mann, Quantitative proteomic

comparison of rat mitochondria from muscle, heart, and liver. Mol. Cell. Proteomics 5,

608 (2006). doi:10.1074/mcp.M500298-MCP200 Medline

18. S. E. Ong et al., Stable isotope labeling by amino acids in cell culture, SILAC, as a simple

and accurate approach to expression proteomics. Mol. Cell. Proteomics 1, 376 (2002).

doi:10.1074/mcp.M200025-MCP200 Medline

19. G. C. Ferreira, T. L. Andrew, S. W. Karr, H. A. Dailey, Organization of the terminal two

enzymes of the heme biosynthetic pathway. Orientation of protoporphyrinogen oxidase

and evidence for a membrane complex. J. Biol. Chem. 263, 3835 (1988). Medline

20. J.-C. Deybach, V. da Silva, B. Grandchamp, Y. Nordmann, The mitochondrial location of

protoporphyrinogen oxidase. Eur. J. Biochem. 149, 431 (1985). doi:10.1111/j.1432-

1033.1985.tb08943.x Medline

21. G. H. Elder, J. O. Evans, Evidence that the coproporphyrinogen oxidase activity of rat liver is

situated in the intermembrane space of mitochondria. Biochem. J. 172, 345 (1978).

Medline

22. B. Grandchamp, N. Phung, Y. Nordmann, The mitochondrial localization of

coproporphyrinogen III oxidase. Biochem. J. 176, 97 (1978). Medline

23. M. Koch et al., Crystal structure of protoporphyrinogen IX oxidase: A key enzyme in haem

and chlorophyll biosynthesis. EMBO J. 23, 1720 (2004). doi:10.1038/sj.emboj.7600189

Medline

24. I. Hamza, H. A. Dailey, One ring to rule them all: Trafficking of heme and heme synthesis

intermediates in the metazoans. Biochim. Biophys. Acta Mol. Cell Res. 1823, 1617

(2012). doi:10.1016/j.bbamcr.2012.04.009

25. R. Nilsson et al., Discovery of genes essential for heme biosynthesis through large-scale gene

expression analysis. Cell Metab. 10, 119 (2009). doi:10.1016/j.cmet.2009.06.012

Medline

26. X. Qin et al., Structural insight into human variegate porphyria disease. FASEB J. 25, 653

(2011). doi:10.1096/fj.10-170811 Medline

27. M. Howarth, A. Y. Ting, Imaging proteins in live mammalian cells with biotin ligase and

monovalent streptavidin. Nat. Protoc. 3, 534 (2008). doi:10.1038/nprot.2008.20 Medline

28. D. S. Lee et al., Structural basis of hereditary coproporphyria. Proc. Natl. Acad. Sci. U.S.A.

102, 14232 (2005). doi:10.1073/pnas.0506557102 Medline

29. A. E. Medlock, T. A. Dailey, T. A. Ross, H. A. Dailey, W. N. Lanzilotta, A pi-helix switch

selective for porphyrin deprotonation and product release in human ferrochelatase. J.

Mol. Biol. 373, 1006 (2007). doi:10.1016/j.jmb.2007.08.040 Medline

30. D. A. Dalton, L. Diaz del Castillo, M. L. Kahn, S. L. Joyner, J. M. Chatfield, Heterologous

expression and characterization of soybean cytosolic ascorbate peroxidase. Arch.

Biochem. Biophys. 328, 1 (1996). doi:10.1006/abbi.1996.0135 Medline

31. J. Rappsilber, M. Mann, Y. Ishihama, Protocol for micro-purification, enrichment, pre-

fractionation and storage of peptides for proteomics using StageTips. Nat. Protoc. 2,

1896 (2007). doi:10.1038/nprot.2007.261 Medline

32. J. Cox, M. Mann, MaxQuant enables high peptide identification rates, individualized p.p.b.-

range mass accuracies and proteome-wide protein quantification. Nat. Biotechnol. 26,

1367 (2008). doi:10.1038/nbt.1511 Medline

33. J. Cox et al., Andromeda: A peptide search engine integrated into the MaxQuant

environment. J. Proteome Res. 10, 1794 (2011). doi:10.1021/pr101065j Medline

34. C. Uttamapinant et al., Fast, cell-compatible click chemistry with copper-chelating azides for

biomolecular labeling. Angew. Chem. Int. Ed. Engl. 51, 5852 (2012).

doi:10.1002/anie.201108181 Medline

35. W. Wen, J. L. Meinkoth, R. Y. Tsein, S. S. Taylor, Identification of a signal for rapid export

of proteins from the nucleus. Cell 82, 463 (1995). doi:10.1016/0092-8674(95)90435-2

36. T. A. Brown et al., Superresolution fluorescence imaging of mitochondrial nucleoids reveals

their spatial range, limits, and membrane interaction. Mol. Cell. Biol. 31, 4994 (2011).

doi:10.1128/MCB.05694-11 Medline

37. D. Kalderon, B. L. Roberts, W. D. Richardson, A. E. Smith, A short amino-acid sequence

able to specify nuclear location. Cell 39, 499 (1984). doi:10.1016/0092-8674(84)90457-4

38. E. L. Snapp et al., Formation of stacked ER cisternae by low affinity protein interactions. J.

Cell Biol. 163, 257 (2003). doi:10.1083/jcb.200306020 Medline

39. A. Keppler, H. Pick, C. Arrivoli, H. Vogel, K. Johnsson, Labeling of fusion proteins with

synthetic fluorophores in live cells. Proc. Natl. Acad. Sci. U.S.A. 101, 9955 (2004).

doi:10.1073/pnas.0401923101

40. A. Chapman-Smith, J. E. Cronan Jr., In vivo enzymatic protein biotinylation. Biomol. Eng.

16, 119 (1999). doi:10.1016/S1050-3862(99)00046-7 Medline

41. G. Kroemer, L. Galluzzi, C. Brenner, Mitochondrial membrane permeabilization in cell

death. Physiol. Rev. 87, 99 (2007). doi:10.1152/physrev.00013.2006 Medline

42. W. R. Patterson, T. L. Poulos, Crystal structure of recombinant pea cytosolic ascorbate

peroxidase. Biochemistry 34, 4331 (1995). doi:10.1021/bi00013a023 Medline

43. K. H. Sharp, M. Mewies, P. C. Moody, E. L. Raven, Crystal structure of the ascorbate

peroxidase-ascorbate complex. Nat. Struct. Biol. 10, 303 (2003). doi:10.1038/nsb913

Medline

44. M. Beck et al., The quantitative proteome of a human cell line. Mol. Syst. Biol. 7, 549 (2011).

doi:10.1038/msb.2011.82 Medline

45. D. Bordo, P. Argos, Suggestions for “safe” residue substitutions in site-directed mutagenesis.

J. Mol. Biol. 217, 721 (1991). doi:10.1016/0022-2836(91)90528-E Medline

46. H. F. Becker, Y. Motorin, R. J. Planta, H. Grosjean, The yeast gene YNL292w encodes a

pseudouridine synthase (Pus4) catalyzing the formation of Ψ55 in both mitochondrial and

cytoplasmic tRNAs. Nucleic Acids Res. 25, 4493 (1997). doi:10.1093/nar/25.22.4493

Medline

47. K. Dresdner, L. A. Barker, M. Orlowski, S. Wilk, Subcellular distribution of prolyl

endopeptidase and cation-sensitive neutral endopeptidase in rabbit brain. J. Neurochem.

38, 1151 (1982). doi:10.1111/j.1471-4159.1982.tb05362.x Medline

48. F. N. Vögtle et al., Intermembrane space proteome of yeast mitochondria. Mol. Cell.

Proteomics 11, 1840 (2012). doi:10.1074/mcp.M112.021105 Medline

49. M. Darshi et al., ChChd3, an inner mitochondrial membrane protein, is essential for

maintaining crista integrity and mitochondrial function. J. Biol. Chem. 286, 2918 (2011).

doi:10.1074/jbc.M110.171975 Medline

50. H. W. Chen et al., Mammalian polynucleotide phosphorylase is an intermembrane space

RNase that maintains mitochondrial homeostasis. Mol. Cell. Biol. 26, 8475 (2006).

doi:10.1128/MCB.01002-06 Medline

51. A. Zhyvoloup et al., Subcellular localization and regulation of coenzyme A synthase. J. Biol.

Chem. 278, 50316 (2003). doi:10.1074/jbc.M307763200 Medline

52. M. Darshi, K. N. Trinh, A. N. Murphy, S. S. Taylor, Targeting and import mechanism of

coiled-coil helix coiled-coil helix domain-containing protein 3 (ChChd3) into the

mitochondrial intermembrane space. J. Biol. Chem. 287, 39480 (2012).

doi:10.1074/jbc.M112.387696 Medline

53. M. Sultan et al., A global view of gene activity and alternative splicing by deep sequencing

of the human transcriptome. Science 321, 956 (2008). doi:10.1126/science.1160342