Noninvasive optical imaging of cysteine protease activityusing fluorescently quenched activity-based probesGalia Blum1, Georges von Degenfeld2, Milton J Merchant2, Helen M Blau2 & Matthew Bogyo1,3

We have generated a series of quenched near-infrared fluorescent activity-based probes (qNIRF-ABPs) that covalently target thepapain-family cysteine proteases shown previously to be important in multiple stages of tumorigenesis. These ‘smart’ probes emita fluorescent signal only after covalently modifying a specific protease target. After intravenous injection of NIRF-ABPs into micebearing grafted tumors, noninvasive, whole-body imaging allowed direct monitoring of cathepsin activity. Importantly, thepermanent nature of the probes also allowed secondary, ex vivo biochemical profiling to identify specific proteases and tocorrelate their activity with whole-body images. Finally, we demonstrate that these probes can be used to monitor small-moleculeinhibition of protease targets both biochemically and by direct imaging methods. Thus, NIRF-ABPs are (i) potentially valuable newimaging agents for disease diagnosis and (ii) powerful tools for preclinical and clinical testing of small-molecule therapeutic agentsin vivo.

Most enzymatic proteins involved in the control of basic biologicalprocesses are regulated at a post-translational level. Thus, simplemeasures of gene expression and even protein expression only providelimited information to aid in our understanding of the function of agiven enzyme target. To address this issue, the field of activity-basedproteomics has been developed to make use of small-moleculeprobes that bind to specific target enzymes using an activity-dependent chemical reaction (for reviews see refs. 1,2). Activity-based probes (ABPs) provide an indirect readout of enzymeactivity that can be monitored in vitro and ex vivo using biochemical,mass spectrometric and fluorescence microscopy methods3. UsingABPs it is possible to determine when, where and which enzymesget activated during a given cellular stimulus or even duringdisease pathogenesis.

Over the past ten years, ABPs have been developed to target anumber of diverse enzyme classes1,2,4,5. Proteases, in particular, havebeen the focus of extensive studies using specific ABPs. In general anABP is comprised of a reactive ‘warhead’ that covalently binds to atarget using an enzyme-catalyzed chemical reaction. This reactivegroup is linked to an element that confers specificity (often a peptidein the case of a protease-directed ABP) and directs binding to thetarget. The final key element is a reporter tag that allows probe-labeledproteins to be directly visualized or purified from a complex mixture.A number of fluorescently labeled ABPs that target proteases havebeen reported6–9. In particular, our group has developed fluorescentprobes that target the papain family of lysosomal cysteine proteases—often referred to as the cysteine cathepsins.

Many diseases, including cancer, arteriosclerosis, inflammation andAlzheimer’s disease, are associated with altered proteolytic activity ofcysteine cathepsins10,11. Recent expression studies strongly suggest that

cathepsins are involved in initiating and promoting tumor formation,growth, invasiveness and metastasis12,13. Furthermore, increasedexpression and activity of cathepsin B and cathepsin L in a varietyof malignancies, including mammary adenocarcinomas, are strongnegative prognostic indicators for individuals with cancer14–17. Theseclinical data have also recently been corroborated by studies in animalmodels of human cancer that suggest that cysteine cathepsins haveimportant roles in the regulation of angiogenesis and invasion duringcancer progression18–21. Thus, cathepsin proteases are gaining recog-nition as promising targets for diagnosis and therapy.

Recent advances in noninvasive optical molecular imaging technol-ogies have now made it possible to monitor disease progression andevaluate therapeutic treatments22–24. In many cases, biomarkers areproteins or nucleic acids that are elevated or mutated in diseased cellsor tissues. A number of elegant new methods have recently beendeveloped that rely on imaging probes that report on the enzymeactivity of specific targets22,25. In addition, several probes have beendesigned to act as substrates of proteases that, when processed, producea signal or change localization26–29. Like ABPs, these ‘smart’ probesprovide a readout of enzyme activity rather than simple proteinabundance. However, probes that are designed based on substratesare limited by their overall lack of selectivity for specific proteases,and the large size of some of the probes makes them difficult touse for intracellular targets. Furthermore, probes that act as sub-strates are not linked to their protease target, hence there is no wayto directly correlate the signals observed in vivo with specifictargets using biochemical methods, which makes analysis of imagingdata difficult.

We describe here the synthesis and application of NIRF-ABPs thattarget the cysteine cathepsins. We have engineered probes that are

Received 30 May; accepted 24 July; published online 9 September 2007; doi:10.1038/nchembio.2007.26

1Department of Pathology, 2Baxter Laboratory in Genetic Pharmacology and 3Department of Microbiology and Immunology, Stanford University School of Medicine,300 Pasteur Dr., Stanford, California 94305, USA. Correspondence should be addressed to M.B. (mbogyo@stanford.edu).

66 8 VOLUME 3 NUMBER 10 OCTOBER 2007 NATURE CHEMICAL BIOLOGY

ART ICL ES

stable in vivo and show highly specific labeling of papain-familyprotease targets in cells and whole animals. When these probes areinjected intravenously into living mice, they produce spatially resol-vable fluorescence in tumors that correlates with levels of activecysteine cathepsins in those tissues. In addition, NIRF-ABPs can beused to monitor inhibition of cysteine cathepsins by small-moleculetherapeutics. Importantly, this method is amenable to fluorescence-based biochemical and microscopic analyses of the explanted tumors,which makes it possible to correlate the measurements acquired in vivowith the actual cathepsin activity measured ex vivo. Thus, NIRF-ABPsrepresent promising new tools for experimental investigationof tumor pathophysiology, for identification of specific therapeutictargets and biomarkers and for monitoring efficacy of small-molecule inhibitors.

RESULTSDevelopment and in vitro evaluation of qNIRF-ABPsWe set out to design a new class of probes that could be used in liveanimals using noninvasive imaging techniques. As a starting point weused the peptide acyloxymethylketones (AOMKs) GB111 (1) andGB117 (2) that were recently described by our group as tools forcell-based imaging of cysteine cathepsin activity30. These probes werean ideal starting point because they had previously been shown to be(i) nontoxic to cells, (ii) reasonably water soluble and (iii) highlyselective for cathepsins B and L, both of which are upregulated innumerous tumor tissues12,31. We converted the original GB111 probeinto the Cy5 (646/664 nm excitation/emission; for structure of Cy5 seeScheme 1)-labeled version to generate GB123 (3), which is bettersuited for in vivo imaging owing to lower background fluorescence in

O

OH

O

NH +NH

20% piperidine

H N

O

O

HO

OH

KFNHFmoc

O

NHBoc

Cl

O

O

O

H N

NHFmoc

NHBoc

SPPS

OH N

O

OO

O

O

H N

N H

O

NHBoc

OH N

O

OO

O

O

H N

N H

O

NHBoc

1% TFA

NHFmocNH2

N H

N H

N H

H2N

HOBt PyBOP DIEA

QSY21 SE

50% TFA Cy5 SE

O

O

O

H N

H N

O6

OH N

N H

O

O

O

HN O

O

N+

N

SO2N

N

SO3

NO3S

O

7

Cy5

QSY21

8 (GB137)

Scheme 1 Synthesis of the fluorescently quenched probe GB137. Fmoc, 9-fluorenylmethylcarbonyl; HOBt, 1-hydroxybenzotriazole; PyBOP, (benzotriazole-

1-yl-oxy)tris(pyrrolidino)phosphonium hexafluorophosphate; DIEA, diisopropylethylamine; Boc, t-butyloxycarbonyl; SPPS, solid-phase peptide synthesis;

TFA, trifluoroacetic acid.

ART ICL ES

NATURE CHEMICAL BIOLOGY VOLUME 3 NUMBER 10 OCTOBER 2007 6 6 9

the near-infrared part of the spectrum. We synthesized a controlprobe, GB125 (4), that carried the Cy5 label but lacked the reactiveAOMK warhead. We also converted the fluorescently quenched probeGB117 to its Cy5-labeled version, GB119 (5) (Fig. 1a). These probeswere prepared using an adaptation of an approach previouslydescribed by us30. We initially tested for sensitivity to serum exposureby preincubating probes either 1 or 4 h before addition to cells.Whereas the unquenched probe GB123 was insensitive to serum evenafter the 4-h preincubation, GB119 completely lost its ability to labeltarget protease even after 1 h of preincubation (Fig. 1b). We thereforedecided to carry out further optimization of the scaffold to generatefluorescently quenched probes that retained activity in serum. One ofthe most significant differences between GB123 and GB119 is thepresence of a glycine spacer adjacent to the reactive carbon methylenegroup of the AOMK warhead. This ester bond linkage is likely to bereadily processed by the high concentrations of esterases found in

mouse serum, thus rendering the probe inactive. To test whether thislinker was the reason for the increased serum sensitivity of the GB119probe, we synthesized two probes containing bulky linkers in place ofthe original glycine of GB119. A probe containing the non-naturalamino acid aminoisobuteric acid (GB135; 6) was synthesized using thesame method described for GB119 (ref. 30), and an analog containinga dimethyl benzoic acid–based linker (GB137; 8) was prepared using anewly developed solid-phase synthesis method (Scheme 1). Testing ofthese probes in the serum sensitivity assays confirmed that GB135 (theprobe containing the aminoisobuteric acid linker) is substantially lessserum sensitive than GB119, although it still lost labeling activity afterprolonged exposure (Fig. 1b). GB137, on the other hand, showedvirtually unchanged activity after prolonged preincubation in serum.The addition of an alkyl spacer between the AOMK reactive group andthe quencher in GB137 also restored some cathepsin B binding thatwas lost in GB119 owing to unfavorable steric interaction between the

Ex/Em 633/670

Inhibitor:

Cells:

Probe:

20

28

36

42

66

98

GB123

Ex/Em 633/670

*

GB119

Serum (h):

Probe:

Inhibitor:1–– 4 4 1–– 4 4–+ – – +

+– +– +– +– +– +–

C2C12/Hras1 U87MG MDA 435 MDA 231

–+ – – +

R2 = Cy5

R2 = Cy5

R2 = IR Dye 800CW

R2 = Cy5

R2 = Cy5

R2 = Cy5

20

28364266

14

GB135

1–– 4 4–+ – – +

GB137

1–– 4 4–+ – – +

NIH3T3 MDA 231 MFP

QSY 21

QSY 21

QSY 21

Ex/Em 633/670

Inhibitor:

Cells:

Probe: GB137

20

28

36

42

66

98+– +– +– +– +– +–

C2C12/Hras1

U87MG MDA 435 MDA 231NIH3T3 MDA 231 MFP

- Cat B- Cat L- Cat L

GB123

3 (GB123) R1 = 5 (GB119) R1 =

6 (GB135) R1 =

8 (GB137) R1 =

4 (GB125) R1 = NH2

11 (GB138) R1 =

a b

c

OH N

N H

R1

O

O

O

HN

O

O

O

H N

O

O

O

O

H N

H N

O

O

6

O

H N

O

R2

Figure 1 Structures and serum sensitivity of the NIRF-ABPs (a) Structures of the nonquenched NIRF-ABPs GB123 and GB138 and the control analogGB125, which lacks the acyloxymethylketone reactive functional group (left), along with the fluorescently quenched probes GB119, GB135 and GB137

(right). Structure of the IRDye 800 is shown in Supplementary Figure 3a. (b) Sensitivity of the probes to serum exposure. Cultures of NIH3T3 were

pretreated with either inhibitor or vehicle and then labeled by addition of 1 mM probe to growth medium for 3 h. Probes were either used directly or

pretreated with mouse serum for 1 or 4 h before addition to growth medium as indicated. Equal amounts of protein from crude detergent lysates were

analyzed on a 12% SDS-PAGE gel and visualized by scanning of the gel with a Typhoon flatbed laser scanner. Probe-labeled cathepsin B (Cat B) and

cathepsin L (Cat L) are indicated based on the reported specificity of the probes30. The asterisk denotes a nonspecific protein that labels in samples

containing serum. (c) Intact monolayers of cells were pretreated either with the cathepsin inhibitor GB111-NH2 (5 mM) or with control DMSO (0.1%) for

1.5 h and labeled by addition of 1 mM GB123 (left panel) or GB137 (right panel) to culture medium for 3 h. Crude detergent lysates were normalized for

total protein, separated by SDS-PAGE and visualized as in b.

ART ICL ES

67 0 VOLUME 3 NUMBER 10 OCTOBER 2007 NATURE CHEMICAL BIOLOGY

quencher and the occluding loop30. In addition, GB137 did not label ahighly abundant background protein at 66 kDa that was observed forall other probes and that was also observed whenever probes wereused for in vivo applications. We therefore chose this probe alongwith the unquenched probe GB123 for advancement into in vivoimaging studies.

To decide on an optimal cell line for use in our tumor imagingstudies, we profiled the cysteine cathepsin activities in a number ofcommonly used cancer cell lines that are suitable for use in tumorgraft studies (Fig. 1c). GB123 and GB137 were applied to a panel ofhuman and mouse cells, and GB111-NH2 (9), a previously reportedpotent cathepsin inhibitor30, was used to control for specificity oflabeling by the probes. In all cell lines tested, specific labeled proteaseswere observed in the 20–30 kDa size range. Notably, overall cathepsinactivity varied over a more than ten-fold range among different celllines. The highest cathepsin activity was observed in a mouse skeletalmyoblast cell line (C2C12) retrovirally transduced with the Hras1

oncogene. These C2C12/Hras1 cells rapidly produce tumors followingsubcutaneous injection into BALB/c nude mice and are thereforesuitable for imaging studies of grafted tumors32. We also observed asimilar pattern, although with reduced intensity, for labeled cathepsinsin the human breast cancer cell lines MDA-MB 231 and MDA-MB435, and in the related human cell line MDA-MB 231 MFP. This line isthe same as the MDA-MB 231 cell line except it was isolated afterin vivo passage of the parent cells33. These cells are advantageousbecause they grow more readily in vivo than the parent line MDA-MB231. We therefore selected the C2C12/Hras1 cells, MDA-MB 231 andMDA-MB 435 cells and MDA-MB 231 MFP for our imaging studiesbecause these lines express similar cathepsins but show markeddifferences in overall activity.

Before moving into in vivo applications, we wanted to verify the cellpermeability and labeling specificity of the probes. Intact C2C12/Hras1 cells were preincubated with either JPM-OEt (10), a generalpapain-family inhibitor34, or vehicle control, labeled with GB123 or

Hours after injection

Hours after injection

Hours after injection

GB125

cm–2

GB123

GB138

3.0

2.5

1.0

1.5

2.0

× 107

b

MDA-MB 231 MFP

MDA-MB 231 MFP0 5 10 15 20 25

GB138

cm–2

GB125

MDA-MB 231 MFP

1.0

0.9

0.5

0.7

0.8

× 108

0.6

0 5 10 15 20 250

10

15

20

5

0

10

15

20

5

Baseline 0.5 h 5 h 23 h

n = 3

n = 4

n = 2

4 h2 h 8 h 10 h 24 h12 hBaseline5.0

4.5

4.0

3.5

× 108

a

MDA-MB#231435

p sec–1 cm–2 sr–1

cGB123

cm–2

0 5 10 15 20 250

10

15

20

5

Baseline 0.5 h 5 h 23 h

2.8

3.2

3.6 × 108

4.0

Baseline 0.5 h 5 h 23 h

p sec–1 c m–2 sr–1

Figure 2 Optical imaging of tumors in live mice using nonquenched NIRF-ABPs. (a) MDA-MB 435 or MDA-MB 231 cells were injected subcutaneously to

the left and right sides of the back of BALB/c nude mice, 3–4 weeks before imaging. GB123 was injected intravenously via the tail vein, and fluorescent

images of the dorsal side of living mice were taken at various time points after injection. The colorometric scale shows photons per second per square

centimeter per steradian (p s–1 cm–2 sr–1) overlaid on bright-light images. (b) MDA-MB 231 MFP cells were injected subcutaneously on the back of BALB/c

mice as in a 2 weeks before imaging. GB123 (1.2 mg g–1) or the identical probe GB138 bearing the IRDye 800 fluorophore in place of Cy5 (1.5 mg g–1) or

the control probe GB125 (1.1 mg g–1) was injected as in a, and fluorescent images of the dorsal side of living mice were taken at various time points.

(c) Signal-to-background ratios were calculated from images in b based on the fluorescence signals of tumor relative to background measured for a region on

the back (red circles). Ratios were normalized to half the surface area of each tumor calculated from tumor weight (assuming tumor as sphere shape and

density to be 1 g cm–3). The number of mice used to derive statistical information for each probe is indicated. Data represent mean ± standard error when

n ¼ 3 and n ¼ 4 mice, and mean ± range when n ¼ 2 mice.

ART ICL ES

NATURE CHEMICAL BIOLOGY VOLUME 3 NUMBER 10 OCTOBER 2007 6 7 1

GB125, washed to remove free probe and visualized using a fluores-cence microscope (Supplementary Fig. 1 online). As expected, thespecific probe GB123 showed punctate labeling of lysosomal compart-ments that overlapped with the acidotropic lysosomal marker Lyso-Tracker (Molecular Probes). Pretreatment of cells with JPM-OEtcompletely blocked GB123 staining, which suggests that the signalwas due to a specific modification of cysteine cathepsins. In addition,the control probe GB125 showed no labeling in cells after washout butdid accumulate in cells that had not been washed (data not shown),which confirms that the lack of labeling was not due to a loss of cellpermeability. Thus, both probes freely penetrate cells but onlyGB123 is retained, owing to irreversible modification of lysosomalcysteine cathepsins.

In vivo imaging of cathepsin activity in tumor-bearing miceBecause we wanted to monitor the distribution and half-lives of thefluorescent probes in vivo, we focused our initial efforts on thenonquenched probes. We selected the MDA-MB 231 and MDA-MB435 cell lines to use in tumor graft studies because they show similarpatterns of active cathepsins, except that the MDA-MB 435 line showssignificantly less overall activity compared with the MDA-MB 231 line.Cells were injected subcutaneously into ventral and dorsal locations inBALB/c nude mice. After a period of 3–4 weeks of growth in vivo,GB123 was injected via tail vein and animals were imaged using acharge-coupled device (CCD) camera-based imaging system at var-ious times after probe injection (Fig. 2a). As expected, the fluorescentprobe rapidly circulated throughout the animal and high fluorescentsignals could be seen in virtually all tissues, including the tumors.Thereafter, the signal faded continuously over most regions of the

body, but remained elevated in the tumors. Similarly, tumors on theventral side of the animal could be visualized; however, substantiallymore fluorescence was observed outside the tumors, even at the latetime points (Supplementary Fig. 2 online). It should be noted thatthis signal in the main body cavity is likely due to a combination oflabeling of active cathepsins in organs that have high cathepsin activityand retention of the free probe in tissues. Regardless, the overallbackground signal decreased more rapidly than the signal in thetumor tissues.

We next performed similar imaging studies using the nonquenchedGB123 and compared it to GB138 (11), a similar probe labeled withthe IRDye 800 dye (LI-COR; structure shown in SupplementaryFig. 3 online). This dye absorbs and emits higher wavelength light(774 nm excitation, 804 nm emission) than Cy5 and thereforeproduces images with less background resulting from tissue auto-fluorescence (Fig. 2b). In addition, we compared these two probeswith the Cy5-labeled control probe GB125, which lacks the reactivewarhead group and fails to label any targets in cells grown in culture(Fig. 2b). All probes were injected in multiple animals to allow properstatistical analysis of images. As expected, all three probes circulatedthroughout the animals and showed general background fluorescenceat the short time points. Importantly, only the two active probes thatcontain the AOMK warhead (GB123 and GB138) showed accumula-tion in the tumors. The control probe GB125, on the other hand, wasrapidly cleared and no fluorescent signal above background could beobserved by the 5 h time point. These data confirm that the signalobserved in tumors and other tissues results from modification ofspecific protein targets and is not due to nonspecific accumulation ofthe probes. Quantification of the overall signal-to-background ratio

0

25

5

20

30

40

35

15

10

45 Tumor flux in image

123

MDA-MB 231Tumor origin:

*

Probe GB#:

97

66

42

36

28

20

C2C12/Hras1435

125 123123

123

MDA-MBTumor origin:

Tumor image:

Probe GB#:

C2C12/Hras1435

125 123123

c

Input S P S P

Ex/Em 633/670Ex/Em 633/670

Cat B

α Cat B α Cat L

Cat L

97

66

42

36

28

20

ba

n = 3

0

20

30

40

10

50

60

123

231MDA-MBTumor origin:

Probe GB#:

C2C12/Hras1435

125 123123

Intensity of labeling of 20–30 kDa

Arb

itra

ry u

nits

n = 2

n = 4

n = 4

n = 3n = 3

n = 3

n = 2

p s

ec

–1

cm

–2

× 1

09

0.5 cm

Gel image:

*

231

Figure 3 Biochemical characterization of in vivo–

labeled proteases. (a) MDA-MB 231, MDA-MB

435 or C2C12/Hras1 cells were injected

subcutaneously on the back of BALB/c nude

mice, 2–4 weeks before imaging. Mice with

similarly sized tumors were injected intravenously

via the tail vein with 25 nmol GB123 or GB125

as indicated. 24 h after probe injection, tumors

were excised and crude lysates were normalized

for total protein, separated by SDS-PAGE and

visualized by fluorescence scanning of the gel

with a flatbed laser scanner. (b) Lysates

(170 mg) from the C2C12/Hras1 tumors were

immunoprecipitated using anti-cathepsin B or

anti-cathepsin L antibodies as indicated.Supernatant and pellet samples were separated

by SDS-PAGE and visualized by scanning of the

gel as in a. (c) Quantitative analysis of in vivo

tumor images with labeling intensity of protease

targets. Fluorescence intensity in tumors after

background subtraction was quantified for each

animal expressed relative to tumor surface area

(left panel). Images of representative tumors

corresponding to each condition and probe are

shown. The intensity of labeled cysteine

cathepsins in the 20–30 kDa size range in a was

quantified from the gel image for each animal

(right panel). Data represent mean values ±

standard errors, or mean values ± range for

n ¼ 2. Images of the region of the gel used for

quantification are shown below each bar in the

graph. The number of mice used to derive

statistical information for each conditionis indicated. Asterisk denotes a nonspecific

protein that is labeled in serum.

ART ICL ES

67 2 VOLUME 3 NUMBER 10 OCTOBER 2007 NATURE CHEMICAL BIOLOGY

for each probe shows that both the GB123 and GB138 probes producestrong specific signals in tumors and that the higher wavelength IRDye800 dye provides higher overall signal-to-background ratios (Fig. 2c).We also found that the IRDye 800 changed the overall distribution ofthe probe, thereby leading to enhanced labeling of cathepsins intumors and reduced labeling of cathepsins in other tissues such asliver, spleen and kidney (Supplementary Fig. 3). In addition, theseresults indicate that specific accumulation of the probe in tumors isnot fluorophore dependent, as multiple probes produced similarresults with comparable signal-to-background ratios.

One of the primary advantages of using covalent activity-basedprobes is that it is possible to directly monitor the labeling of specificproteins by simple SDS-PAGE analysis of tissue homogenates collectedfrom animals treated with the probes. Thus, to identify proteasetargets and to further exclude the possibility that specific signalsobserved in the whole-body images resulted from nontarget-mediatedaccumulation of the probes, we performed biochemical analysis oftumors derived from MDA-MB 231, MDA-MB 435 and C2C12/Hras1cells that had been in vivo–labeled by either the active probe GB123 orthe control probe GB125 (Fig. 3a). After noninvasive imaging of the

123Probe GB#:

20

28

36

42

66

14

137

Tumor

*

123137 123137123137123137

Liver Kidney Spleen Brain

97

200

0.5 h 1.5 h 5 h 12 h 24 h

1.0

2.5

2.0

1.5

3.0

× 108

× 108

GB

137

GB

123

Signal to background

cm–2

Hours after injection

GB137

GB123

0 6 8 10 12 1442

6

8

10

4

2

2.5

4.0

3.5

3.0

4.5

n = 3

MDA MB231 MFP

MDA MB231 MFP

Baseline

p sec–1 cm–2 sr–1

a

b c

Figure 4 Direct comparison of the nonquenched and quenched NIRF-ABPs GB123 and GB137. (a) Optical imaging of tumors in live mice. MDA-MB 231

MFP cells (1 � 106) were injected subcutaneously into the back of BALB/c nude mice 2 weeks before imaging. GB123 (1.2 mg g–1) or GB137 (2 mg g–1)

dissolved in DMSO/PBS was injected intravenously via the tail vein, and fluorescent images of the dorsal side of living mice were taken at various time pointsafter injection using an IVIS 200 imaging system equipped with a Cy5.5 filter. Images are presented using a colorometric scale based on photons per second

per square centimeter per steradian (p s–1 cm–2 sr–1) overlaid on bright-light images. (b) Quantification of signal-to-background ratios expressed relative to

tumor surface area. Data represent mean values ± standard errors for each time point; three independent experiments including two mice each were

performed. (c) Biochemical characterization of tissues from mice treated with GB123 and GB137. The indicated tissues were collected and lysed by dounce

in detergent buffer. Crude lysates were normalized for total protein, separated by SDS-PAGE and visualized by fluorescence scanning of the gel with a

Typhoon flatbed laser scanner. A nonspecific 66 kDa serum protein is indicated with an asterisk.

ART ICL ES

NATURE CHEMICAL BIOLOGY VOLUME 3 NUMBER 10 OCTOBER 2007 6 7 3

tumors on the live mice, tissues were collected and homogenized andnormalized protein samples were analyzed by SDS-PAGE followed byscanning of the gel with a laser scanner. Three predominant labeledproteins were observed in the 20–30 kDa size range. This labelingpattern was similar to the pattern of labeled cathepsins observed in theparent cell lines (Fig. 1c). The 30 kDa protein was confirmed to becathepsin B, and the 23 kDa protein was identified as the heavy chainform of cathepsin L by immunoprecipitation using specific antisera(Fig. 3b). Though the identity of the 25 kDa protein could not bedetermined with available antibodies, it is likely to be a cysteinecathepsin given that its labeling could be blocked by treatment ofanimals with specific inhibitors. Importantly, the overall intensity oflabeling of the cathepsins observed by gel analysis strongly correlatedwith the intensity of the signal observed in the whole-body images(Fig. 3c), which suggests that the signals observed in the live animalswere due to specific probe modification of active cysteine cathepsins.

In vivo analysis of quenched NIRF-ABPsOur initial results with the nonquenched probes suggested that theyare effective tools for in vivo imaging applications. However, thesenonquenched probes suffer from high background labeling thatrequires extended waiting periods before images with sufficientsignal-to-background ratios can be generated. We reasoned that thequenched probe GB137 should have significant benefits at early timepoints after introduction of the probe. We therefore performed adirect comparison of GB123 and GB137 in mice bearing the MDA-MB231 MFP tumors (Fig. 4a). Animals were injected with the probes andimages were collected over a 24 h period. Unlike the nonquenchedprobe, the quenched probe GB137 produced a specific signal in thetumors and showed virtually no background labeling starting at theearliest time point. This specific signal in the tumors increased withtime and reached a maximum by 6–8 h after probe injection.

Quantification of the overall signal-to-background ratios for multipleanimals for each probe indicated that GB123 and GB137 producesimilar overall signal-to-background ratios; however, the quenchedprobe achieved its maximum much more rapidly than thenonquenched probe (Fig. 4b). To confirm that the signal observedin animals treated with GB137 was due to specific modification ofprotease targets, we collected tissues from various organs and fromtumors and analyzed these samples by SDS-PAGE (Fig. 4c). Asobserved for intact cells (Fig. 1b), the GB137 probe labeled thesame three cathepsin protease species in tumors, liver, spleen andkidney tissues. In addition, GB137 did not show any labeling of thehighly abundant serum protein at 66 kDa, whereas GB123 showedstrong labeling of this protein. This result confirms that the 66 kDabackground protein does not affect images generated in vivo becauseboth probes produced similar noninvasive images. We believe thisbackground protein is likely to be serum albumin, which contaminatesthe samples when tissues are collected. Overall, these datafurther support the notion that these NIRF-ABPs produce specificsignals in tumors as a result of activity-dependent modification ofprotease targets.

Imaging of in vivo efficacy of small-molecule inhibitorsTo further confirm that signals observed in whole-body images weredue to protease activity and to show that these probes have value astools for in vivo monitoring of target inhibition, we used a broad-spectrum papain-family inhibitor to reduce protease activity in theanimals before imaging with the quenched and nonquenched probes.We initially began these studies using the general inhibitor JPM-OEtbecause it is characterized by broad reactivity with the cysteinecathepsins and has been used in mouse models of cancer toblock activity of this family of proteases19. However, all of ourinitial attempts to reduce cysteine cathepsin activity in tumors by

123 GB#:

20

28364266

Tumor

*

123137 123 137123 137123 137Liver Kidney Spleen Muscle

97200

137K11777 – + – + – + – + – + – + – + – + – + – +

K11777 VehicleGB137

K11777 VehicleGB123

GB137 GB123

K11777 VehicleGB123

K11777 VehicleGB137

K11777 Vehicle05

2530

1520

10

K11777 Vehicle0

12

8

4

163.5

3.0

1.5

2.0

2.5

1.0

0.6

0.7

0.8

× 109

× 109

1.1

0.9

2.4

1.6

2.0

3.2

2.88.0

7.0

4.0

5.0

6.0× 10

8

× 109

9.0

10 h 10 h

n = 3 n = 3

n = 2 n = 2

K11777 Vehicle

GB137 Tumor gel quantification

020406080

100

120140

K11777 Vehicle

GB123 Tumor gel quantification

p sec–1

cm–2 sr–1

p sec–1

cm–2 sr–1p sec–1

cm–2 sr–1p sec–1

cm–2 sr–1

p se

c–1

cm–2

× 10

7

p se

c–1

cm–2

× 10

8

Per

cent

age

rela

tive

to v

ehic

le

020406080

100120140

Per

cent

age

rela

tive

to v

ehic

le

a

d e

b c

Figure 5 Imaging of small-molecule inhibitor efficacy. (a) MDA-MB 231 MFP cells were injected subcutaneously on the back of BALB/c nude mice

2 weeks before imaging. Animals were treated with the general inhibitor K11777 (50 mg kg–1 twice daily) or DMSO vehicle for 5 d before labeling. Mice

were injected intravenously with 25 nmol GB123 or GB137 and then imaged at various times up to 10 h. Images for representative mice at the 10 h time

points are shown. Images use a colorometric scale based on photons per second per square centimeter per steradian (p s –1 cm–2 sr–1) overlaid on bright-

light images. (b) Total fluorescence signal relative to tumor surface area after background subtraction was calculated for each animal. Data represent mean

values ± standard errors for n ¼ 3 or mean values ± range for n ¼ 2. (c) Tumors from inhibitor- and vehicle control–treated mice in a were surgically

removed and tumor masses imaged as in a. (d) Tissues from vehicle- and inhibitor-treated mice were collected, homogenized and analyzed by SDS-PAGE

followed by scanning of the gel using a laser fluorescent scanner. (e) Intensities of the four predominant labeled proteins in tumor tissues in d were

quantified and plotted as mean values relative to vehicle treatment (designated as 100%) with calculated standard error. Error bars represent range for

vehicle samples. Two separate experiments including two animals treated with vehicle and three animals treated with K11777 were performed.

ART ICL ES

67 4 VOLUME 3 NUMBER 10 OCTOBER 2007 NATURE CHEMICAL BIOLOGY

intraperitoneal injection of JPM-OEt failed. Even after treatment withhigh doses of drug (100 mg kg–1 day–1) over a 3-d period beforeimaging, we observed no reduction of fluorescent signals in tumors. Inagreement with the imaging result, we observed no reduction inlabeling of cathepsins in tumor tissues assayed biochemically bySDS-PAGE analysis of tissue extracts (data not shown).

We reasoned that our inability to block cysteine cathepsin activitywas due to poor distribution and cellular uptake of the inhibitor intografted tumor tissues. In our previous studies of tumors in atransgenic model of pancreatic cancer, this same inhibitor effectivelyblocked cysteine cathepsins19. These findings suggest that the lack ofinhibition is due to the nature of grafted tumors, which may generallybe less accessible to this class of small-molecule inhibitors than tumorsthat develop in situ in their native microenvironment.

We therefore decided to switch to the MDA-MB 231 MFP cells anduse a broad-spectrum inhibitor that has more favorable cell perme-ability properties. The inhibitor K11777 (12) is a dipeptide vinylsulfone35 that has been used for extensive in vivo preclinical animalstudies owing to its potential use as a drug for treatment of Chagasdisease (J. McKerrow, University of California, San Francisco, personalcommunication). In addition, we treated mice for 5 d before imagingin order to more completely reduce cathepsin activity in tumortissues. Imaging showed that the specific signal in tumors at the10 h time point was reduced in mice treated with K11777 (Fig. 5a).Importantly, we obtained similar results with both the GB123 andGB137 probes. Quantification of total tumor fluorescence indicatedthat signal was reduced by 60–80% in mice treated with K11777relative to control mice (Fig. 5b). To further confirm this reduction insignal in tumors and to reduce background, we removed solid tumormasses from the mice and imaged them using the same CCD camerasystem as described above (Fig. 5c). Again the signal intensity in thetissues of the K11777-treated mice was reduced, corresponding to theimages obtained in the live animals.

Finally, we analyzed the tumor tissues, as well as tissue from liver,kidney, spleen and muscle, by SDS-PAGE to confirm that loss of signalin our whole-body images was due to a reduction in labeled cysteinecathepsins (Fig. 5d). Labeling profiles in all tissues from both treatedand control mice showed the same pattern of active cathepsinsobserved previously with the addition of a specific signal around45 kDa that was particularly pronounced in liver. This additionalsignal is likely to be the pro form of a cysteine cathepsin that can belabeled by the active site probe if it accumulates at sufficient con-centrations. Such proenzyme labeling has been observed for thecysteine cathepsins and is thought to result from the reversiblebinding nature of the inhibitory prodomain36. Importantly, quantifi-cation of the biochemical gel data indicated that labeling of specificcysteine cathepsins was reduced in tissues from the mice treated withK11777 by 50–80% (Fig. 5e), which correlates well with the observedimaging data. Overall, these data further support the notion thatNIRF-ABPs provide a fluorescent readout of protease activity in aliving animal and that the modulation of this activity using small-molecule therapeutics can be monitored by direct whole-bodyimaging methods.

DISCUSSIONProteases are among the most extensively studied enzyme familiesowing to their involvement in regulation of diverse cellular processesand their potential value as disease biomarkers and as targets forsmall-molecule therapeutics. In this study we present classes offluorescently quenched and nonquenched NIRF-ABPs that allow theactivity of the cysteine cathepsins B and L to be visualized in living

subjects. The covalent nature of the probes enables a direct linkbetween imaging data and the biochemical readout of labeled targets,thus permitting assignment of imaging signals to specific proteasetargets. Importantly, these probes allow direct in vivo analysis of drugefficacy and pharmacodynamic properties.

There have been a number of recent technological advances thathave enhanced our ability to monitor enzyme activity using non-invasive imaging methods22–24. These new methods have mainly reliedon substrates that become fluorescent or change cellular localizationwhen processed by a given enzyme target26–29. While they are valuableadditions to the imaging toolbox, these methods suffer from a numberof shortcomings that limit their utility for studies of specific proteasetargets. In particular, the large number and diversity of proteases in agiven cell or organism hinders the development of highly specificsubstrates. Although there are examples of peptide substrates thatshow relatively high selectivity for a given protease target37,38, it isgenerally difficult to control protease processing of peptide-basedsubstrates. Furthermore, there is no direct way to determine whichprotease is responsible for the processing of a given substrate in vivo.ABPs have advantages over substrates in that they can be designed tobe highly specific for distinct subfamilies of proteases. The AOMK-based probes used in the studies presented here are highly selective forcysteine proteases and have been engineered to target only a smallsubset of this family30,39. The main drawback of using a covalent probeis the lack of amplification of signal since the target proteases areinactivated upon binding the probe. Although we were initiallyconcerned that the general abundance of active proteases wouldseverely limit the utility of an ABP for imaging applications, wehave found that sufficient concentrations of these proteases exist intumor tissues to allow the generation of reasonable contrast imagesusing noninvasive methods. Notably, the images generated with ourNIRF-ABPs show signal-to-background ratios that are similar to thosepublished previously for substrate-based probes27,29, which suggeststhat amplification of signal by processing of multiple reportermolecules may not constitute a limiting factor in the control ofprobe sensitivity.

In addition to being valuable tools for monitoring protease activityin living subjects for the purpose of disease diagnosis, NIRF-ABPshave potential for assessing the specificity of therapeutic agentsdesigned to target a particular protease. In many cases, drug leadsare advanced or dropped in preclinical testing based on efficacy data ina particular model of disease. When compounds fail to show efficacy,it is often difficult to determine whether this is due to a poor choice oftargets or to problems with distribution of the compound in vivo. Forexample, in the study presented here, we made several attempts toreduce probe labeling by treating mice with high doses of the proteaseinhibitor, JPM-OEt. Our initial failed attempts to detect a reduction insignal with increased JPM-OEt could have led us to believe that thesignals obtained by imaging in tumors were the result of nonspecificbackground. Because we could correlate the imaging results withbiochemical analyses of tumor tissues, this failure could be ascribedto the drug treatment, which did not lead to the expected reduction inprotease activity in the tumor tissues. This example illustrates theproblem of drug access associated with the use of the graftedtumor model. More importantly, it highlights the value of ourprobes, which allow validation, in vivo, of the pharmacodynamicproperties of small molecules and other therapeutic agents by acombination of complementary assays. The potential to establish theefficacy of drug targeting in preclinical and clinical settings, whichthis technology enables, should help to accelerate the drugdevelopment process.

ART ICL ES

NATURE CHEMICAL BIOLOGY VOLUME 3 NUMBER 10 OCTOBER 2007 6 7 5

In conclusion, NIRF-ABPs are promising new tools that can now befurther developed for applications in noninvasive imaging. In parti-cular, the quenched probes can be used to image specific proteaseactivity at considerably earlier time points than can be used forsubstrate-based methods or nonquenched ABPs. Thus, quenchedprobes have the potential to be used for intraoperative proceduresin which rapid production of specific signal is required. In addition,the approach described here can be adapted to target other classes ofproteases as well as other families of enzymes. Indeed, a number ofnewly validated classes of ABPs have been reported in the last fiveyears, and more will inevitably be developed with potential for use asimaging probes2. A current limitation of this technology is that it isonly applicable to superficial tissues, and the high levels of signal inlarge organs with high cathepsin activity such as liver, kidney andspleen make imaging of specific locations within the central bodycavity difficult. However, one can envision improvements in imagingtechnologies that will allow fluorescently labeled probes to be used indeeper tissue targets in larger subjects. Furthermore, improvements influorescence tomography methods will also allow these probes to beused for imaging in specific locations with high levels of ‘background’labeling. In addition, ABPs can be coupled to radioactive reporters toallow radiological imaging (for example, PET and SPECT) of probe-modified targets. Work on the development of (i) additional qNIRF-ABPs that make use of the higher wavelength IR800 dyes and (ii) PETand SPECT ABPs is currently underway and will likely lead toincreased utility for these valuable reagents in the near future.

METHODSGeneral methods. Unless otherwise noted, all resins and reagents were

purchased from commercial suppliers and used without further purifications.

All solvents used were HPLC grade. All water-sensitive reactions were pre-

formed in anhydrous solvents under positive pressure of argon. Reactions were

analyzed by LC-MS using an API 150EX single-quadrupole mass spectrometer

(Applied Biosystems). Reverse-phase HPLC was conducted with an AKTA

explorer 100 (Amersham Pharmacia Biotech) using C18 or C4 columns.

Fluorescent gels were scanned using a Typhoon 9400 flatbed laser scanner

(GE Healthcare). Male BALB/c nude mice (4–8 weeks old) were obtained from

the Stanford University Department of Comparative Medicine. All animal

protocols were approved by the Stanford Administrative Panel on Laboratory

Animal Care, and the procedures were performed in accordance to their

guidelines. Statistical analysis was preformed using Microsoft Excel, and

s.e.m. was calculated by dividing the s.d. by the square root of n.

Cell cultures. NIH3T3 mouse fibroblast cells (see Acknowledgments), mouse

myoblast cells C2C12 retrovirally transduced to express the Hras1 oncogene

(using the ecotropic FNX packaging cell line), and U87MG cells were cultured

in DMEM (GIBCO) supplemented with 10% fetal bovine serum (FBS;

GIBCO), 100 units ml–1 penicillin and 100 mg ml–1 streptomycin (GIBCO).

MDA-MB 231 and MDA-MB 435 human epithelial adenocarcinoma cells and

MDA-MB 231 MFP human epithelial adenocarcinoma were cultured in

Leibovitz’s L-15 medium (GIBCO) supplemented with 10% FBS, 100 units

ml–1 penicillin and 100 mg ml–1 streptomycin at 37 1C (with no CO2 addition).

Labeling of intact cells with NIRF-ABP. MDA-MB 231 cells and MDA-MB

231 MFP (200,000 cell well–1) and NIH3T3, C2C12/Hras1, U87MG and MDA-

MB 435 cells (240,000 cell well–1) were seeded in a 6-well plate 1 d before

treatment. Cells were pretreated with the papain-family protease inhibitor

GB111-NH2 (ref. 30) (5 mM) or with control DMSO (0.1%) for 1.5 h and

labeled by addition of GB123 or GB137 (1 mM) in growth medium DMSO

(0.1%) for 3 h; the final DMSO concentration was maintained at 0.2%. Cells

were washed with phosphate-buffered saline (PBS) and lysed by addition of

sample buffer (10% glycerol, 50 mM Tris HCl, pH 6.8, 3% SDS and 5%

b-mercaptoethanol). Crude lysates were spun down and protein content was

measured. Equal amounts of protein per lane were separated by 12%

SDS-PAGE, and labeled proteases were visualized by scanning of the gel with

a Typhoon flatbed laser scanner (excitation/emission 633/680 nm).

Mouse serum sensitivity. NIH3T3 (240,000 cells well–1) were seeded in a 6-well

plate 1 d before treatment. Cells were pretreated either with general papain-

family protease inhibitor JPM-OEt (50 mM) or with control DMSO (0.1%)

for 1 h and then labeled with 1 mM (final concentration) probe for 3 h. Probes

(1 ml) were either incubated in 9 ml mouse serum (Invitrogen) for 1 or 4 h

at 37 1C or incubated at 37 1C without serum, before addition to cells for

labeling. Equal numbers of cells were lysed by addition of sample buffer (10%

glycerol, 50 mM Tris HCl, pH 6.8, 3% SDS and 5% b-mercaptoethanol) and

then separated by SDS-PAGE. Labeled proteases were visualized by scanning of

the gel with a Typhoon flatbed laser scanner (excitation/emission 633/680 nm).

In vivo imaging. MDA-MB 231, MDA-MB 435, MDA-MB 231 MFP or C2C12/

Hras1 cells were grown to subconfluencey and then detached with trypsin, spun

down and resuspended in 0.5% BSA (Fisher Biotech) in sterile PBS (GIBCO) at

1 � 108 cells ml–1. Cells (1–4 � 106 per spot) were injected subcutaneously at

the indicated locations on 4–8 week old male BALB/c nude mice under

isoflurane anesthesia. 2–4 weeks after cell implantation, probe (25 nmol;

1.2 mg g–1 GB123, 1.1 mg g–1 GB125, 2.0 mg g–1 GB137 or 1.5 mg g–1 GB138)

dissolved in 66% DMSO in PBS at a total volume of 100 ml was injected

intravenously via the tail vein to tumor-bearing mice. Mice anesthetized with

isoflurane were imaged at various time points after probe injection using the

IVIS 200 imaging system (Xenogen) equipped with a Cy5.5 filter (for GB123-,

GB137- and GB125-treated mice) or an ICG filter (for GB138-treated mice).

Relative fluorescence of equal-sized areas of tumor and back were measured

using Living Image (Xenogen). After the last time point of imaging, mice were

anesthetized with isoflurane and killed by cervical dislocation. Tumors, liver

and kidney, spleen and muscle, or brain were surgically excised and frozen in

liquid nitrogen. Organs were either lysed using bead beater, dounced in PBS

buffer, pH 7.2 (1% Triton X-100, 0.1% SDS, 0.5% sodium deoxycholate, 0.2%

sodium azid), or imaged ex vivo using the IVIS 200 imaging system before lysis

of tissues. Total protein extracts (120 mg) were separated by SDS-PAGE and

visualized by scanning of the gel with a Typhoon flatbed laser scanner

(excitation/emission 633/680 nm). Intensity of bands was measured using

NIH Image (http://rsb.info.nih.gov/nih-image/). For inhibitor studies, com-

pounds were injected intraperitoneally twice daily in 40% DMSO/sterile PBS in

a final volume of 100 ml. Inhibitor (K11777, 100 mg kg–1 day–1) or vehicle was

administered over 5 d before probe injection.

Cathepsin immunoprecipitation. Total protein extracts (170 mg) from C2C12/

Hras1 tumors locally injected with DMSO and in vivo–labeled with GB123

were diluted to 500 ml in RIPA buffer (50 mM Tris HCl, pH 7.4, 150 mM NaCl,

5 mM EDTA, 0.5% deoxycholate, and 0.1% SDS). Anti-mouse cathepsin B (E

Weber Halle) or anti-mouse cathepsin L (R&D Systems) were added and mixed

for 30 min at 4 1C. Protein A/G Plus Agarose beads (10 ml, Santa Cruz

Biotechnology) were added and mixed overnight at 4 1C. Samples were

centrifuged and supernatant and beads were collected. Beads were washed

with RIPA buffer and boiled after addition of sample buffer 2X (20% glycerol,

100 mM Tris HCl, pH 6.8, 6% SDS, and 10% b-mercaptoethanol). Acetone at

0 1C (1 ml) was added to the supernatant and kept at –80 1C for 2 h.

Precipitated proteins were collected by centrifugation in the cold for 30 min,

dried, and dissolved by boiling in sample buffer. Samples without immuno-

precipitation (input), pellets (P) and supernatant (S) were separated on a 14%

Tris-glycin gel (Invitrogen) and visualized by scanning of the gel with a

Typhoon flatbed laser scanner (excitation/emission 633/680 nm).

Chemical synthesis and characterization of NIRF-ABPs. See Supplementary

Methods online.

Note: Supplementary information and chemical compound information is available onthe Nature Chemical Biology website.

ACKNOWLEDGMENTSWe thank T. Doyle and S. Keren from the Stanford Small Animal Imaging Facilityand T. Troy from Caliper Life Sciences for expert advice and technical assistance.We thank K.B. Sexton and S. Verhelst for valuable discussions and technicalassistance. The authors thank C. Gilon for helpful advice on peptide synthesis.

ART ICL ES

67 6 VOLUME 3 NUMBER 10 OCTOBER 2007 NATURE CHEMICAL BIOLOGY

We thank P. Jackson (Stanford University) for the mouse fibroblast cells,G.P. Nolan (Stanford University) for the ecotropic FNX packaging cell line,S. Gambhir (Stanford University) for the U87MG cells, X. Chen (StanfordUniversity) for the MDA-MB 435 human epithelial adenocarcinoma cells, andB. Cravatt (The Scripps Research Institute) for the MDA-MB 231 MFP humanepithelial adenocarcinoma. This work was supported by the US NationalInstitutes of Health National Technology Center for Networks and Pathways(grants U54 RR020843, R01 EB005011 and P01 CA072006), the US Departmentof Defense Breast Cancer Center of Excellence (grant DAMD-17-02-0693; toM.B.), a Susan G. Komen postdoctoral fellowship (to G.B.) and a grant from theDeutsche Forschungsgemeinschaft (DE 740/1-1; to G.v.D.).

COMPETING INTERESTS STATEMENTThe authors declare no competing financial interests.

Published online at http://www.nature.com/naturechemicalbiology

Reprints and permissions information is available online at http://npg.nature.com/

reprintsandpermissions

1. Berger, A.B., Vitorino, P.M. & Bogyo, M. Activity-based protein profiling: applications tobiomarker discovery, in vivo imaging and drug discovery. Am. J. Pharmacogenomics 4,371–381 (2004).

2. Evans, M.J. & Cravatt, B.F. Mechanism-based profiling of enzyme families. Chem. Rev.106, 3279–3301 (2006).

3. Sadaghiani, A.M., Verhelst, S.H. & Bogyo, M. Tagging and detection strategies foractivity-based proteomics. Curr. Opin. Chem. Biol. 11, 20–28 (2006).

4. Saghatelian, A. & Cravatt, B.F. Assignment of protein function in the postgenomic era.Nat. Chem. Biol. 1, 130–142 (2005).

5. Schmidinger, H., Hermetter, A. & Birner-Gruenberger, R. Activity-based proteomics:enzymatic activity profiling in complex proteomes. Amino Acids 30, 333–350(2006).

6. Greenbaum, D. et al. Chemical approaches for functionally probing the proteome. Mol.Cell. Proteomics 1, 60–68 (2002).

7. Patricelli, M.P., Giang, D.K., Stamp, L.M. & Burbaum, J.J. Direct visualization of serinehydrolase activities in complex proteomes using fluorescent active site-directed probes.Proteomics 1, 1067–1071 (2001).

8. Saghatelian, A., Jessani, N., Joseph, A., Humphrey, M. & Cravatt, B.F. Activity-basedprobes for the proteomic profiling of metalloproteases. Proc. Natl. Acad. Sci. USA 101,10000–10005 (2004).

9. Sieber, S.A., Niessen, S., Hoover, H.S. & Cravatt, B.F. Proteomic profiling of metallo-protease activities with cocktails of active-site probes. Nat. Chem. Biol. 2, 274–281(2006).

10. Lecaille, F., Kaleta, J. & Bromme, D. Human and parasitic papain-like cysteineproteases: their role in physiology and pathology and recent developments in inhibitordesign. Chem. Rev. 102, 4459–4488 (2002).

11. Mueller-Steiner, S. et al. Antiamyloidogenic and neuroprotective functions of cathepsinB: implications for Alzheimer’s disease. Neuron 51, 703–714 (2006).

12. Jedeszko, C. & Sloane, B.F. Cysteine cathepsins in human cancer. Biol. Chem. 385,1017–1027 (2004).

13. Mohamed, M.M. & Sloane, B.F. Cysteine cathepsins: multifunctional enzymes incancer. Nat. Rev. Cancer 6, 764–775 (2006).

14. Foekens, J.A. et al. Prognostic significance of cathepsins B and L in primary humanbreast cancer. J. Clin. Oncol. 16, 1013–1021 (1998).

15. Harbeck, N. et al. Prognostic impact of proteolytic factors (urokinase-type plasminogenactivator, plasminogen activator inhibitor 1, and cathepsins B, D, and L) in primarybreast cancer reflects effects of adjuvant systemic therapy. Clin. Cancer Res. 7,2757–2764 (2001).

16. Jagodic, M., Vrhovec, I., Borstnar, S. & Cufer, T. Prognostic and predictive valueof cathepsins D and L in operable breast cancer patients. Neoplasma 52, 1–9 (2005).

17. Lah, T.T. et al. Cathepsin B, a prognostic indicator in lymph node-negative breastcarcinoma patients: comparison with cathepsin D, cathepsin L, and other clinicalindicators. Clin. Cancer Res. 6, 578–584 (2000).

18. Gocheva, V. et al. Distinct roles for cysteine cathepsin genes in multistage tumorigen-esis. Genes Dev. 20, 543–556 (2006).

19. Joyce, J.A. et al. Cathepsin cysteine proteases are effectors of invasive growthand angiogenesis during multistage tumorigenesis. Cancer Cell 5, 443–453(2004).

20. Urbich, C. et al. Cathepsin L is required for endothelial progenitor cell-inducedneovascularization. Nat. Med. 11, 206–213 (2005).

21. Vasiljeva, O. et al. Tumor cell-derived and macrophage-derived cathepsin B promotesprogression and lung metastasis of mammary cancer. Cancer Res. 66, 5242–5250(2006).

22. Choy, G., Choyke, P. & Libutti, S.K. Current advances in molecular imaging: non-invasive in vivo bioluminescent and fluorescent optical imaging in cancer research.Mol. Imaging 2, 303–312 (2003).

23. Licha, K. & Olbrich, C. Optical imaging in drug discovery and diagnostic applications.Adv. Drug Deliv. Rev. 57, 1087–1108 (2005).

24. Shah, K. & Weissleder, R. Molecular optical imaging: applications leading to thedevelopment of present day therapeutics. NeuroRx 2, 215–225 (2005).

25. Baruch, A., Jeffery, D.A. & Bogyo, M. Enzyme activity–it’s all about image. Trends CellBiol. 14, 29–35 (2004).

26. Bremer, C. et al. Optical imaging of spontaneous breast tumors using protease sensing‘smart’ optical probes. Invest. Radiol. 40, 321–327 (2005).

27. Jiang, T. et al. Tumor imaging by means of proteolytic activation of cell-penetratingpeptides. Proc. Natl. Acad. Sci. USA 101, 17867–17872 (2004).

28. Tung, C.H., Mahmood, U., Bredow, S. & Weissleder, R. In vivo imaging of proteolyticenzyme activity using a novel molecular reporter. Cancer Res. 60, 4953–4958(2000).

29. Weissleder, R., Tung, C.H., Mahmood, U. & Bogdanov, A., Jr In vivo imaging of tumorswith protease-activated near-infrared fluorescent probes. Nat. Biotechnol. 17,375–378 (1999).

30. Blum, G. et al. Dynamic imaging of protease activity with fluorescently quenchedactivity-based probes. Nat. Chem. Biol. 1, 203–209 (2005).

31. Ledakis, P. et al. Cathepsins D, B, and L in malignant human lung tissue. Clin. CancerRes. 2, 561–568 (1996).

32. D’Orazi, G. et al. Exogenous wt-p53 protein is active in transformed cells but not intheir non-transformed counterparts: implications for cancer gene therapy withouttumor targeting. J. Gene Med. 2, 11–21 (2000).

33. Jessani, N., Niessen, S., Mueller, B.M. & Cravatt, B.F. Breast cancer cell lines grownin vivo: what goes in isn’t always the same as what comes out. Cell Cycle 4, 253–255(2005).

34. Bogyo, M., Verhelst, S., Bellingard-Dubouchaud, V., Toba, S. & Greenbaum, D.Selective targeting of lysosomal cysteine proteases with radiolabeled electrophilicsubstrate analogs. Chem. Biol. 7, 27–38 (2000).

35. Palmer, J.T., Rasnick, D., Klaus, J.L. & Bromme, D. Vinyl sulfones asmechanism-based cysteine protease inhibitors. J. Med. Chem. 38, 3193–3196(1995).

36. Smith, S.M., Kane, S.E., Gal, S., Mason, R.W. & Gottesman, M.M. Glycosylation ofprocathepsin L does not account for species molecular-mass differences and is notrequired for proteolytic activity. Biochem. J. 262, 931–938 (1989).

37. Coombs, G.S. et al. Substrate specificity of prostate-specific antigen (PSA). Chem.Biol. 5, 475–488 (1998).

38. Liu, S., Bugge, T.H. & Leppla, S.H. Targeting of tumor cells by cell surface urokinaseplasminogen activator-dependent anthrax toxin. J. Biol. Chem. 276, 17976–17984(2001).

39. Kato, D. et al. Activity-based probes that target diverse cysteine protease families. Nat.Chem. Biol. 1, 33–38 (2005).

ART ICL ES

NATURE CHEMICAL BIOLOGY VOLUME 3 NUMBER 10 OCTOBER 2007 6 7 7