|Home | About | Journals | Submit | Contact Us | Français|
Imaging agents that enable direct visualization and quantification of apoptosis in vivo have great potential value for monitoring chemotherapeutic response as well as for early diagnosis and disease monitoring. We describe here the development of fluorescently labeled activity based probes (ABPs) that covalently label active caspases in vivo. We used these probes to monitor apoptosis in the thymus of mice treated with dexamethasone (dex) as well as in tumor-bearing mice treated with the apoptosis inducing monoclonal antibody Apomab. Caspase ABPs provided direct readouts of the kinetics of apoptosis in live animals, whole organs and tissue extracts. The probes produced a maximum fluorescent signal that could be monitored non-invasively and that coincided with the peak in caspase activity as measured by gel analysis. Overall, these studies demonstrate that caspase-specific ABPs have the potential to be used for non-invasive imaging of apoptosis in both pre-clinical and clinical settings.
As the number of therapies that function by inducing or inhibiting apoptosis continue to grow, imaging tools capable of tracking cell death will become increasingly important. Not surprisingly, strategies for monitoring apoptosis have been developed based on a wide range of surrogate biomarkers. These include specific apoptosis signaling molecules such as the caspases, as well as markers of downstream events in the apoptosis cascade.
One commonly used probe of apoptosis, Annexin V, is a protein with high affinity for phosphatidylserines exposed during the late stages of apoptosis. Annexin V has been labeled with a number of tags including fluorochromes1–3 and radioactive nuclides4. Labeled versions of Annexin have been used for imaging studies in mouse models of human cancer1 as well as for PET and SPECT based studies in humans5–7,8. While Annexin V represents a valuable label of apoptotic cells, its use in humans has had limited success partly due to its slow clearance in vivo, leading to high background signals and incompatibility with radioactive nuclides with short half-lives. As an alternative to Annexin V, small amphipathic molecules that accumulate in apoptotic cells have been developed9. A radiolabeled version of one such probe is currently being evaluated in a Phase IIa trial for PET-based imaging of ischemic stroke. However, the limited information regarding mode of action makes it difficult to determine the overall applicability of this probe to diverse types of cell death observed in other disease systems.
As direct mediators of the early stages of apoptosis, caspases are obvious targets for molecular imaging probes. Because they are proteases, substrate processing can be used as a readout of activity. Currently, several classes of luminescent10, fluorescently-quenched11 or radiolabeled12 substrate based probes are in development. Perhaps the biggest challenge for these probes is developing sequences that are specific for caspases. The substrate sequence most commonly used for these probes, Asp-Glu-Val-Asp (DEVD), while optimal for caspases-3/713, is also efficiently recognized by several other cysteine proteases including the cathepsins and legumain14,15. Cross-reactivity with cathepsins is especially problematic due to their high constitutive expression in multiple tissues and organs including the liver, kidney and spleen. As a result, substrate-based reagents containing this or similar selectivity sequences may prove problematic for apoptosis imaging in vivo.
In addition to substrates, several classes of small molecule inhibitors have been used to label caspase activity in vivo. These include reagents such as WC-II-89, a non-peptidic competitive inhibitor derived from isatin sulfonamide analogs16, M808, an irreversible inhibitor based on a caspase substrate17and the fluorescently labeled peptide fluoromethyl ketones (FLICA)18,19. While WC-II-89 has been shown to inhibit caspase activity in vitro, it remains difficult to determine its selectivity for caspases in vivo. Likewise, M808 has been shown to exhibit high, nonspecific labeling in an in vivo model of cycloheximide induced hepatocyte apoptosis17. The FLICA probes, on the other hand, have been used mainly for FACS-based studies of apoptosis and are now commercially available. However, these probes suffer from lack of selectivity20 and are effective inhibitors of various cathepsins21. We have also found that FMK-based probes of the caspases produce high levels of background labeling when used in simple in vitro apoptosis systems14.
We have previously described the development of irreversible inhibitors and active site probes of the caspases that exhibit both broad and narrow selectivity within this family of proteases22. While we found initial acyloxymethyl ketone (AOMK) probes designed based on our earlier work to be effective labels of caspases in vivo, these reagents showed significant cross-reactivity with cathepsins and legumain. In this study, we identify optimal sequences that show reduced legumain reactivity and a complete lack of reactivity towards the cathepsins. Importantly, when these optimized probes are labeled with near infrared fluorescent (NIRF) tags they allowed caspase labeling to be monitored in vivo using non-invasive imaging methods. Furthermore, addition of a cell permeable peptide sequence to the probe increased uptake into apoptotic cells resulting in enhanced overall signal in apoptotic cells and tissues. Overall, these studies demonstrate that ABPs that target caspases can be used to track the early stages of apoptosis and that probe signal can be monitored using methods that allow whole body, non-invasive imaging of apoptosis.
In our past studies, we designed a number of AOMK-based probes that showed efficient labeling of caspases in whole cell extracts22. For our first generation ABP, we converted the most potent and broad-spectrum peptide sequence (6-E-8-D; AB28) to a fluorescent probe by replacement of the P4 amino acid with a linker labeled with the NIRF fluorophore Cy5. We initially tested this probe, AB46-Cy5 in a syngeneic lymphoma model in which tumorigenesis is driven by conditional overexpression of the Myc oncogene23 (Sup. Fig. 1a,b). These initial studies indicated that the probe efficiently labeled caspase-3 and -7 but also labeled cathepin B and legumain (Sup. Fig. 1c,d), consistent with previous studies 14,15. In order to decrease cross-reactivity of AB46-Cy5 with cathepsin B we made use of our earlier finding that a proline residue in the P2 position of legumain probes abolished binding to cathepsin B15. Using this information we developed an ABP containing the EPD-AOMK sequence labeled with the Cy5 fluorophore (AB50-Cy5; Sup. Fig. 1e). This probe showed labeling of caspase-3 and legumain with virtually no detectible cathepsin B labeling. Indirect competition experiments produced similar results (Sup. Fig. 2). In order to reduce the potency of our caspase probes towards legumain we conducted a screen for P3 amino acids that directed selectivity away from legumain. We identified a series of sequences that enhanced potency towards caspase-3 and away from legumain (Sup. Fig 3). We synthesized a total of 11 inhibitors containing non-natural amino acids that directed selectivity away from legumain (Sup. Fig. 4). From this set of optimized inhibitors, AB53-Cy5, which contained a P3 biphenylalanine (Bip), showed the most selectivity towards caspases, with a greater than 10-fold reduction in legumain binding relative to AB46-Cy5 or AB50-Cy5 (Sup. Fig. 1). However, labeling of intact cells indicated that it had relatively poor cell permeability (Sup Fig. 5). We therefore chose to carry out our in vivo studies using AB50-Cy5.
In order to enhance the cell permeability of AB50-Cy5 we synthesized a version of the probe containing a Tat peptide. This peptide makes use of multiple positively charged amino acids to carry attached cargo across membranes and has previously been used to increase the cell uptake of caspase substrates11,12. The Tat probe, tAB50-Cy5, differs from AB50-Cy5 in that the Cy5 fluorochrome is moved to a lysine side chain and the Tat peptide is coupled through a cysteine residue to a maleimide group at the N-terminus of the probe (Fig 1a). We also generated control versions of AB50-Cy5 (AB50-Ctrl) and tAB50-Cy5 (tAB50-Ctrl) that contain an amide in place of the reactive AOMK warhead. As expected, the active probes efficiently labeled recombinant caspase-3, while control versions of the probes did not (Fig. 1b). In addition, we tested all four probes for their ability to label caspases in intact cells treated with an antibody to Fas. Cells were either activated by the Fas antibody and directly labeled with probes or pretreated with probes and then washed prior to activation of apoptosis (Fig. 1c). These results indicated that only the active probes AB50-Cy5 and tAB50-Cy5 labeled caspases and furthermore, only tAB50-Cy5 showed labeling of caspases after pretreatment and washout. However, we also found that addition of the tat peptide resulted in a significant increase in the labeling of legumain. This increase in labeling of a lysosome-resident protease is likely due to uptake of the probe via the endo-lysosomal route prior to release into the cytosol as has been suggested for similar carrier peptides24.
To demonstrate the utility of caspase ABPs for imaging apoptosis, we monitored apoptosis in CD4+/CD8+ thymocytes of mice treated with dexamethasone25. We chose this system because the kinetics of apoptosis have been explored in previous studies using a number of common apoptosis markers including TUNEL26 and Annexin V27,28. We monitored caspase activation in mice treated with dexamethasone (dex) for 6, 12, or 24 hours (n=3 for each time point). The AB50-Cy5 and tAB50-Cy5 probes were injected intravenously (IV) two hours prior to removal of thymi and imaging using the IVIS 200 imaging system (Fig 2a). Following imaging, we processed the tissues and analyzed labeling in total extracts using SDS-PAGE followed by visualization of labeled proteins using a flat-bed laser scanner (Fig 2a). This allowed us to biochemically characterize the target proteases labeled by in vivo application of the probe. We confirmed by immunoprecipitation that the probes labeled both caspase-3 and legumain (data not shown). We then quantified total caspase and legumain labeling and compared these values to total fluorescence signals observed in intact thymi (Fig. 2b). These data indicate that legumain activity is low and remains largely unchanged in the first 12 hours after dex treatment. Caspase-3 activity is observed at 6 hours post treatment and peaks at 12 hours and then sharply drops to background levels at 24 hours after injection. Importantly, the overall trend in levels of labeled caspases directly correlate with the overall signal observed for intact thymi using the IVIS system, suggesting that the fluorescence observed in whole organs can be used as a direct readout of total probe-labeled proteases. These data also agree with previous studies26 that showed a peak in TUNEL+ thymocyte staining 16 hours after dex treatment followed by a sharp decrease at 18 and 24 hours. These data suggest that caspases are likely to be activated at early time points (i.e. 6 h) and therefore may serve as effective markers of the early stages of apoptosis.
Comparison of the data sets for the AB50 and tAB50 probes suggested that while the overall trend in fluorescent signals in the intact thymi as well as the labeling patterns of caspases and legumain were the same for both, the overall signal intensity of labeled legumain and caspases as well as fluorescence emitted from intact thymi was increased for the Tat labeled probe (Fig. 2b). Quantification of fluorescent signals indicated that peak fluorescence was greater in the tAB50-Cy5 labeled thymi. Interestingly, we did not observe a similar increase in tAB50-Cy5 signal in the samples treated with vehicle suggesting increased uptake of the probe only into apoptotic cells. In further support of this hypothesis, we analyzed thymocytes from the 12 hour time point by flow cytometry (Sup. Fig. 6). This allowed us to monitor levels of Cy5 fluorescence in both apoptotic (i.e. CD4+/CD8+ thymocytes) as well as non-apoptotic cells (i.e. CD4+ or CD8+ thymocytes). These data confirmed that dex treatment specifically induced apoptosis in CD4+/CD8+ cells, as measured by a drop in this cell population over time. Furthermore, the tAB50-Cy5 probe accumulated only in dying cells, and probe positive cells were also positive for Annexin V (Sup. Fig. 6c).
To further characterize the properties of our probes in the dex model, we monitored uptake of active and control probes into intact thymi at 12 hours after treatment with dex (Fig. 3a). These data indicated that both AB50-Cy5 and tAB50-Cy5 showed a highly significant accumulation in dex treated thymi relative to vehicle treated tissues (p values 0.003 for AB50-Cy5, <0.0005 for tAB50-Cy5). Interestingly, both control probes showed an accumulation in dex treated samples, albeit at lower overall signal compared to the active probe. SDS-PAGE analysis of the total thymus extracts confirmed that the control probes failed to label either legumain or caspases and these signals therefore were the result of an increase in levels of free probes in dex treated samples (Fig. 3b). Thus, our ability to image apoptosis was enhanced by overall increased probe uptake into apoptotic cells. Finally, histology of thymus tissues showed that signals observed in tissue sections closely matched the signals observed for intact thymi (Fig. 3c). Interestingly, the signals for both the active and control tat probes showed strong nuclear staining of cells and only in dex treated tissues (Fig. 3c, insets). Therefore, we believe that there is significant uptake of tat-labeled probes when cells become apoptotic and this uptake results in nuclear retention of the probes.
In addition to the dex model, we wanted to evaluate our probes in a more relevant model of human disease. We therefore chose to monitor apoptosis in xenografted human tumor tissues that had been induced to undergo apoptosis by treatment with the monoclonal antibody Apomab. This reagent induces the extrinsic apoptosis pathway by binding to death receptor 5 and is currently in phase II clinical trials as a chemotherapy agent for lung cancer29,30. We felt this was an ideal model system because the antibody induces apoptosis in a way that is highly distinct from the dex-induced intrinsic apoptosis in CD4+/CD8+ thymocytes. As a starting point we determined the in vivo kinetics of caspase activation in tumor cells in response to Apomab. We treated mice with Apomab, waited 2, 5, 8, 11 and 17 h and then injected AB50-Cy5 and allowed it to circulate for an additional hour. We then imaged whole tumors ex vivo and analyzed caspase labeling by SDS-PAGE (Fig. 4a). Quantification of total tumor fluorescence, as well as caspase and legumain labeling intensity, indicated that total fluorescent signals in the tumors closely mirrored levels of labeled caspase-3/7 as measured by gel analysis (Fig. 4b).
Furthermore, since the overall levels of legumain labeling were constant throughout the time course, this cross reactivity did not hinder our ability to monitor dynamic changes in caspase activity. Overall these results indicated that optimal caspase activity was observed 12 hours after Apomab treatment.
Since each probe has different clearance rates, we needed to determine the optimal time for imaging after probe injection. We therefore treated mice with Apomab or vehicle for 12 hours and injected each of the active probes and non-invasively monitored tumor fluorescence over a range of time points (Sup Fig. 7). These data indicated that the Tat labeled probes show significantly brighter signals but had slow clearance from all tissues resulting in low signal to background at the early time points. Optimal contrast was observed at 5 hours after probe injection. AB50-Cy5, on the other hand, cleared rapidly and showed good signal to background even at the early time points (i.e. 50 minutes). Thus, we monitored apoptosis in tumors 50 minutes after injection of AB50-Cy5 and AB50-Ctrl and 5 hours after injection of tAB50-Cy5 and tAB50-Ctrl (Fig. 4c). These images showed specific probe labeling of apomab treated tumors with a high contrast for both tAB50-Cy5 and AB50-Cy5 (4.5-fold and 3.2-fold contrast, respectively). We also found that the control probes showed some degree of apoptosis-specific uptake into tumors. However, the AB50-Ctrl probe showed substantially weaker signal and overall, there was no significant difference in signals of apomab-treated tumors relative to vehicle treated tumors (p value = <0.27). Interestingly, tAB50-Ctrl showed nearly identical accumulation in apomab treated tumors as tAB50-Cy5 but also produced higher background fluorescence in vehicle treated tumors. Thus, although the control probes show some accumulation in apoptotic cells, only the active probes showed significant contrast between apoptotic and non-apoptotic tumors.
Finally, to confirm that the signal observed by non-invasive imaging was due to labeling of active caspases, we performed ex vivo imaging of intact tumors (Fig. 5a), followed by analysis of labeled proteins by SDS-PAGE (Fig. 5b). The ex vivo images closely matched the images obtained in live animals. In addition, gel analysis showed specific labeling of caspases only in Apomab treated samples labeled with AB50-Cy5 and tAB50-Cy5. We also confirmed specific labeling of apoptotic cells by histology of tumor tissues (Fig. 5c). In agreement with the results obtained for the dex model (Fig. 3c), we found that only tumor cells from Apomab treated animals showed probe staining. These data therefore demonstrate that caspase-specific probes can be used to non-invasively monitor apoptosis in tumors treated with chemotherapy agents.
The development of reagents that can monitor apoptosis in vivo has potentially great value for monitoring therapeutic efficacy of drugs as well as for diagnosing early stages of diseases involving this critical pathway. To address the need for novel tools, several strategies have been implemented with mixed success. As the first responders to apoptotic stimuli, caspases are an ideal target for imaging agents for use in both pre-clinical and clinical applications. Here we described the development of fluorescent activity based probes and their application in vivo. Starting from a first generation probe AB46-Cy5 that shows strong cross-reactivity with both cathepsins and legumain, we were able to improve selectivity of the probes towards caspases and show that they could be used to monitor apoptosis in vivo using both invasive and non-invasive methods. In addition, since the probes covalently modify target proteases, they allow direct biochemical analysis of the kinetics of apoptosis in vivo.
In addition to finding that the caspase-specific probes could be used to monitor the kinetics of apoptosis in multiple model systems, we also found that control versions of the probes that lacked the reactive warhead group showed some degree of specific uptake in apoptotic cells. We found that addition of the tat peptide caused slower clearance and greater uptake into cells. However, we found significantly higher uptake of the Tat control probe compared to the non-Tat control. Histology of tissues from probe treated mice showed that both active and control Tat probes accumulated in the nucleus as has been proposed for the Tat peptide in other studies31. While the control probes may still be capable of binding caspases as substrates, we believe that the use of tat peptides may be beneficial for apoptosis imaging applications. Our results suggest that the non-tat labeled probes provide the best direct readout of caspase activity, but that the Tat-labeled probes boost overall specific signal and may therefore be useful for carrying other contrast agents into apoptotic cells.
While we were able to develop a probe with decreased legumain cross-reactivity, the overall degree of selectivity for caspases relative to legumain was modest. We believe that it may be difficult to completely avoid cross–reactivity with legumain. However the issue of labeling of this off-target protease can potentially be reduced by pre-treatment with inhibitors that show exquisite specificity for legumain. For example, pre-blocking of legumain activity using aza-epoxide containing a P1 asparagine32 should allow subsequent exclusive labeling of caspases using the AB50 probe series. Alternatively, it may not be necessary to use probes with absolute caspase selectivity for imaging apoptosis in vivo. We found that overall levels of legumain in the dex model seem to increase at late time points in the apoptosis cascade. This increase may be due to an increase in the overall number of legumain-expressing immune cells that respond to increased numbers of dead or dying cells. Thus probes that label both legumain and caspases may prove to be optimal for monitoring both early (caspase) and late (legumain) stages of apoptosis. Furthermore, the fact that legumain activity remains low and constant during the early stages of caspase activation suggest that probes can still be used to monitor dynamic changes in caspase activity at early time points. Finally, if tags such as tat provide selective access to apoptotic cells, it may be desirable to have multiple targets that bind probes and lead to their long-term retention within the cell.
While a number of methods have been explored to globally monitor levels of apoptosis, most make use of reporters that cannot be used to directly determine which protease targets are responsible for the production of a fluorescent or radioactive signal. One of the main benefits of the ABPs is their ability to form permanent covalent bonds with target proteases. Thus, signals can be imaged using non-invasive methods and these signals can then be associated with specific target proteases by biochemical analysis of labeled tissues. In the two examples presented here, we demonstrate that overall labeling signals in whole tissues and in live animals correlate with the signal intensity of all labeled proteases in those tissues.
Overall the data presented here demonstrate that activity based probes can be used for direct, non-invasive in vivo imaging of the kinetics of apoptosis in multiple mouse models. For both the dexamethasone and apomab systems, caspase activity can be detected at early time points and peaks at 12 hours. This is in contrast to other markers of apoptosis such as Annexin V and TUNEL staining which serve as markers of the later stages of apoptosis. Thus caspases are potentially important markers for apoptosis in vivo and agents that can be used to monitor dynamic changes in their activity are likely to have great value for both pre-clinical and clinical applications.
All inhibitors and activity based probes were synthesized using solid phase synthesis methods previously reported for P1 Asp-AOMK compounds22,33. Cy5 fluorochrome (Invitrogen) was coupled using a previously described method34. Fmoc-aminohexanoic acid (3 equiv.) and maleimidopropionic acid (3 equiv.) were coupled in the same manner as amino acids. The Tat peptide (Arg-Lys-Lys-Arg-Arg-Orn-Arg-Arg-Arg-Cys, all d-amino acids, except cysteine) was custom synthesized by the Stanford PAN peptide synthesis facility. Tat was coupled via its C-terminal cysteine to the N-terminal maleimide in DMSO (100mM final concentration) and DIEA (9 equiv) while agitating in the dark. The coupling reaction was monitored every 30 minutes by LC-MS analysis and purified upon completion (typically after 2 hours). AB50-Ctrl and tAB50-Ctrl were synthesized on Rink resin. The purity and identity of all compounds were assessed by LC-MS analysis using an Agilent HPLC coupled to an API 150 mass spectrometer (Applied Biosystems/SCIEX) equipped with an electrospray interface.
All animal experiments were approved by the Stanford Administrative Panel on Laboratory Animal Care and strictly followed their specific guidelines. Female, 4–6 week old BALB/c mice were obtained from the Stanford University Department of Comparative Medicine and housed in the Research Animal Facility. Animals were injected in the peritoneal cavity with water-soluble dexamethasone (Sigma) dissolved (50mg/kg final dexamethasone concentration) in 100µL sterile PBS 24, 12, and 6 hours prior to euthanization. Two hours prior to euthanization, mice were injected via tail vein with fluorescent probes (50nmol) in 10% DMSO in sterile PBS (100µL final volume). Animals were anesthetized with isoflurane and humanely euthanized by cervical dislocation. Thymi were harvested and fluorescence was visualized using the IVIS 200 system with a Cy5.5 filter and Living Image software (Xenogen). Thymus lysates were made using a bead beater in buffer containing 1% Triton X-100, 0.1% SDS, 0.5% sodium deoxycholate in PBS, pH7.2, as described elsewhere34. Total protein lysate (100µg) was analyzed by SDS-PAGE using 15% polyacrylamide gels. Probe labeling was visualized by scanning gels on a GE Typhoon flat-bed laser scanner (excitation/emission 633/670 nm). For control probe experiments, mice were treated with dex for 11 hours. Probes were then injected and allowed to circulate for 50 minutes for non-tat probes and 5 hours for tat probes. One lobe of each thymus was then imaged and analyzed by gel, and the other was used for histological analysis.
Female, 6-week old nude mice were obtained from Charles River and housed in the Research Animal Facility. Human colorectal cancer COLO205 cells (3×106) were injected subcutaneously on the back of each mouse in 30µL of 0.5% BSA in PBS. Tumors were established in 8–10 days. Apomab (10mg/kg; Genentech Inc.) or vehicle (10mM histidine/0.8% sucrose/0.02% Tween-20, pH6) was administered intravenously in 100µL volume as reported previously29 for 2, 5, 8, 11, or 17 hours. Fluorescent probes (50nmol) were then injected intravenously in 10% DMSO/PBS in 100µL volume. After one hour, tumors were removed, imaged using the IVIS 200 system, and analyzed by gel as described for the thymi. For noninvasive imaging, mice were injected with probe or control (50nmol), anesthetized with isoflurane and imaged with the IVIS system over time. Optimal time for probe clearance was determined to be 50 minutes for non-tat probes and 5 hours for tat-containing probes. Tumors were then removed, imaged ex vivo, and cut in half. One half was used for gel analysis and the other was processed for histology.
Thymi were fixed on ice for 5 hours in 4% paraformaldehyde and PBS, transferred to a 30% sucrose solution and rocked overnight at 4°C followed by embedding in O.C.T. Compound (Tissue-Tek). Tumors were placed directly into O.C.T. without prior fixing. Frozen sections (10µm) were cut by the Histology Lab in the Department of Comparative Medicine at Stanford. Sections were revived in PBS and mounted with Vectashield Mounting Medium with DAPI (Vector Laboratories). Images (10× and 40×) were obtained using Zeiss Axiovert 200M microscope.
Recombinant caspase-3 (100nM active in caspase buffer [100mM Tris, 10mM DTT, 0.1% CHAPS, 10% sucrose, pH 7.4]) was incubated with the indicated probes at the concentrations listed for 30 min. Samples were analyzed as described above. For wash-out experiments, Jurkat cells (5×106) were pre-treated with 1µM AB50-Cy5 or tAB50-Cy5 or their respective controls for 30 or 120 minutes followed by washing of the cells with warm RPMI 3 times. After washout, the CH11 antibody to Fas was added and cells were incubated for an additional 3 hours. For non-pretreated samples, Jurkat cells (5×106) were treated with the antibody to Fas for 3 hours with probe incubation for the last 30 or 120 minutes without washout. Cells were lysed on ice in hypotonic lysis buffer and labeled proteins were analyzed as above.
The authors would like to thank Dr. Guy Salvesen from the Burnham Institute for Medical Research for kind gift of recombinant caspases and for creative input on the project. We thank Dr. Bonnie Sloane, Wayne State University for kind gift of cathepsin antibodies, Dr. Colin Watts of University of Dundee for the kind gift of legumain antibodies. We thank R. Weimer for critical discussion of the data and help with protocols for the use of Apomab. We thank A. Fan and D. Felsher for assistance with the MYC mouse model. We thank the Molecular Imaging Program at Stanford (MIPS) and the Stanford Small Animal Imaging Facility for assistance with non-invasive imaging studies. This work was funded by NIH grants U54 RR020843 and R01 EB005011 (to M.B.).