Translational Molecular Imaging
By: Martin G. Pomper
Russell H. Morgan Department of Radiology and Radiological Science
Department of Rheumatology, Johns Hopkins Bayview Campus
March 6, 2009
The evolution of diagnostic imaging
Diagnostic imaging is evolving rapidly. Anatomic techniques such as CT and MR that still form
about 95 percent of clinical practice are being transferred into functional modalities. For
example, CT perfusion studies, MR perfusion and functional MR imaging and diffusion tensor
imaging. These are functional applications of a formerly anatomic technique.
Hybrid technologies that combine the anatomic techniques of CT and MR, with the functional
techniques of PET and SPECT, are a big area of current molecular imaging research. The goal is
to perform molecular imaging looking at cellular and molecular processes in vivo non-
invasively, with no probes. A nice picture of the physiology in a completely unperturbed
environment is one of the main benefits of doing imaging to study cellular and molecular
Molecular imaging: a multidisciplinary enterprise
At Johns Hopkins, there are two Centers:
• The in vivo cellular and molecular imaging center (ICMIC) led by Zaver Bhujwalla, an MR
biophysicist. There are experts in each of the different areas and they try to improve on what
they do with imaging, using different techniques to study cancer; and
• A small animal imaging resource program. This program is oriented toward nuclear medicine
as opposed to the ICMIC, which is more towards MRI. It's got three different nodes of
activity with a physics group that develops hardware and software for imaging small animals.
A chemistry group, about a dozen or so chemists, develops new probes for imaging. A
molecular biology group generates molecular genetic reporter systems for imaging, looking at
signal transduction cascades. Most molecular imaging research is pre-clinical.
Utility of molecular imaging for clinical medicine
What can molecular imaging do?
• Early detection of changes occurring in tissue
• Enables changes in individual patient management in real time (personalized medicine)
• Facilitates drug development
All the major drug companies have in-house molecular imaging programs (most pre-clinical,
some are actually clinical), conducting pharmacokinetic, pharmacodynamic studies with various
Molecular imaging modalities
Modality Agents Human Rat Primary uses Examples
FMT Fluorescent X X gene expression, tagging GFP
proteins superficial structures RFP
BLI Luciferin X gene expression, fLuc
therapeutic monitoring rLuc
Tc X X site-selectivity, protein 99m
Tc-annex in V
I labeling 123
C X X site-selectivity, gene 11
F expression, drug 124
I development 64
Spectroscopy Endogenous X X CNS, prostate, heart , NAA
metabolites breast Cr
Contrast agents Gd X X cell trafficking, poly-L-lysine
Mn enzymatic activation dendrimers
Contrast agents perfluorinated X drug-delivery, gene human albumin
microbubbles transfection (Optison)
The techniques in the table above are not exclusive of one another; they're complimentary. They
have different capacities for being translated from rodents to patients. They also have different
sensitivities, for instance:
• Ultrasound has very good temporal resolution.
• Nuclear techniques do not because of the accumulated radioactivity counts on the detector to
generate the image.
• Ultrasound and Nuclear techniques don’t have the kind of special resolution MRI has, which
could be 100 microns.
These all balance each other out, which is why multi-modality is important.
The ICMIC tends to focus on the radiopharmaceutical-based techniques because they represent
the best way to translate new probes from rodents to the clinic and they have high sensitivity for
detecting molecular phenomena. Micromolar concentrations at the very lowest concentrations
are necessary for detecting molecules.
PET is the premiere translational molecular
imaging technique right now. It works by a
positron-emitting nucleoid that's attached to
a molecule of interest homing to a certain
process. The nucleoid achieves stability and
decays by emitting a positron that collides
with an electron about a millimeter away
from the decay event to create annihilation
photons that are detected in coincidence
around the patient. Combine that with CT
for the PET CT (Figure). The scan doesn’t
detect the decay, rather the annihilation.
Because of this, resolution is limited to
about a millimeter.
Translational Molecular Imaging
Four mature projects that have a translational component:
1. Mechanism-based imaging of prostate cancer
2. New aspects of bioluminescence imaging
3. Imaging bacteriolytic cancer therapy and infection
4. Cancer imaging and therapy via viral lytic induction
1. Mechanism-based imaging of prostate cancer
Background (reasons to study/image prostate cancer):
• One man dies of hormone-refractory disease every 20 min. in US alone
• Clinical understaging in 25% of men who have undergone radical prostatectomy
• Prostate-specific antigen (PSA) in normal tissues/sampling error from biopsy
• To direct biopsy and/or therapy (conformal radiotherapy or brachytherapy)
• Difficult to obtain metastatic biopsy specimens – bone scanning lags tumor response and is
• Preoperative pelvic nodal staging
• Detection of recurrence in post-prostatectomy patients with rising PSA
SPECT Imaging with ProstaScint is the mechanism-based technique for imaging the tumor. The
process, a monoclonal antibody, was developed about ten years ago. Monoclonal antibodies do
not make very good imaging agents:
• They are big molecules: 55,000–100,000 molecular weight.
• They tend to circulate in the blood for a long time.
• A concurrent blood pool image needs to be subtracted out.
• The wait is five days.
• A radioisotope that doesn't have very good imaging characteristics is used.
However, the monoclonal antibody binds to a very valuable target––the prostate-specific
membrane antigen, which is a marker for prostate cancer, particularly androgen-independent
disease, and tumor neovasculature. If an imaging agent can be developed for this, it can be used
for a lot of different indications.
PSMA is a very interesting “conventional: target because there's an enzymatic active site on the
outside of the cell. Processing tends to bind to an internal epitope so the cells have to be dead for
the antibody to gain access. The most important aspect of molecular imaging research is to find
the right target. PSMA is considered a conventional target because it’s being treated as if it's just
a receptor that's hanging off the cell.
When developing an imaging agent the function of the subject and what it does is key. The
monoclonal antibody for PSMA basically cleaves an acetyl-aspartyl-glutamate to glutamate and
N-acetylaspartate. Why this happens in the prostate is not clear. It also happens in the brain and
the GI tract. However, this enzyme has a fairly limited expression, which also makes it a good
target. In other words, it's not all over the body, therefore an imaging agent will not have a lot of
• Performing a retro synthesis on NAAG to get back to compounds that are a higher affinity
and more stable than the natural substrate.
• Developing urea-based PSMA inhibitors and functionalizing them for imaging. They are all
very low molecular weight, and the way you know it’s done is by looking at the natural
substrate. Since then the crystal structures have been made apparent.
Cyril Barinka at the NCI Frederick has crystallized the protein. The three domains contribute to
the active site. It's a dimer when it's on the cell surface. (Usually this kind of work doesn't help
that much in developing new imaging agents, but in this case having the crystal structure was
Looking more carefully at the enzyme
(figure), numerous molecules that have been
made are now overlapping. This is a
computational docking study. Two different
pockets to the enzyme can be seen. There's
one pocket where everything overlaps and
that has to be a certain structure. There's
another pocket that's a little bit more open to
taking larger substrates, which is good
because that's where chelaters and other
things toxic moieties can be inserted for
[11C]DCMC: radiosynthesis: Before the three domains and enzyme pocket information was
know, in 2002, the natural substrate was made into a urea and inserted with a sulfhydryl group.
The sulfhydral group attacks the carbon-11 methyl iodine and creates a direct probe for PSMA.
This compound/probe has a 20-minute physical half-life, carbon-11, so it's really just for proof of
principle. It needs to be purified, sterilized, injected into the patient and imaged within an hour.
Using a small animal PET scanner, a coronal image is
produced (figure) showing a PSMA-producing tumor.
There is a 12-to-1 target-to-nontarget or tumor-to-
muscle ratio. The process then has about a 2-to-1
[18F]DCFBC: The fluoro-18 version was developed because it has a 2-hour physical half-life as
opposed to 20 minutes. This can be shipped around the eastern half of the United States and
other centers can use it. Unlike the antibody, these low molecular weight agents have a little bit
of live uptake. The Cancer Imaging Program has a program called the Decide Program
(http://imaging.cancer.gov/) where they pay for the toxicity. They're paying for the toxicity on
this and this agent is about to enter the clinic.
YC-I-37: Using the crystal structure and knowledge of the necessary linkers, other moieties can
be studied. By doing docking studies, other interactions with this iodine can be identified. What
residues that this is going to interact with resulting in very high affinity agents with 20-to-1
target-to-nontarget ratios. This is specific binding in the kidneys and very little background if
things wash out.
[125I]YC-VI-11: There are other agents that just bind to the tumor. A day or so with this 125I-label
compound gives very high target-to-nontarget ratios with essentially no background.
Single amino acid chelate (SAAC) concept
Sangeeta Ray and her colleagues developed the SAAC concept:
The concept can be used to link up to these PSMA ligands, these ureas. Technetium can be put
out there for imaging; rhenium for therapy. Some of these are actually inherently fluorescent like
the bisquinolines because they're highly conjugated.
Using molecular modeling the chain link and the hydrophilicity of this particular chain (figure
below) was optimized.
This is on the order of 50-to-1 target-to-nontarget ratios with this molecule. And it's also
convenient. The workhorse of nuclear medicine is not PET. The workhorse is technetium and
that's because technetium is generated using a generator; in fact, every nuclear medicine
department has technetium as part of their armamentarium. A technetium agent is both desirable
and convenient. It’s injected, and then the patient can come back a few hours later and get the
The translational part comes is that the company, Molecular Insight, that licensed the patent, they
went ahead and developed a slight variation of one of the iodinated compounds and gave it to
patients. They've done about half a dozen patients. There’s less uptake in the kidneys as
compared to the rodents.
PSMA = glutamate carboxypeptidase II (GCPII)
1. PSMA turns out to be the same as an enzyme called glutamate carboxypeptidase II. There's
disregulation of this enzyme in schizophrenia and if inhibited, the enzyme ameliorates some
of the symptoms in the PCP model of schizophrenia.
2. Using a radiolabeled version of one of the PSMA ligands in an in vitro autoradiogram,
binding specificity was proven by first blocking the receptor and then treating with the
radioactivity and seeing no uptake.
Based on these two studies, we can look at the brain using our prostate agents.
The figure above shows there are differences in certain brain regions between normals and
schizophrenics or between schizophrenics and bipolar disorder. This is a good target for
psychiatric disease. The next step is to try to get compounds that don't have carboxylate groups,
that don't have all these charges because they can get across the blood brain barrier.
2. Bioluminescence imaging
Bioluminescence imaging is for looking at unconventional targets such as signaling cascades.
The hedgehog signaling cascade is being worked on at Hopkins in about eight different labs.
This is a developmental pathway; if there's a mutation in one of the proteins (this is a sonic
hedgehog ligand) there is a kind of disregulation of the pathway and eventually overproduction
of the transcription factor known as gli. Overproduction of gli causes patients to develop
glioblastoma, medulloblastoma, and prostate cancer pancreatic cancer.
This is a big target for the drug companies to be able to modulate, to turn it off. An in vivo
model to test a drug is necessary. To do molecular genetic imaging to look a signal transduction
cascade, use a reporter gene. Instead of imaging the gene of interest, link its activity to the
reporter. The only difference here is is that the reporter is a thymidine kinase protein that can be
detected using an external imaging device, like a PET scanner.
Thymidine kinase is produced whenever gli is turned on. The thymidine kinase phosphorylates
radioactive nucleoside analogs, like FIAU, which get phosphorylated and trapped in the cells. If
gli becomes active, turn on three different proteins: firefly luciferase for bioluminescence
imaging, red fluorescent protein for cell sorting and thymidine kinase for radiopharmaceutical-
based imaging. Any cell that has this gene in it that is active in terms of hedgehog is going to
turn on these three proteins and then the imaging devices can detect those proteins.
For example, the mouse in the figure has a
U87 tumor that has that gene in it. If U87
produces gli, turn on the reporter; inject
luciferine into the animal and then the
luciferine reacts with the firefly luciferase
that's being produced by the reporter gene.
The mouse is basically glowing with
The next step is to come in and turn that
light off with various drugs to prove the
Treatment with a hedgehog antagonist, HhAntag-691 increases BLI light output
The HhAntag actually made the light brighter. When given to the animal, the tumors start to
light up brighter. HhAntag was inhibiting a pump that normally pumps luciferin out of cells.
When it does that, more luciferin can get into the cells and react with the firefly luciferase
creating stronger light output. The only pumps that gave the high light output was a compound
called fumitremorgin C, which is specific for the ABCG2 breast cancer resistant protein pump.
This really throws off using bioluminescence imaging for drug development unless one can
account for the presence of ABCG2 in cells. Other implications include:
• BLI is the most common method for imaging transgene expression in vivo
• D-luciferin is the most common substrate for BLI
• All imaging studies, in vitro or in vivo, that employ fLuc-mediated BLI must be interpreted
in the context of BCRP activity of the cells under study
• A high-throughput screening test for BCRP substrates and inhibitors is suggested
Screening the Johns Hopkins Clinical Initiative Library
In bioluminescence imaging there's not a lot of background so it makes for a very nice screening
test. Cells can be put in the 96-well plates and one can basically add all sorts of compounds to
them and see which ones cause the light to increase. If the light increases, they're inhibitors of
ABCG2. So it's a screening test for new ABCG2 inhibitors. There's no clinical ABCG2 inhibitor
that's ever been used in patients because they tend to be a little bit toxic.
Using Jin Liu's compound library many compounds were tested, and about 60 of them increased
that light output. They are all FDA-approved. Therefore, they can be given with standard cancer
chemotherapy to inhibit ABCG2 and get more of the chemo into the tumor.
3. Bacteriolytic therapy: potential for molecular imaging
Heterogeneous oxygenation of tumors
Tumors become somewhat hypoxic when they outgrow their blood supply. Bert Vogelstein is
adding anaerobic bacteria by injecting it intravenously as spores. They go all over the body, but
they only germinate in the hypoxic core of tumors (unless there is an abscess of something). The
intent is for the bacteria to consume the tumors from the inside out.
This is kind of a crazy idea, but it's not a new idea. What Vogelstein brought to the program was
a systemic analysis of all the different anaerobes and found that C. novyi (type A) got in there
and consumed tumors.
The only problem with this therapy, particularly for clinical translation, is the fact that the C.
novyi tends to be lethal. The lethal toxin can be deactivated, which produces C. novyi non-toxic.
This is the stage where imaging is attempted to prove that the bacteria is under control.
Tumors can be imaged with radiolabeled FIAU, once bacteria have homed to them. FIAU was
not developed as an antibacteria agent; it was developed as an antiviral agent to treat AIDS
patients with hepatitis. It was lethal to some of these patients in the early '90s and its use was
forbidden. However, it can be used for imaging again because of the tracer principle. If bacteria
can be killed by FIAU, they can be imaged by [125I]FIAU. In other words, the bacteria have the
same thymidine kinase and it'll turn that over and accumulate in the bacteria.
Bacterial strains imaged after i.m. injection
Organism Clinical significance
E. coli Adult and infantile diarrhea, urinary tract
infection, pneumonia, meningitis and abscess
E. faecalis 49532 Nosocomial infection including vancomycin-
resistant enterococci, urinary tract infection,
endocarditis, abscess and meningitis
S. pneumoniae 49619 Pneumonia, meningitis, sinusitis, osteomyelitis
S. aureus 29213 and 25293 Cellulitis, indwelling medical infection,
diabetic ulcer, postsurgical wounds,
osteomyelitis, endocarditis, meningitis,
mastitis, phlebitis, pneumonia, boils, furuncles
S. epidermis F362 Endocarditis, cellulitis, urinary tract infection
and indwelling medical device infection
Mechanism-based bacterial imaging: infected prosthetic joints
Once a new way to image bacteria was developed, the first indication would be to look at
infective prosthetic joints because orthopedic surgeons are often fraught with patients that have
painful prosthetic joints. They don't know if it's infected or if it's just loose and those are going to
have two very different therapeutic outcomes.
It's a very sensitive technique, but not very specific. The bacterial imaging technique just looks at
bacteria because living, breathing bacteria is necessary to metabolize at that FIAU. Six to ten
patients with staph aureus were imaged using PET CT with [124I]FIAU. We can go from
preclinical with 125I to clinical with 124I.
Translation: By starting out looking at bacteria as cancer therapy, we now have a new way to
look at infection. There's not a lot of good ways to look at infection. Right now, the state of the
art in nuclear medicine is to use a tag white blood cell study. There are problems with that
because of using the right materials when reinjecting.
4. Cancer imaging and therapy via viral lytic induction
Imaging human herpesvirus lytic infection
EBV-positive tumor cells could be targeted for destruction by inducing the switch from latent to
lytic infection with pharmacologic agents.
• To show that the EBV-TK can be used as a reporter gene for the [125/124I]FIAU reporter probe
• To image EBV lytic gene expression in vivo
This is where the personalized medicine comes into play:
• How can we predict who's going to respond to this kind of therapy?
• What happens when the virus goes from latent to lytic, are any genes turned on?
• Yes, the Epstein - Barr virus has a thymidine kinase that is turned on in going from latent to
lytic. It needs the TK to do that.
The goal was to show that the Epstein - Barr virus TK can be imaged like the viral and the
bacterial TKs by using the same probe (the FIAU).
Imaging bortezomib-induced lytic gene induction
The first step was to take cells that are constitutively expressing the EBV-TK and show they can
be imaged using radiolabeled FIAU. Importantly, a TK- or an EBV-associated tumor like a Rael
lymphoma injected with FIAU doesn’t show anything in the tumor because it doesn't bind to
mammalian TK. If the virus is woken up using bortezomib, TK is produced and it can be imaged
This is kind of a noninvasive way to do it, and it can be done repeatedly. That's where the
personalized medicine aspect comes in. Different species of lymphoma can be imaged using this
The [125I]FIAU is sequestered for a fairly long time in the tumor. It goes away from all the other
organs and remains in the tumor, much like a therapy––long contact time with the drug.
To go from 125I to 131I produces a therapeutic nuclide, because 131I gives off beta particles. What's
nice about beta particles is they're like a hand grenade. They're very indiscriminate about what
they kill, which gives a bystander effect. Not every cell in the tumor has to express TK and more
cells will be killed than just the ones expressing it with 131I (figure).
This is referred to as bortezomib-induced enzyme-targeted radiotherapy. An increased dose can
actually kill those tumors. Interestingly, tumors that are comprised of 10 percent TK producing
cells as opposed to 50 percent result in a similar cell kill, which is indicative of the bystander