Longitudinal and multimodal in vivo imaging oftumor hypoxia and its downstream molecular eventsSteffi Lehmanna, Daniel P. Stiehlb, Michael Honerc, Marco Dominiettoa, Ruth Keista, Ivana Kotevica, Kristin Wollenickb,Simon Ametameyc, Roland H. Wengerb, and Markus Rudina,d,1

aInstitute for Biomedical Engineering, University of Zurich and Swiss Federal Institute of Technology, 8093 Zurich, Switzerland; Institutes of bPhysiologyand dPharmacology and Toxicology, University of Zurich, 8057 Zurich, Switzerland; and cInstitute for Pharmaceutical Sciences, Swiss Federal Instituteof Technology, 8093 Zurich, Switzerland

Edited by Ewald R. Weibel, University of Bern, Bern, Switzerland, and approved July 2, 2009 (received for review February 24, 2009)

Tumor hypoxia and the hypoxia-inducible factors (HIFs) play acentral role in the development of cancer. To study the relationshipbetween tumor growth, tumor hypoxia, the stabilization of HIF-1�,and HIF transcriptional activity, we have established an in vivoimaging tool that allows longitudinal and noninvasive monitoringof these processes in a mouse C51 allograft tumor model. We usedpositron emission tomography (PET) with the hypoxia-sensitivetracer [18F]-fluoromisonidazole (FMISO) to measure tumor hypoxiaover 14 days. Stabilization of HIF-1� and HIF transcriptional activitywere assessed by bioluminescence imaging using the reporterconstructs HIF-1�-luciferase and hypoxia response element-lucif-erase, respectively, stably expressed in C51 cells. Interestingly, wedid not observe any major change in the level of tumor hypoxiathroughout the observation period whereas HIF-1� levels and HIFactivity showed drastic temporal variations. When comparing thereadouts as a function of time we found a good correlationbetween HIF-1� levels and HIF activity. In contrast, there was nosignificant correlation between the [18F]-FMISO PET and HIF read-outs. The tool developed in this work allows for the longitudinalstudy of tumor hypoxia and HIF-1� in cancer in an individual animaland will be of value when monitoring the efficacy of therapeuticalinterventions targeting the HIF pathway.

hypoxia-inducible factor � positron emission tomography �bioluminescence � reporter gene � tumor allograft model

As a result of their high proliferation rates and abnormalgrowth patterns, tumor cells outgrow their vascular supply

territories, causing intratumoral areas of hypoxia, a hallmark ofmost solid tumors (1). Tumor hypoxia is accompanied by thestabilization of hypoxia-inducible factors (HIFs), oxygen-regulated transcription factors that mediate the adaptation ofcells to decreased oxygen availability (2). Typically, accumula-tion of HIFs in solid tumors is associated with poorer patientprognosis, more aggressive tumor phenotypes, and an increasedmetastatic potential (3). Furthermore, HIF activity has beenimplied to decrease the effectiveness of chemotherapy andradiation therapy (4, 5). Inhibiting tumor progression by target-ing the HIF signaling pathway in combination with other treat-ments thus appears to offer attractive options in the develop-ment of cancer therapeutics (6).

HIFs are heterodimeric transcription factors consisting of a�-subunit and 1 of 3 �-subunits (HIF-1�, HIF-2�, and HIF-3�).Whereas the HIF-� subunit is not affected by changes in oxygenavailability, HIF-� subunits are subject to proteasomal degra-dation under normoxic conditions, but become stabilized duringhypoxia. Details regarding the mechanism of HIF-� oxygen-dependent regulation have been described (7).

In view of the important role of tumor hypoxia and HIFs incancer development, great efforts are being taken to furtherunderstand and elucidate their contribution to tumor growth.Noninvasive in vivo imaging techniques are of particular interestin this context because they allow longitudinal monitoring ofcellular and molecular processes in the same subject. Here, we

report on the development of a hypoxia imaging tool, whichcombines nuclear imaging strategies and optical imaging ofreporter genes to study the relationship between tumor hypoxiaand its molecular consequences in a mouse tumor allograftmodel. We stably transfected the murine colon cancer cell lineC51 (8) with either a HIF-1�-luciferase fusion construct(pcDNA3.1-mHIF-1�-luciferase) or a hypoxia response element(HRE)-driven luciferase reporter gene [pGL(P2P)95bp]. Afterevaluation of these reporter constructs in vitro, tumor allograftswere established by s.c. inoculation of nude mice with the stablytransfected C51 reporter cells. In longitudinal in vivo experi-ments we daily assessed (i) tumor hypoxia by quantitativelymeasuring the uptake of [18F]-f luoromisonidazole (FMISO),(ii) the levels of HIF-1� (HIF-1 stability), and (iii) HIF tran-scriptional activity by using bioluminescence imaging over up to14 days.

ResultsAssessment of Tumor Hypoxia by Measuring Uptake of [18F]-FMISO.Accumulation of [18F]-FMISO in the C51 allograft tumors wasanalyzed from days 6 to 14 after tumor inoculation in 2 groupsof 6 animals by using an interleaved observation scheme. Twogroups of mice had to be used for the longitudinal study becauseof experimental restrictions (number of anesthesia episodes peranimal). Uptake of [18F]-FMISO tended to increase with in-creasing tumor volumes, implying a higher degree of hypoxia, asillustrated by transversal positron emission tomography (PET)images of a representative animal recorded longitudinally ondays 7, 10, 12, and 14 (Fig. 1A). After inoculation, tumors grewrapidly, with fastest growth rates between days 11 and 13 (Fig.1B). In [18F]-FMISO experiments, tissue is considered hypoxicwhen the activity displays a tissue-to-background ratio �1.4.Taking muscle as reference tissue, we calculated the tumor-to-muscle retention ratio (TMRR) for all hypoxic tumor voxels, i.e.,all tumor voxels displaying a TMRR �1.4 (9). In both groups ofanimals, the TMRR of the hypoxic tumor fraction (hypoxicTMRR) showed no major change in signal intensity over theentire observation period. Only a slight increase was observedbetween days 9 and 14 (Fig. 1C). On average, at least 50% of alltumor voxels displayed TMRR values indicative of hypoxia,clearly showing that the C51 tumor allografts were highlyhypoxic irrespective of the absolute tumor mass.

In Vivo Monitoring of the Stabilization of HIF-1� Using Biolumines-cence Imaging. To gain further insight into the complex anddynamic events driving the stabilization of HIF in cancer we

Author contributions: S.L. and M.R. designed research; S.L., D.P.S., M.H., M.D., R.K., and I.K.performed research; D.P.S., K.W., S.A., and R.H.W. contributed new reagents/analytic tools;S.L., M.H., R.H.W., and M.R. analyzed data; and S.L., D.P.S., M.H., and M.R. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

1To whom correspondence should be addressed. E-mail: rudin@biomed.ee.ethz.ch.

This article contains supporting information online at www.pnas.org/cgi/content/full/0901194106/DCSupplemental.

14004–14009 � PNAS � August 18, 2009 � vol. 106 � no. 33 www.pnas.org�cgi�doi�10.1073�pnas.0901194106

Dow

nloa

ded

by g

uest

on

June

10,

202

0

generated a reporter construct to monitor HIF-1� stability bymeans of bioluminescence imaging. HIF-1� was C-terminallyfused to firefly luciferase via a short linker sequence (pcDNA3.1-mHIF1�-luciferase; Fig. 2A). To analyze the functionality andlocalization of the fusion protein mouse embryonic fibroblasts(MEFs) were transiently transfected with the reporter construct,followed by treatment with dimethyloxalylglycine (DMOG), achemical substance that mimicks hypoxia by inhibiting thedegradation of HIF-� subunits (10). Immunofluorescence stain-ing of these cells using an antibody against firefly luciferaserevealed that the fusion protein was localized in the cell nucleus

as expected (Fig. 2B). To examine the oxygen-dependent regu-lation of the fusion construct, MEF wild-type or HIF-1�-deficient cells (MEF-Hif1a�/� and MEF-Hif1a�/�, respectively)were cotransfected with both the HIF-1�-luciferase and a re-porter plasmid (pH3SVB) that drives the expression of �-galac-tosidase from a HIF-responsive promoter. Whereas �-galacto-sidase activity was low in the absence of DMOG for HIF-1�-luciferase-transfected MEF-Hif1a�/� and MEF-Hif1a�/�, thereporter activity increased in both cell lines after DMOGtreatment (Fig. 2C). In cells transfected only with the �-galac-tosidase reporter plasmid, an induction of reporter activity afterDMOG treatment could be observed exclusively in the MEF-Hif1a�/�, reflecting endogenous HIF-1� activity. Next, we stablyexpressed the reporter construct in C51 colon cancer cells. Singlecell clones were screened for the highest induction in responseto hypoxia and high luciferase counts. To study HIF-1� stabili-zation in vivo, a highly expressing clone was injected s.c. into thenecks of nude mice. C51 cells stably expressing firefly luciferasefrom a CMV promoter (pcDNA3.1-luciferase) were used as areference; these cells should yield an almost constant lightoutput per viable tumor cell and could therefore be used fornormalizing the HIF-1�-luciferase-induced bioluminescent sig-nals. Bioluminescence imaging after the administration of D-luciferin revealed an increase in photon output in animalscarrying HIF-1�-luciferase-expressing tumors as the tumorsgrew (Fig. 2D). Two groups of 5 animals were analyzed by usingan interleaved scheme starting on days 5 or 6. Tumor growth wassimilar in both the control luciferase and the HIF-1�-luciferasetumors (Fig. S1). Luciferase activity was estimated by normal-izing (total photon counts) measured in both HIF-1�-luciferaseand luciferase control tumors to the theoretical counts calcu-lated for luciferase control tumors of the same volume (forfurther details on the normalization procedure see Fig. S2 andSI Text). Normalized intensities are expressed relative to thevalues measured on day 5 for both HIF-1�-luciferase andluciferase tumors (Fig. 2E). The normalization procedure com-pensates to some extent for variations in signal intensity asso-ciated with changes in tumor volume, blood absorption, ornecrosis. Intensities in Fig. 2D represent total photon countsmeasured in the tumors and hence differ from the normalizedvalues shown in Fig. 2E. The signal detected in the luciferasecontrol tumors deviates from the value 1 mainly for reasons ofbiological variability: (i) clonal differences may lead to varyinglevels of luciferase expression and (ii) tumor shapes may differ,both factors influencing the signal intensity measured on thesurface.

A

C

B

6 8 10 12 140

5

10

15

20

group 1group 2

days after tumor inoculation

htworg ro

mu t dezilamron

day 7 day 12 day 14day 10

6 8 10 12 141.5

2.0

2.5

3.0

3.5

4.0

days after tumor inoculation

RR

MT cixopyh

mean group 1mean group 2

animal 1-12

Fig. 1. Assessment of tumor hypoxia by measuring the uptake of [18F]-FMISO. (A) Four consecutive PET images representing cross-sections of thesame tumor-bearing mouse on days 7, 10, 12, and 14 are shown. Tumor regionsare indicated by the black dotted line and show increased uptake of [18F]-FMISO. (B) Tumor volume as function of time relative to the values measuredon day 6; mean � SEM (n � 6). (C) Average tumor-to-muscle retention signal(hypoxic TMRR) for the hypoxic fraction in the tumor defined by a TMRR �1.4(10); mean � SEM (black), individual animals (gray).

A

B

C Dfirefly lucCMV HIF-1

firefly lucCMV

HIF-1 -luc

luciferase

luciferase DAPI overlay

0.0

0.1

0.2

0.3

Hif-1a+/+Hif-1a-/-HIF-1 -lucpH3SVB ++ ++

+

]mn024[ .

U.A

+ + + +++ +

controlDMOG [2mM]

Hif-1a+/+Hif-1a-/-

E140012001000

800600400200

*

ytisne tni dez ilamron

*

6 8 10 12 14 160123456789

10 HIF-1 -luc group 1HIF-1 -luc group 2luciferase group 1luciferase group 2

days after tumor inoculation

*

Fig. 2. In vivo monitoring of the stabilization of HIF-1� using bioluminescence imaging. (A) pcDNA3.1-mHIF-1�-luciferase (HIF-1�-luc) and pcDNA3.1-luciferase(luciferase) reporter constructs. (B) Immunofluorescence staining of firefly luciferase in MEF cells transiently transfected with pcDNA3.1-mHIF-1�-luciferase.Nuclei were stained with DAPI. (Magnification: � 300.) (C) Assessment of oxygen-dependent regulation and transcriptional activity of the HIF-1�-luciferase fusionconstruct. MEF-Hif1a�/� or MEF-Hif1a�/� were cotransfected with pcDNA3.1 mHIF-1�-luciferase, and the reporter plasmid pH3SVB, which drives the expressionof �-galactosidase from a HIF responsive promoter. �-Galactosidase activity was assessed. Data are representative of 2 independent experiments. (D)Bioluminescent images of a mouse carrying a HIF-1�-luciferase C51 tumor in the neck. Four consecutive images of the same animal are shown. Color bar indicatestotal photon counts. (E) Normalized bioluminescence photon counts (mean � SEM) relative to day 5 values for both the HIF1�-luciferase and the luciferase controltumors. Values indicated by * significantly (P � 0.05) differ from values measured in control groups.

Lehmann et al. PNAS � August 18, 2009 � vol. 106 � no. 33 � 14005

MED

ICA

LSC

IEN

CES

Dow

nloa

ded

by g

uest

on

June

10,

202

0

Measuring HIF Transcriptional Activity Using in Vivo BioluminescenceImaging. A reporter construct for in vivo monitoring of HIFtranscriptional activity was generated to investigate the relation-ship between HIF-1� stability and HIF transcriptional activity.A DNA element comprising the minimal HRE from the humanPHD2 promoter was used to drive expression of the fireflyluciferase gene [pGL(P2P)95bp; Fig. 3A]. Hypoxic induction ofluciferase activity was assessed in MEF-Hif1a�/� and MEF-Hif1a�/� cells transiently expressing the HRE reporter con-struct. Normalized photon counts were increased under hypoxicconditions only in cells expressing HIF-1�, confirming the keyrole of this factor in transcriptional activation of the reportergene (Fig. 3C). C51 cells were stably transfected with thereporter construct, and single cell clones were screened for highluciferase counts and high sensitivity to oxygen levels. Cellularextracts of a selected clone were analyzed for HIF-1� andluciferase expression by immunoblotting (Fig. 3B). In the sameclone, bioluminescence increased with decreasing oxygen levels,revealing the graduated induction of HRE-luciferase (Fig. 3D).Allograft tumors were established in nude mice with the selectedHRE-luciferase clone and C51 control cells stably expressingfirefly luciferase from an SV40 promoter (pGL3prom). Twogroups consisting of 5 animals each were used for the measure-ments between days 5 and 15. Tumor growth rates were com-parable in the HRE-luciferase and the luciferase control tumors(see Fig. S1). Normalized in vivo luciferase activity was calcu-lated as described above by dividing the actual photon countsmeasured for both the HRE-luciferase and the luciferase controltumors by the estimation value for the photon counts of aluciferase control tumor of the same volume. Normalized photoncounts in control C51 tumors did not significantly change duringthe observation period. However, tumors expressing the HREreporter construct displayed an up to 14-fold signal increase�day 10 when compared with the initial measurement on day 5(Fig. 3E). For a more detailed description of the normalizationprocedure please see SI Text.

In Vitro Evaluation of in Vivo Reporter Activity. For confirmation ofthe reporter activities measured in vivo, sections from tumorsisolated on days 7, 8, 9, 11 and 14 were analyzed by immuno-histochemistry. The sections were stained for the nitroimidazolederivative pimonidazole (11), HIF-1�, GLUT1 [a typical HIF-1�target gene (12)], and CD31 as a marker for endothelial cells, i.e.,angiogenesis (13) (Fig. 4A). Higher-magnification images ob-tained from tumor sections harvested on day 14 demonstrate thecellular localization of the antigens detected (Fig. 4B). Whereaspimonidazole adducts were observed in both the cytoplasm and

nucleus, HIF-1� expression was predominantly nuclear. Asexpected, GLUT1 and CD31 were localized to the plasmamembrane. These in vitro stainings supported the results of thein vivo measurements: HIF-1� and GLUT1 protein levels andthe hypoxia marker pimonidazole showed a decrease from days7 to 8 and then recovered to increase throughout the remainingobservation period. In contrast, there was only little CD31detected on day 7. CD31 expression started to increase from day8, concomitant with the onset of tumor growth. Interestingly, thestaining patterns of pimonidazole, HIF-1�, and GLUT1 wereoften comparable but not identical. In particular, large tumorsdisplayed significant discrepancies in the spatial correlation ofhypoxia markers: regions with high HIF-1� and GLUT1 expres-sion but hardly any pimonidazole staining were observed, im-plying that the HIF cascade in tumors may not be exclusivelyactivated by low oxygen tensions. Furthermore, there werewell-vascularized tumor areas that nevertheless showed intensivepimonidazole staining, indicating that the vascularization inthose regions may not (yet) be fully functional (days 11 and 14).This interpretation was confirmed by Hoechst perfusion exper-iments and immunofluorescence stainings: regions that accu-mulated pimonidazole and were also positive for CD31 tendedto be weakly perfused (Fig. 4C). However, we also found regionspositive for CD31, pimonidazole, and Hoechst, indicating that inthese areas cells were hypoxic because of increased metabolicactivity or there was plasma flow with only low intravascularoxygen content.

Comparison of in Vivo Tumor Hypoxia, HIF Stability, and HIF ActivityMeasurements. We then compared the readout for tumor hypoxiawith the stabilization of HIF-1� and HIF transcriptional activityin the C51 allograft tumor model over 14 days (Fig. 5). Tumorhypoxia as assessed by measuring the average TMRR in thehypoxic tumor fraction, showed only a slight increase in traceruptake between days 6 and 14. In contrast, the HIF-1� stabilityreporter signal (HIF-1�-luc) peaked around days 9 and 10 forboth groups, decreased until day 12, and then reached a plateau.For the HIF activity readout, maximum photon counts wereobserved around day 10 in both groups of animals, i.e., slightlydelayed when compared with the HIF-1� stability reportersignal; toward the end of the observation period (day 12 andlater) the activity readout significantly decreased similar to theHIF-1� signal (Fig. 5A). Quantitative correlations of the varioushypoxia-related signals over time (Fig. 5 B–D) revealed areasonable correlation between HIF-1� stability and HIF activ-ity signals with Spearman r � 0.6364 and P � 0.05: the higher theHIF-1� stability, the more HIF transcriptional activity was

HIF-1

luciferase

oxygen [%] 20 1.0

DMOG [mM] - 2

firefly lucHRE

A

B

C ED

unspec.

unspec. 0

1.0

2.0

3.0

4.0

5.0

6.0

7.0

oxygen [%] 20 4 12 0.2

**

*

0.000

0.025

0.050

0.075

0.100

0.125

oxygen [%] 20 0.2 20 0.2

Hif1a+/+ Hif1a-/-

*

6 8 10 12 14 160

5

10

15

20

luciferase group 1luciferase group 2

HRE-luc group 1HRE-luc group 2

days after tumor inoculationstnuoc notohp dezila

mron

**

*

*

*

*

**

*

01 x nieto rp g/stnuoc

4

ytisnetni dezilamron

Fig. 3. Measurement of HIF transcriptional activity reporter. (A) Reporter construct pGL(P2P)95bp (HRE-luciferase). (B) Immunoblot analysis of stablytransfected C51 cells cultured at different oxygen concentrations or treated with DMOG. Unspec. refers to an unspecific band on the Western blots used as loadingcontrol. (C) In vitro luciferase activity measurement. MEF-Hif1a�/� and MEF-Hif1a�/� cells were transiently transfected with the HRE-luciferase reporter constructand exposed to normoxic or hypoxic conditions. (D) Luciferase activity (photon counts) normalized to total amount of protein in HRE-luciferase C51 reporter cellsexposed to varying oxygen concentrations. (E) Normalized bioluminescence photon counts (mean � SEM) relative to day 5 values for both the HRE-luciferaseand the luciferase control tumors. * indicate significant differences (P � 0.05).

14006 � www.pnas.org�cgi�doi�10.1073�pnas.0901194106 Lehmann et al.

Dow

nloa

ded

by g

uest

on

June

10,

202

0

observed. However, only weak or no correlations were observedbetween the luciferase reporters readouts and the degree ofhypoxia as measured by [18F]-FMISO PET (Spearman r ��0.5357, P � 0.24 for HIF-1�-luciferase and Spearman r �0.2143, P � 0.62 for HRE-luciferase, respectively).

DiscussionBy combining nuclear and bioluminescence imaging approacheswe have generated a tool for longitudinal assessment of tumorhypoxia, HIF-1� stability, and HIF activity in a mouse allograftmodel. This multimodal imaging approach allows investigatingthe relationship between tumor hypoxia and HIF signaling,which up to now had been only poorly understood. It may alsobe used for monitoring therapeutic interventions targeting theHIF pathway in a semiquantitative fashion. Remarkably, thelevel of overall tumor hypoxia only slightly increased with tumorgrowth over 14 days. However, the HIF-related readouts indi-cated dramatic changes in HIF-1� levels and HIF activity in theearly phase of tumor development, before the onset of massivetumor growth. Toward the end of the observation period, when

tumors have reached a volume of �1 cm3, these readoutsdisplayed a drastic decrease in signal intensity. Decreased HIFactivity, mediated by a HIF-induced negative feedback mecha-nism, has also been observed in in vitro experiments whenexposing cells to chronic hypoxia conditions (14). To whichextent, if at all, the decrease in HIF stability and HIF activityobserved in our in vivo model is regulated by such a negativefeedback mechanism remains to be investigated. Even thoughdirect comparison of an allograft model and spontaneouslyarising tumors in cancer patients is difficult, variability in HIFactivity is likely to also occur in the clinical situation. BecauseHIFs play an important role in mediating resistance to chemo-therapy and radiation therapy, our results, in combination withthe study of HIF activity in other tumor models, may allow forthe future identification of a time window for most effectivetreatment (15).

Interestingly, a correlation (P � 0.05) was found onlybetween the HIF-1� stability and the HIF activity readout,whereas this was not the case when comparing HIF-1� stabilityor HIF activity to the [18F]-FMISO PET readout (P � 0.24 and0.62, respectively). This observation was supported by in vitroimmunohistochemical stainings for pimonidazole, HIF-1�,and the HIF target GLUT1 in tumor sections: even though itwas possible to identify regions with comparable distributionpatterns, the overlay between pimonidazole and HIF-1� orHIF target proteins was generally poor. This finding is in linewith earlier studies demonstrating that there is no significantcorrelation in the degree of hypoxia and expression of HIF andits target proteins in human or xenograft tumor sections(16–18). Notably, oncogenic signaling pathways such as PI3Kor MAPK signaling in response to activation by oncogenic Rashave been shown to activate the HIF pathway independently oftumor oxygenation (3). At least in part this may account for thediscrepancies observed between the hypoxia and HIF read-outs. Alternatively, transient changes in oxygen concentration,

day 7

day 11

day 14

day 9

day 8

A

B

max

min

max max

min

max

min

HIF-1 GLUT1 CD31 pimonidazole H&E

HIF-1 GLUT1 CD31 pimonidazole H&E

min

C i ii iii iv

Fig. 4. Immunohistochemical analyses of tumor sections. (A) Immunohisto-chemical stainings of tumor sections extracted on days 7, 8, 9, 11, and 14.Images of whole tissue sections are shown. (Magnification: � 0.8.) (B) Tumorregions of a section from a tumor isolated on day 14 are shown with highmagnification to confirm the subcellular localization of the detected antigens.(Scale bars: 20 �m for H&E, pimonidazole, HIF-1�, and GLUT1 stainings; 50 �mfor the CD31 image.) (C) Immunofluorescence stainings. (i) Pimonidazolestaining. (ii) CD31 staining. (iii) Hoechst 33342 (perfusion marker). (iv) Overlayof i–iii. (Scale bar: 100 �m.)

Fig. 5. Comparisons of in vivo tumor hypoxia, HIF-1� stability, and HIFactivity measurements. (A) Tumor hypoxia as assessed by [18F]-FMISO PET,HIF-1� stability, and HIF activity readouts as a function of time. HIF-1� stabilityand HIF activity reporter values were normalized to the values measured onday 5 (left y axis). Tumor hypoxia is given by the hypoxic TMRR (right y axis).For each readout, mean � SEM values of the 2 groups measured are displayed.(B–D) Quantitative correlations of the different hypoxia readouts. (B) Spear-man r � 0.6364, P � 0.05. (C) Spearman r � �0.5357, P � 0.24. (D) Spearmanr � 0.2143, P � 0.62. For each readout, mean � SEM values of the 2 groupsmeasured are displayed.

Lehmann et al. PNAS � August 18, 2009 � vol. 106 � no. 33 � 14007

MED

ICA

LSC

IEN

CES

Dow

nloa

ded

by g

uest

on

June

10,

202

0

which cause the accumulation of [18F]-FMISO, but may bedetected with the HIF readouts only if of sufficient duration,might account for the missing correlation observed in ourstudy. To investigate whether and to what extent C51 tumorswould undergo such acute, transient changes in tumor oxy-genation, we performed dynamic PET scans to assess theuptake of [18F]-FMISO over 4 h (Fig. S3 a and b). We did notobserve significant temporo-spatial f luctuations in [18F]-FMISO activity pattern during this time window, indicatingthat there were no major changes in tumor hypoxia over 4 h inthe C51 allograft model. These results were in agreement withexperiments involving sequential injection (1-h delay) of 2hypoxia markers, CCl-103F and pimonidazole (19), in C51tumor-bearing mice (Fig. S3c). Immunof luorescence analysisof tumor sections from these animals did not reveal anyprofound changes in the overall distribution of the 2 hypoxiamarkers. We concluded that transient changes in tumor hyp-oxia occurring at pO2 � 10 mm Hg are unlikely to account forthe lack in correlation between the hypoxia and HIF-relatedimaging readouts in our study. However, we cannot excludethat tumor oxygenation transiently changes at pO2 levels �10mm Hg. These f luctuations would not be detected withbioreductive marker molecules, but could nevertheless lead tothe activation of the HIF system (20).

Even though HIF-1� stability and HIF activity readouts wereclearly correlated, discrepancies were observed also betweenthese readouts, which is not surprising considering the probabletemporal delay between the activation of the 2 reporters. More-over, the HIF activity construct used is driven by the HREisolated from the human PHD2 promoter. Although it has beenshown that this enzyme is induced predominantly by HIF-1 andnot HIF-2 (21) and our in vitro results indicate a strongregulation of the reporter construct by HIF-1, we cannot rule outsome activation by HIF-2 or other unknown transcription fac-tors, although the use of a minimized regulatory element shouldhave alleviated such interplay.

In summary, we have developed a multimodal imagingstrategy for studying the molecular events induced by tumorhypoxia in a time-resolved manner. Our findings revealedsignificant discrepancies between [18F]-FMISO PET and theHIF-1� stability or HIF transcriptional activity readout, whichhas to be further investigated in view of the importance of thePET approach in clinical tumor diagnostics. The HIF signalingpathway has, besides other pathways, emerged as an attractivetarget in the development of anticancer drugs. The imagingtool proposed in this work might support these developmentsby enabling visualization of mechanistic aspects of drug action.

Materials and MethodsPlasmid Constructions. A short linker sequence was used to C-terminally fuseHIF-1� to firefly luciferase, generating pcDNA3.1 HIF-1�-luciferase. The fireflyluciferase reporter vector pGL(P2P)95bp (HRE-luciferase) is driven by a trun-cation of the previously described human PHD2 promoter (22) and wasconstructed by placing synthetic oligonucleotides encompassing 90 nt of thecore region of the HIF binding site into pGL3 basic (Promega). To generate thepcDNA3.1-luciferase control plasmid, HIF-1� was excised from the pcDNA3.1HIF-1�-luciferase construct and the vector was subsequently recircularized.The �-galactosidase reporter vector pH3SVB is a subversion of pH3SVL (23)where a lacZ ORF replaces the firefly luciferase gene. �-Galactosidase isexpressed under the control of a minimal SV40 promoter flanked by 3 HIF-responsive elements from the human transferrin gene (24).

Generation of Stably Transfected Cell Lines. Four stable cell lines were generatedby cotransfecting mouse colon carcinoma C51 cells with either pcDNA3.1-mHIF-1�-luciferase, pcDNA3.1-luciferase, pGL3prom (Promega), or the pGL(P2P)95bpreporter plasmid, the latter two in combination with neomycin resistance gene-

containing pcDNA3.1 in a ratio of 10:1. Stable transfectants were selected byadding G418 (400 �g/mL). Resistant clones were isolated by limited dilution andin the case of pcDNA3.1-mHIF-1�-luciferase and pGL(P2P)95bp they werescreened for (i) good oxygen-dependent regulation and (ii) high absolute lucif-erase photon counts by using luciferase assays. pcDNA3.1-luciferase andpGL3prom luciferase clones were screened only for high luciferase activity.

In Vivo Allograft Tumor Models. All animal protocols were approved by theCantonal Veterinary Office in Zurich (129/2007 XIMO�Y2). To establish allografttumors, we injected 1 � 106 reporter cells into the neck of 8-week-old BALB/cnude mice (Charles River) that were maintained under optimized hygienic con-ditions. The termination criteria were reached and animals had to be euthanizedwhen tumors showed a volume of 2 cm3. Caliper measurements allowed deter-mination of tumor length and width. From these parameters tumor volumeswere calculated by using the formula: tumor volume � (length � width2)/2 .

[18F]-FMISO PET Experiment. The radiosynthesis of [18F]-FMISO was carried outaccording to the 2-step procedure reported by Lim and Berridge (25). The totalsynthesis time was �120 min, and radiochemical purity was �99% as assayedby HPLC. Specific radioactivities obtained immediately after the synthesiswere always �100 GBq/�mol. PET experiments were performed on the 16-module variant of the quad-HIDAC tomograph (Oxford Positron Systems) withperformance characteristics as described (26). Animals were lightly restrainedand injected with 5–20 MBq of the radiotracer (100–120 �L per injection) viaa lateral tail vein. Animals were anesthetized with isoflurane (Abbott) in anair/oxygen mixture at 80 min after injection and monitored as described (27).PET data were acquired in list-mode from 90 to 120 min after injection andreconstructed in a single time frame with a voxel size of 1 mm and a matrix sizeof 120 � 120 � 200 mm. Regions of interest (ROIs) were manually defined byusing the dedicated software PMOD (PMOD Technologies). ROIs were drawnfor the whole tumor on all coronal planes containing tumor tissue yielding avolume of interest. Reference tissue ROIs were drawn on 5–10 subsequentcoronal planes containing muscle tissue at the contralateral forelimb. Thequantification of [18F]-FMISO uptake was based on the TMRR. This ratio wascalculated for all hypoxic tumor voxels that were determined in analogy to themethod described by Koh et al. (9): according to their definition a tumor voxelwith a TMRR �1.4 defines the presence of hypoxia. Additionally, the fractionalhypoxic volume of the tumors was computed that is represented by thepercentage of hypoxic voxels (with a TMRR �1.4) of all voxels within a tumorvolume of interest (VOI). For visual inspection and comparison of [18F]-FMISOtumor uptake PET images were normalized to the injected dose per g of bodyweight.

Bioluminescence Imaging. Mice were gas-anesthetized by using 3% isoflurane(MINRAD) and oxygen as a carrier gas. Each mouse was given an i.p. injectionof 100 �L of luciferin in PBS (15 mg/mL; Caliper Life Sciences). Ten minuteslater, the animals were placed in a light-tight chamber equipped with acharge-coupled device imaging camera (IVIS 100; Xenogen). Photons werecollected between 5 and 300 s depending on the reporter line that wasanalyzed: (i) HIF-1�-luciferase, 300 s, (ii) HRE-luciferase, 60 s, (iii) pcDNA3.1-luciferase, 120 s, and (iv) pGL3prom-luciferase, 10 s. Images were analyzedwith Living Image software (Xenogen) and IGOR image analysis software(Xenogen). Total photon counts were determined by drawing an ROI aroundthe peak of photon emission. The border of a ROI was formed by those pixelswhose signal intensity was 5% of the maximal signal in the ROI. To correct forthe loss of signal associated with bigger tumor volumes in the oxygen-regulated HIF-1�-luciferase and the HRE-luciferase tumors, we normalizedtotal counts from those tumors to pcDNA3.1-luciferase or pGL3prom-luciferase control tumor counts, respectively.

For more detailed information regarding standard methods used in thisstudy please refer to SI Text.

ACKNOWLEDGMENTS. We thank Prof. Wilhelm Krek and Dr. Ian Frew(Institute of Cell Biology, Swiss Federal Institute of Technology) for helpfuldiscussion and help with immunohistochemical analysis, Prof. BurkhardBecher (Institute of Neuroimmunology, University of Zurich) for supportwith the bioluminescence measurements, and Prof. Jean-Marc Fritschy(Institute of Pharmacology and Toxicology, University of Zurich) and mem-bers of his laboratory for introduction to the different light microscopes.This work was supported by the Swiss National Science Foundation, theNational Center for Research Resources Neural Plasticity and Repair, andthe National Center for Research Resources Computer-Aided and Image-Guided Medical Interventions.

1. Vaupel P (2004) Tumor microenvironmental physiology and its implications for radi-ation oncology. Semin Radiat Oncol 14:198–206.

2. Semenza GL (1999) Regulation of mammalian O2 homeostasis by hypoxia-induciblefactor 1. Annu Rev Cell Dev Biol 15:551–578.

14008 � www.pnas.org�cgi�doi�10.1073�pnas.0901194106 Lehmann et al.

Dow

nloa

ded

by g

uest

on

June

10,

202

0

3. Pouyssegur J, Dayan F, Mazure NM (2006) Hypoxia signaling in cancer and approachesto enforce tumor regression. Nature 441:437–443.

4. Unruh A, et al. (2003) The hypoxia-inducible factor-1 � is a negative factor for tumortherapy. Oncogene 22:3213–3220.

5. Wirthner R, et al. (2008) Impaired DNA double-strand break repair contributes tochemoresistance in HIF-1�-deficient mouse embryonic fibroblasts. Carcinogenesis29:2306–2316.

6. Semenza GL (2003) Targeting HIF-1 for cancer therapy. Nat Rev Cancer 3:721–732.7. Coleman ML, Ratcliffe PJ (2007) Oxygen sensing and hypoxia-induced responses. Essays

Biochem 43:1–15.8. Corbett TH, et al. (1975) Tumor induction relationships in development of transplant-

able cancers of the colon in mice for chemotherapy assays, with a note on carcinogenstructure. Cancer Res 35:2434–2439.

9. Koh WJ, et al. (1992) Imaging of hypoxia in human tumors with [F-18]fluoromisonida-zole. Int J Radiat Oncol Biol Phys 22:199–212.

10. Jaakkola P, et al. (2001) Targeting of HIF-� to the von Hippel-Lindau ubiquitylationcomplex by O2-regulated prolyl hydroxylation. Science 292:468–472.

11. Raleigh JA, et al. (1998) Hypoxia and vascular endothelial growth factor expression inhuman squamous cell carcinomas using pimonidazole as a hypoxia marker. Cancer Res58:3765–3768.

12. Brahimi-Horn MC, Chiche J, Pouyssegur J (2007) Hypoxia and cancer. J Mol Med85:1301–1307.

13. DeYoung BR, et al. (1995) CD31 immunoreactivity in mesenchymal neoplasms of theskin and subcutis: Report of 145 cases and review of putative immunohistologicmarkers of endothelial differentiation. J Cutan Pathol 22:215–222.

14. Stiehl DP, et al. (2006) Increased prolyl 4-hydroxylase domain proteins compensate fordecreased oxygen levels: Evidence for an autoregulatory oxygen-sensing system. J BiolChem 281:23482–23491.

15. Semenza GL (2002) HIF-1 and tumor progression: Pathophysiology and therapeutics.Trends Mol Med 8:62–67.

16. Mayer A, Hockel M, Wree A, Vaupel P (2005) Microregional expression of glucosetransporter-1 and oxygenation status: Lack of correlation in locally advanced cervicalcancers. Clin Cancer Res 11:2768–2773.

17. Mayer A, et al. (2004) Lack of correlation between expression of HIF-1� protein andoxygenation status in identical tissue areas of squamous cell carcinomas of the uterinecervix. Cancer Res 64:5876–5881.

18. Vordermark D, Brown JM (2003) Evaluation of hypoxia-inducible factor-1� (HIF-1�) asan intrinsic marker of tumor hypoxia in U87 MG human glioblastoma: In vitro andxenograft studies. Int J Radiat Oncol Biol Phys 56:1184–1193.

19. Ljungkvist ASE, et al. (2000) Changes in tumor hypoxia measured with a double hypoxicmarker technique. Int J Radiat Oncol Biol Phys 48:1529–1538.

20. Cardenas-Navia LI, et al. (2008) The pervasive presence of fluctuating oxygenation intumors. Cancer Res 68:5812–5819.

21. Aprelikova O, et al. (2004) Regulation of HIF prolyl hydroxylases by hypoxia-induciblefactors. J Cell Biochem 92:491–501.

22. Metzen E, et al. (2005) Regulation of the prolyl hydroxylase domain protein 2 (phd2/egln-1) gene: Identification of a functional hypoxia-responsive element. Biochem J387:711–717.

23. Wanner RM, et al. (2000) Epolones induce erythropoietin expression via hypoxia-inducible factor-1� activation. Blood 96:1558–1565.

24. Rolfs A, Kvietikova I, Gassmann M, Wenger RH (1997) Oxygen-regulated transferrinexpression is mediated by hypoxia-inducible factor-1. J Biol Chem 272:20055–20062.

25. Lim JL, Berridge MS (1993) An efficient radiosynthesis of [18F]fluoromisonidazole.Appl Radiat Isot 44:1085–1091.

26. Missimer J, et al. (2004) Performance evaluation of the 16-module quad-HIDAC smallanimal PET camera. Phys Med Biol 49:2069–2081.

27. Honer M, et al. (2004) Dynamic imaging of striatal D2 receptors in mice using quad-HIDAC PET. J Nucl Med 45:464–470.

Lehmann et al. PNAS � August 18, 2009 � vol. 106 � no. 33 � 14009

MED

ICA

LSC

IEN

CES

Dow

nloa

ded

by g

uest

on

June

10,

202

0