Quantitation of Pulmonary Transgene Expression with PET Imaging

MATERIALS AND METHODS

The present study was approved by the Washington University School of Medicine’s Animal Studies Committee.

RESULTS

The relationship between adenoviral dose and several methods of quantifying reporter gene expression are presented in Figure 3. In vitro assays of mutant Herpes simplex virus-1 thymidine kinase (mHSV1-TK) activity and eGFP expression (the 2 products of the fusion reporter gene) were both linearly correlated with viral dose (P < 0.01; Figs. 3A and 3B). Likewise, there was an excellent correlation of mHSV1-TK activity versus eGFP expression levels by ELISA (R² = 0.84; data not shown). In contrast, although the pulmonary uptake of ¹⁸F-FHBG was greater at all doses of viral vector than that in rats administered null vector, there was also no correlation between uptake and adenoviral dose (Fig. 3C), nor was there any significant correlation of ¹⁸F-FHBG uptake versus mHSV1-TK activity (R² = 0.29).

Figure 4 combines data from the current study with similar data from a previous study that reported a linear correlation between ¹⁸F-FHBG uptake and mHSV1-TK activity (18). ¹⁸F-FHBG uptake increases rapidly at low levels of reporter gene expression (assayed as thymidine kinase activity) but then plateaus and essentially remains steady as mHSV1-tk expression increases.

The accuracy of pulmonary ¹⁸F-FHBG tissue radioactivity in lung ROIs on the PET images was assessed with correlations to γ-counter radioactivity measurements obtained on ex vivo lung samples (R² = 0.98; data not shown). The slope of the relationship was 0.28, comparable to that previously reported in a similar comparison (18). A slope less than unity likely represents the lack of correction for lung density on the imaging measurements, given that attenuation corrections have not yet been incorporated into microPET image reconstruction algorithms (the PET data are expressed as per milliliter lung, whereas the γ-counter data are expressed as per gram lung; normal lung density is 0.3–0.4 g/mL lung in rats (27)).

The clearance of ¹⁸F-FHBG from blood, based on direct blood sampling and expressed as the mean ± SD for all animals in this study, is shown in Figure 5. After a peak in blood radioactivity after tracer injection, a biphasic decrease was observed with a complex first phase followed by a monoexponential slower phase beginning 30 min after tracer injection (mean half-life ± SD, 31 ± 4 min). At the end of the experiment, blood radioactivity levels were nearly 2 orders of magnitude lower than the measured peak value and showed only minimal variability among experimental animals (¹⁸F-FHBG 0.04 ± 0.01 %ID/g blood [mean ± SD]).

Examples of fitted lung time–activity curves using the 3-compartment model (Fig. 2 and Eq. 2) are shown in Figure 6 for 2 representative animals. Figures 6A and 6B were obtained from an animal infected with Ad-CMV-mNLS-HSV1sr39tk-egfp and in a control animal infected with the null vector, respectively. Plotting of residuals over time (Figs. 6C and 6D), used as an estimate of goodness of fit (28), showed a random distribution of residuals about 0 in all animals of the study.

Patlak plots also showed excellent linearity by linear regression analysis for all animals (Fig. 7). The mean ± SD for the coefficient of determination, R², was 0.999 ± 0.001 for Ad-CMV-mNLS-HSV1sr39tk-egfp–infected animals and 0.58 ± 0.09 for the 2 control animals.

Overall, estimates of the influx constant, K_i, by compartmental modeling and by Patlak analysis were strongly and linearly correlated (R² = 1, P ≤ 0.0001; data not shown). A similar linear correlation was observed between the influx constant obtained by compartmental modeling and ¹⁸F-FHBG %ID (R² = 0.94, P ≤ 0.0001).

No relationship between either the net uptake of tracer (quantified as the slope of the Patlak relationship) or k₃ (the rate constant presumed to represent “trapping” of tracer within tissues expressing the tk reporter gene) could be identified with the assay of TK activity (as an in vitro index of reporter gene expression (Fig. 8)). This lack of correlation cannot be attributed to poor data quality given the results presented in Figures 5–7. Note once again, however, that values obtained in all rats infected with the adenovector carrying the reporter gene were consistently higher than values obtained in the rats infected with the null vector (Fig. 8; Table 1).

In rats treated with ANTU, levels of reporter gene expression (assessed with in vitro assays) were not significantly different than those in the normal rats (1.38% ± 0.95% vs. 1.29% ± 0.39% ³H-penciclovir phosphorylation per min/μg protein). However, in contrast to normal rats (Fig. 4), ¹⁸F-FHBG uptake was linearly correlated with in vitro assays of TK activity (Fig. 9A; R² = 0.56, P < 0.05). Furthermore, total lung uptake (expressed as %ID in the whole lung) was positively correlated with normalized lung weight (as a surrogate for severity of edema due to increased permeability) in the ANTU group (Fig. 9B) but not in the normal rats (Fig. 9C). The level of mHSV1-tk gene expression, on the other hand, achieved in the ANTU group (assessed with in vitro assays) did not correlate with normalized lung weight (R² = 0.25, P > 0.05; data not shown), demonstrating that these were likely independent factors determining the uptake of ¹⁸F-FHBG in the lung.

DISCUSSION

The main finding of this study is that lung tissue accumulation of the PET reporter probe ¹⁸F-FHBG is not linearly related to the tissue expression of the mutant PET reporter gene HSV1-tk over the full range of transgene expression obtained in lung tissue. In addition, our results appear to indicate that it is the pulmonary endothelium that may be acting as the potentially critical barrier, limiting access of the tracer to target cells expressing the viral thymidine kinase.

These results appear to contradict the results of several other studies that sought to validate image quantitation of gene expression against in vitro assays of HSV1-TK activity. For instance, highly linear correlations have been reported between in vitro assays of gene expression and the tissue accumulation of both pyrimidine (such as 2′-fluoro-2′-deoxy-1-β-d-arabinofuranosyl-5-iodouracil [FIAU]) and acycloguanosine derivatives (such as the ¹⁸F-FHBG used in the current study) (2,29).

More recently, several studies have noted a possible plateauing of tracer uptake as tissue measurements of HSV1-TK activity (the in vitro measure of gene expression) increased (18,19,30). However, the low number of experimental points and the relatively limited range of transgene expression in these studies precluded any definite conclusions. For instance, using a closely related PET reporter gene and the identical PET reporter probe as used in the current study, the upper limit of TK activity was 0.5% ³H-penciclovir phosphorylation per min/μg protein in 2 previous reports (19,31) (compared with 6-fold higher levels in this study; Figs. 4 and 9).

Our study addresses these issues is several ways. First, we combined the results of the present study with those of a previous study by our group using identical experimental protocols (Fig. 4) (18). Second, as just noted, we were able to achieve a wide range of tissue gene expression, measured as TK activity in vitro. Third, the scattergram shown in Figure 4 is the largest (n = 28) single comparison of tissue accumulation of a PET reporter probe versus an in vitro assay of gene expression. Note that in Figure 4, the plateau effect only occurs above TK activity levels of 0.5% ³H-penciclovir phosphorylation per min/g protein.

Several factors might account for this apparent plateau effect. Competition for phosphorylation between ¹⁸F-FHBG and cellular thymidine is one possibility, as recently noted by Min et al. (14). However, for this to have been a factor here, thymidine concentrations would had to have increased as a function of increasing expression of the transgene; there is no reason to believe that this was the case.

Another possible cause is that tracer availability to the viral transgene was restricted in some way. For instance, ¹⁸F-FHBG was shown recently to accumulate much more slowly in mammalian cells expressing the wild-type HSV1-TK in vitro than other radiolabeled pyrimidine nucleoside derivatives (13), and it was hypothesized that ¹⁸F-FHBG uptake might be limited by cell membrane transport (13). However, the mechanisms contributing to the final levels of radiolabeled probe uptake are complex. For example, pyrimidine nucleoside derivatives (i.e., FIAU) show higher cross-reactivity with endogenous TK than acycloguanosine derivatives (such as FHBG), and standard cell assays to determine levels of trapped radiolabeled nucleotides will not distinguish between product arising from the 2 different enzymes. Furthermore, pyrimidine and acycloguanosine derivatives have different affinities for known nucleoside transporters. Thus, whole cell assays cannot differentiate the effects of transport from enzyme efficiency.

To isolate different factors potentially involved in limiting pulmonary ¹⁸F-FHBG uptake despite evidence for increasing levels of gene expression, we analyzed the tissue and blood time–activity data with a standard 3-compartmental model that assumes both a transport step into target cells and a trapping (phosphorylation) step (Fig. 2). However, neither the net uptake of tracer (K_i) nor trapping per se (estimated by the rate constant k₃) correlated with the in vitro assay of gene expression (Fig. 8).

This absence of a significant correlation between k₃ and an independent measurement of the underlying physiologic process it allegedly represents calls into question the validity of the standard 3-compartment model assumed to describe pulmonary ¹⁸F-FHBG kinetics. We thus hypothesized that the model identified in Figure 2 was too simple and that, though transport into target cells per se might not be rate limiting, transport into the immediate extracellular environment of these tissues might be. For instance, the viral vector is administered intratracheally, resulting primarily in alveolar epithelial cell infection and gene transfer. However, the PET reporter probe is administered intravenously. It must therefore cross pulmonary vascular endothelial and interstitial compartments before it can gain access to alveolar epithelial cells. Trapping of tracer that makes it to this point might indeed be a function of transgene expression (i.e., “k₃”), but overall uptake might be limited by the barrier posed by the endothelium–interstitium.

To test this possibility, we used a low dose of the thiourea ANTU to increase the permeability of the pulmonary endothelial barrier. We used a dose that has been reported to increase vascular permeability with minimal development of pulmonary edema (which itself might interfere with tracer access to infected cells) (32–34). Although such an intervention is hardly precise, we nevertheless observed a linear relationship in ANTU-treated rats between ¹⁸F-FHBG uptake and the in vitro assay of mHSV1-TK activity (Fig. 9A vs. Fig. 4). Furthermore, we also observed a linear correlation between tracer uptake and lung weight (used as a surrogate for severity of change in endothelial permeability) in the ANTU-treated rats but not in the normal rats (Fig. 9B vs. Fig. 9C).

As it turns out, the fact that K₁ (the transport step) can limit net uptake of tracer (K_i), despite increasing levels of an enzyme that is designed to trap the tracer, is entirely consistent with the standard 3-compartment model. To illustrate this point, Figure 10 shows a simple simulation in which K_i was plotted against k₃ for different values of K₁ (where the ratio of K₁/k₂ was kept constant). Note that at low levels of K₁, net uptake of tracer plateaus despite increasing levels of k₃. K₁ must be increased by 50-fold before K_i and k₃ are linearly correlated over the full range of k₃ values in this illustration.

The change in the relationship between pulmonary uptake of ¹⁸F-FHBG and in vitro measures of gene expression after vascular permeability was increased by the administration of ANTU is consistent with an increase in K₁. The result is a linearization of the relationship (Fig. 9). Accordingly, we speculate that the kinetics of ¹⁸F-FHBG uptake in the lungs are better exemplified by a 4-compartmental model (Fig. 11), in which the second compartment represents the endothelial–interstitial barrier between tracer in the vascular space and tracer in cells expressing the viral kinase.

Since the pulmonary endothelium–interstitium appears to pose a barrier for intravenously administered ¹⁸F-FHBG uptake by pulmonary epithelial cells expressing a reporter imaging transgene, its possible that inhalational delivery of the radiotracer might improve the correlation between imaging signal and in vitro assays of reporter transgene expression. However, though this strategy might work well in normal lungs, it would be difficult to ensure that all portions of the lungs were equally exposed to radiotracer (a necessary condition if quantitation is to be meaningful) during disease states in which regional ventilation would be expected to vary considerably. Of course, the kinetics of ¹⁸F-FHBG uptake by the lungs after intravenous administration of the radiotracer might be quite different if endothelial cells rather than epithelial cells were transduced by the gene vector.

There are potentially multiple applications for reporter transgene expression imaging in the lungs, but they will depend on appropriate modifications to the vector expression cassette. For instance, it should now be possible to evaluate the transcriptional expression of an endogenous gene within the lungs of transgenic animals (35,36) by using the corresponding promoter/enhancer of the endogenous gene to drive expression of the reporter gene. Alternatively, by linking a therapeutic gene to a PET reporter gene, one should also be able to infer the level of expression of the therapeutic gene from the expression level of the reporter. Such a strategy could be used to monitor gene therapy for lung diseases such as cystic fibrosis. However, a predictable relationship between the imaging signal obtained with PET and the physiologic effect of any therapeutic gene has yet to be demonstrated.

Although the results of this study present a challenge for applying the strategy of monitoring reporter transgene expression with imaging, it would be inappropriate to assume that this result eliminates the potential value of this PET reporter gene/PET reporter probe combination specifically or the value of the concept underlying the use of PET imaging for “gene expression imaging” more generally. Alternative combinations of imaging reporter genes or reporter probes, possibly applied in different tissues, will likely yield different results with respect to in vitro assays of gene expression.

Furthermore, as we have already shown, measures of ¹⁸F-FHBG uptake are extraordinarily sensitive for detecting mHSV1-tk expression (18,19). Accordingly, this (and other similar systems) can be used to study both the onset and the duration of gene expression noninvasively, as recently demonstrated (18,19). This combination of mHSV1-tk and ¹⁸F-FHBG has also been used effectively to study the intrapulmonary distribution of vector delivery and to compare different vector vehicles (37).

CONCLUSION

In conclusion, uptake of the reporter probe ¹⁸F-FHBG appears to be a function of both transport into tissues expressing the reporter transgene mHSV1-tk as well as the level of transgene expression itself. Whether this specific combination of reporter gene and reporter probe can be used to monitor a “therapeutic” transgene depends on the level of expression that is required to achieve a physiologic effect. Until such studies are performed, this—as well as several other potential reporter combinations—should continue to be evaluated.