Introduction:

The most examined biomarker¹ in Dynamic-Contrast-Enhanced Magnetic Resonance Imaging (DCE-MRI) is the pharmacokinetic parameter K^trans, the contrast agent (CA) extravasation transfer constant. Perhaps the prevailing expectation is that K^trans measures microvascular wall CA “permeability.” By far the most common CAs are monomeric Gd(III) chelates. For these, the Renkin-Crone expression for CA extravasation blood flow-dependence is generally in the permeability limit: K^trans is essentially equal to the P_CA•S product; where P_CA is the capillary wall CA permeability coefficient and S is the total vascular surface area per voxel volume.² Thus, only one K^trans factor, P_CA, actually measures permeability. Another “pure” permeability measure is the unidirectional rate constant for CA extravasation, k_pe, which is given by K^trans/v_p or 4P_CA/d.³ In these latter expressions, v_p is the blood plasma volume fraction and d is the mean capillary diameter. Here, with implanted human U87 glioma in rat brain, we show that K^trans changes essentially reflect changes in only S, not CA permeability.

Methods

Seven male athymic nude rats were anesthetized, and human U87 cells stereotactically injected into the right caudate nucleus. Four to ten weeks later, tumor‑bearing animals were re-anesthetized, and coronal-equivalent DCE-MRI scans of the head were performed at 11.75 T.⁴ For each animal, a dose of only 50 μmol/kg of GdDTPA-BMA was injected via tail vein catheter.⁴ The three-slice, fast-gradient-echo DCE-MRI sequence parameters were: TR/TE, 25/1.4 ms, flip angle 20º, slice thickness 1.0 mm, rectangular FOV (4.48 x 2.24) cm², with a 1.6 s intersampling interval. The nominal voxel size is 1.0 x 0.35 x 0.35 mm³. The data were analyzed with the shutter-speed paradigm pharmacokinetic model assuming three ¹H₂O signals,^5,6 applied in an adaptive (i.e., recursive or successive approximation) manner.⁴

Results

Figure 1 displays coronal-equivalent tumor parametric maps for one rat. A histological slice from the same animal brain, comparable to the DCE image slice (1a) is also shown (1d): the darker brown color shows human mitochondrial staining. Of particular interest are the v_b (1c) and K^trans (1b) maps: v_b is the blood volume fraction [often called CBV]. Although these exhibit heterogeneity, the patterns are similar: elevated in the upper left and lower right quadrants. Much v_b map noise arises from compromised DCE tumor v_b precision when using an extravasating CA:⁴ the precision is high for normal brain.³ Thus, we averaged K^trans and v_b over the entire tumor evident in the image slice [e.g., <K^trans>_sl] for each of the seven animals. Assuming a microvascular hematocrit of 0.3, we calculated pixel v_p values as 0.7 v_b. The <K^trans>_sl value divided by <v_p>_sl gives the <k_pe>_sl value for each tumor. Figure 2 plots <k_pe>_sl vs. <K^trans>_sl, for all seven rats. Four-to-ten weeks after tumor seeding, the <K^trans>_sl value [abscissa] varies by almost eight-fold. However, the <k_pe>_sl value [ordinate] varies by less than two [much due to v_p imprecision], and is effectively constant at ~0.04 s^-1.

Discussion

Since normal brain K^trans ≈ 10^‑5 min^-1,⁷ normal capillary k_pe [often called k₁ in the tracer literature] is effectively zero on the Fig. 2 ordinate scale (red point). Thus, within the first four to ten weeks of growth, tumor capillaries with k_pe ≈ 0.04 s^-1 have been established. One possibility is they result from alterations of normal brain capillaries. If so, surely P_CA increases during the alteration. Recall k_pe is proportional to the P_CA/d ratio, and it seems unlikely that the mean capillary diameter d would decrease. More importantly, however, the tumor capillary k_pe is so constant over time and animals that it appears regulated. Therefore, the more likely possibility is that the tumor begins by growing new capillaries (angiogenesis) with k_pe ≈ 0.04 s^-1. As K^trans increases (say, with tumor size and/or age), it seems to do so by increasing capillary density (increasing S), not by increasing microvessel permeability (P_CA): the capillary pore size and density seem fixed. For essentially all DCE‑MRI studies of gliomas and other cancers, the K^trans responses to therapies (mostly antiangiogenic) are disappointingly qualitatively, and quantitatively, very similar.¹ The K^trans value is essentially always decreased by therapy. Our results suggest this most likely happens by pruning tumor capillary density. Though it is often hoped that K^trans changes reflect P_CA changes, this is likely not the case. The K^trans biomarker is an ordinary intensive property, while k_pe is a supra‑intensive property.³ Precise DCE tumor v_b can be mapped with an intravascular CA.³ Combining this with K^trans from a Gd(III) CA will allow k_pe mapping, which could show regional k_pe variations. Our results suggest these will be small.

Acknowledgements

Grant Support: NIH: UO1-CA154602; R44 CA180425.

References

1. O’Connor, Jackson, Parker, Roberts, Jayson, Nat Rev Clin Oncol 9:167-177 (2012). 2. Li, Springer, Jerosch-Herold, NMRB 22:148-157 (2009). 3. Rooney, Li, Sammi, Bourdette, Neuwelt, Springer, NMRB 28:6-7-623 (2015). 4. Li, Rooney, Varallyay, Gahramanov, Muldoon, Goodman, Tagge, Selzer, Pike, Neuwelt, Springer, JMR 206:190-199 (2010). 5. Li, Rooney, Springer, MRM 54:1351-1359 (2005). 6. Li, Priest, Woodward, Siddiqui, Beer, Garzotto, Rooney, Springer, JMR 218:77-85 (2012). 7. Njus, Li, Springer, Taylor, Greisel, Telang, Coyle, Rooney, PISMRM 16:3431 (2008).