The concept of a barrier between blood vessels and the brain and spinal cord had existed since the late 1800s, when immunologist Paul Ehrlich found that intravenously administered dyes failed to stain certain regions of the brain, whereas other body tissues were stained. Ehrlich thought the dyes did not have a staining affinity for the brain, but his student, Edwin Goldman, showed that the dyes could stain brain tissues but could not cross a barrier into the brain.

But until 1967, it was not clear whether the structural basis for the barrier was at the level of the endothelium, the astrocytes or glial cells in the brain, or the basement membrane. It took the high resolution of electron microscopy, the development of sensitive tracer methods, and a fortuitous lunch date between Thomas Reese and Morris Karnovsky to show that the endothelial cells in the brain vasculature, and more precisely the junctions between these cells, form the cellular correlate for the so-called blood–brain barrier (Reese and Karnovsky, 1967).

“The study was a result of having a place where faculty could get together and talk,” says Karnovsky. “I met Reese by chance at the faculty dining room at Harvard. He was working on the anatomy of the brain and I was developing better tracers. As we were walking back from our lunch, we decided that the problem that needed to be tackled next was the blood–brain barrier.”

Karnovsky had already extended the horseradish peroxidase (HRP) tracer method of Werner Straus (1958) to the light and electron microscope. HRP was relatively small, and it could be detected by allowing the enzyme to act on a suitable substrate to yield an electron-opaque reaction product. The product could then be localized by microscopy of tissues fixed at various times after HRP injection (Graham and Karnovsky, 1966).


In the brain (left), the peroxidase reaction product (top, black) cannot get past endothelial junctions (arrow), but in the heart (right) peroxidase flows down a cleft (C) between endothelial cells.


The first experiments were disillusioning: after intravenous injection of mice with HRP and fixing by perfusion, the authors saw nothing. They came close to giving up, even considering that Ehrlich may have been right that some dyes, or in this case HRP, didn't work in the brain because of a lack of affinity for the tissue. But then Reese found a vesicle in the endothelium with HRP reaction product in it. That led them to do a slice immersion experiment where they cut out a chunk of tissue and dropped it in fixative. With this experiment they could, finally, see the HRP in the lumen of the vessels. This convinced them that the experiment was working but HRP could not cross the endothelium, and, incidentally, was being washed out of the lumen by the perfusion fixation.

They could see that the reaction product was blocked by tight junctions (Reese and Karnovsky, 1967) and concluded that the blood–brain barrier existed at the level of the vascular endothelium. In a paper Karnovsky published the same year in the JCB, he determined what made the brain special. Peroxidase was seen passing through the vascular endothelium of heart and skeletal muscle, apparently through or around the looser cell junctions in these tissues (Karnovsky, 1967), with a possible contribution from vesicles (Palade, 1953).

“Since passage of peroxidase through the vascular endothelium could occur in other tissues, we proposed that the blood–brain barrier was due to the fact that the cell junctions in the vascular endothelium in the brain were tight,” says Karnovsky. Indeed, unlike cell junctions found in other endothelia, the cell junctions of endothelial cells in the brain appeared to be extensive and were surmised to form an unbroken belt between cells.

A second characteristic feature of the endothelium of cerebral vessels observed by Reese and Karnovsky was the low frequency of vesicles associated with the transport of materials across endothelia, but the authors did not think this played a major role in the blood–brain barrier. “Even those vesicles that were there rarely seemed to fill with peroxidase, and no peroxidase seemed to penetrate beyond the luminal surface of the endothelium, so our feeling was that the junctions were the main barrier,” explains Karnovsky. LB

Graham, R.C., and M.J. Karnovsky.
J. Histochem. Cytochem.

Karnovsky, M.J.
J. Cell Biol.

Palade, G.E.
J. Appl. Phys.

Reese, T.S., and M.J. Karnovsky.
J. Cell Biol.

Straus, W.
J. Biophys. Biochem. Cytol.