Genetically engineered fluorescent protein lights up the course of electrical signals in mouse hearts

Drugs, being molecules themselves, relieve and cure through molecular interactions. Yet hypertension, arrhythmia, and other medical disorders show up on the scales of organs and bodies. To gain a complete picture of how organs work—or fail to work—biologists want to see the coordinated function of molecules in an entire organ, preferably in vivo.

Michael Kotlikoff of Cornell University and his collaborators have devised a way to do just that in mouse hearts. ¹ Using genetic engineering, they induced lab mice to make a specially designed protein in heart muscle. The protein fluoresces green under blue light—but only when the local concentration of calcium ions, which trigger muscle contraction, is high enough. Thanks to the protein, Kotlikoff and his colleagues filmed, for the first time, the ebb and flow of Ca²⁺ ions in the beating heart of a live mammal.

Richard Gray of the University of Alabama in Birmingham, who does electrophysiological imaging of hearts ex vivo, points out the advantage of the genetic approach. “One doesn’t have to deal with the one-time shot of doing experiments, ripping the heart out, and injecting dye.”

There’s another advantage. Because gene expression can be induced in embryos, Kotlikoff and his collaborators could follow how the calcium-signaling system changes and grows as the heart itself develops.

Circular permutation

The fluorescent protein was designed by Kotlikoff’s collaborator, Junichi Nakai of the RIKEN Brain Institute in Wako, Japan. Nakai’s starting point was green fluorescence protein (GFP), which is found in the jellyfish Aequorea victoria. Since the late 1980s, biologists have used GFP to “report” the success of transplanting a gene. They append the GFP gene to the one they want to transplant. If the two genes are expressed, GFP will be made and fluoresce.

Biochemists prize GFP above other reporters because its fluorescence derives solely from its configuration of amino acids: 11 so-called beta sheets arranged in a cylinder. By contrast, other light-emitting proteins depend on either an extrinsic pigment molecule or the binding of an additional molecule.

In the late 1990s, Roger Tsien of the University of California, San Diego, wondered whether the structural foundation of GFP’s fluorescence was robust enough to tolerate mutations that might modify the fluorescence or allow it to be switched on and off.

Tsien picked points along GFP’s genetic sequence where he could cut the sequence in two. Then, using a technique called circular permutation, he rejoined the formerly free ends with a short sequence whose corresponding amino acids would hold the mutant GFP together.

Surprisingly, Tsien found 10 circular permutations whose sequence of amino acids would not only fold reliably into a protein but would also preserve the fluorescence. The fluorescence even survived the insertion of a second protein, calmodulin, into GFP’s structure. ²

In the presence of four Ca²⁺ ions, a calmodulin molecule binds to a variety of proteins, including a specific peptide sequence from the muscle protein myosin light chain kinase (MLCK). Here, thought Tsien, was a GFP mutant that could potentially act as a fluorescent probe of Ca²⁺ concentration. He went on to develop a calmodulin-based probe, but its signal was quite weak.

Calcium

Tsien was interested in calcium because of its role in signaling. Of the various small ions that pass in and out of cells, calcium has a uniquely high concentration gradient. Outside cells and within certain cellular compartments, its concentration is 10 000 times higher than it is inside cells.

A modest number of Ca²⁺ ions flowing into a cell can therefore boost concentration by a significant factor. Cells exploit calcium’s hair-trigger sensitivity to drive muscle contraction.

In 2000, Nakai extended Tsien’s GFP work by inserting not just calmodulin but also its binding target, the MLCK peptide. In the absence of calcium ions, calmodulin and the MLCK peptide separate, causing GFP’s cylindrical structure to open and the ability to fluorescence to disappear. Adding calcium restores GFP’s structure and fluorescence.

Unfortunately, the first attempts at making calcium-sensitive GFP suffered from low intensity and poor heat tolerance. The problem is that GFP’s fluorescence not only depends on the protein’s detailed structure but is also influenced by its environment. Predicting in advance what mutation mitigates the environment or enhances fluorescence is difficult.

Indeed, that’s why Nakai, like Tsien, resorted mostly to trial and error to find successful mutations. He’d replicate the protein in bacterial colonies using a deliberately error-prone technique and screen for brightness and temperature stability. Eventually, in 2005, he found the bright, temperature-stable mutant, called GCaMP2, shown in figure 1.

Gene expression

Designing the protein was the first step. The second step, getting heart muscle cells, or myocytes, to express the gene for GCaMP2 in the right amount was also tricky. In general, a cell will start making a protein when the appropriate promoter protein binds to the appropriate stretch of DNA. Creating a gene for GCaMP2 and its promoter is straightforward, as is incorporating the gene—or transgene, to be precise—into a single-cell mouse embryo. However, the result would be too much GCaMP2 and a sick mouse.

To regulate GCaMP2 production, Kotlikoff exploited a standard and ingenious technique called tet-off. A special sequence, tet0, is added to the GCaMP2 transgene upstream of where transcription starts. The insertion creates an additional, necessary condition for gene expression: Not only must the promoter bind to the DNA but so too must a protein called tTA, which attaches to the tet0 sequence.

Doxycycline, a potent antibiotic, also binds to tTA. Feeding mice doxycycline provides tTA with alternative binding sites, thereby forestalling tTA–tet0 binding and, with the right dose, regulating GCaMP2 production.

Kotlikoff and his Cornell team introduced the transgene into single-cell mouse embryos and reared mice whose myocytes expressed a safe amount of GCaMP2. Preparation for filming the mice involved first anesthetizing and immobilizing them, then cutting open their chests to expose the beating heart.

The signaling molecule worked so well that a commercial camcorder sufficed to watch the heart turning green with each contraction. At 128 frames per second, the temporal resolution was high enough to record the conduction of Ca²⁺ waves from cell to cell throughout the heart.

Figure 2 shows the result in the case of an adult mouse. The sequence begins at the top left as the left and right atria prepare to contract. Just after the middle of the top row, the atrial Ca²⁺ concentration and contraction peak. Then, by about the middle of the bottom row, the ventricular Ca²⁺ concentration and contraction peak.

Because the gene for GCaMP2 is expressed, not injected, it’s possible to track the development of the embryonic heart. Strong Ca²⁺ signals first appear 10 days after embryonic cells start to divide. At that point, the heart has only two chambers connected by a narrow canal.

Like any other pump, the embryonic heart needs time to fill up before discharging. Adult hearts introduce a delay at the atrioventricular (AV) node. The heart’s pacemaker, the sinoatrial node, lies in the right atrium and transmits electrochemical signals to the AV node, which lies between the right atrium and right ventricle. When an electrical signal reaches the AV node, high resistance slows the signal’s retransmission to the ventricles. The delay—about 0.1 s in the human heart—makes pumping possible.

But an embryonic heart starts pumping before its AV node develops. How does it achieve the necessary delay? Biologists had found a patch of specialized cells near the interchamber canal of embryonic hearts that looked as though it might serve as the delay’s source.

Kotlikoff and his collaborators filmed embryonic hearts at various stages. At 10.5 days after the onset of cell division, Ca²⁺ signals flow quickly over the single atrial chamber, slow by a factor of 10 at the interchamber canal, then pick up speed again over the single ventricular chamber. By 13.5 days, the AV node has formed and the Ca²⁺ signals no longer evince direct conduction between the chambers. Other data suggest the specialized cells undergo programmed cell death.

Arrhythmias arise in fetal and adult hearts when the heart’s signaling system doesn’t work properly. If those malfunctions arise early, Kotlikoff hopes his technique can find them.