COVID-19 antibody tests are raising as many questions as they answer

In January, we didn’t know much about the novel coronavirus that would eventually be called SARS-CoV-2. We had just determined its genetic sequence. We were still learning about the vast array of symptoms it could cause. And we didn’t yet know how widespread, and how deadly, the disease COVID-19 would become.

That same January, after several days at a large international biotech conference in San Francisco, shaking hands, sharing elevators, and sitting across from people in tiny hotel rooms, I got sick. My head hurt. I was exhausted. I toggled between allergy meds, work, and long naps. And then, after about a week, I got better.

A couple of months later, while reporting on less conventional symptoms of COVID-19, I was struck by the more common ones, like headache and fatigue. Thinking back to my illness in January with the hindsight that soon after I left, the Bay Area became an early epicenter for COVID-19 in the US, I wondered: Could I have had a mild case of this disease? To find out, I started to look for an antibody test, a blood test to detect the immune signal that would tell me whether at some point in the past, I had been infected with SARS-CoV-2.

Now, as the US and several other countries emerge from the isolation orders intended to slow the spread of the virus, antibody testing is being touted as a way to know whether locked-down communities are reopening safely. Some view these tests as a way individuals can understand their risk in re-entering society, and when used routinely, a tool for businesses, schools, and the like to determine whether they are keeping their workers and their visitors healthy. But the scale of tests needed, and the scope of their use in social and economic decision-making could be ethically and logistically messy, says Shobita Parthasarathy, a science and public policy expert at the University of Michigan.

“The bottom line is that there is so much demand,” she says about workplace testing, “from universities, from prisons, from employers, from small businesses, and large businesses and schools.”

Performing this kind of widespread testing would be taking clinical measurements, like the presence or absence of antibodies to SARS-CoV-2, and using them to promote personal behavior and policy decisions affecting millions of people. As Parthasarathy and others told C&EN, it’s a lot of pressure to put on a simple test. After all, the tests can’t tell us how well we responded to the virus, or whether those antibodies will keep us from getting reinfected. And some of the tests have not been reliable, calling into question whether the results can even be trusted.

Yet the tests are getting better, says Alex Greninger, a virologist and diagnostics expert at the University of Washington. Months after allowing antibody tests into the market without regulatory vetting, the US Food and Drug Administration is finally demanding the tests be well-validated. Over time, he says, we should be able to get the reliable data we need to understand just how widespread COVID-19 has been.

But “over time” doesn’t jibe with the timelines of many lawmakers, who are impatient to restore the stalled economy and to resume normal life. The push to relax social distancing measures prompted Anthony Fauci, the top infectious disease expert in the US, to tell Congress that if we reopen too soon, “there is a real risk that you will trigger an outbreak that you may not be able to control.”

So, to better understand antibody tests, and how they might dictate a post-isolation world, I walked into a Quest Diagnostics laboratory on April 28, the day the company announced direct-to-consumer antibody testing. Under the skilled hands of a young phlebotomist who told me he had been working by himself all day in a waiting room of people in masks wanting to learn whether they had ever had COVID-19, I watched my blood whoosh into a tube.

If you build it, will the antibodies come?

Patricia Slev runs ARUP Laboratories, one of the four major reference laboratories in the United States (the others are Quest, LabCorp, and the Mayo Clinic). Affiliated with the University of Utah, ARUP runs thousands of different diagnostic tests, including common tests to diagnose diabetes and certain cancers, rarely requested tests for one-of-a-kind diseases, and more often than not these days, tests for COVID-19. As a reference lab, her staff also help independently validate that diagnostic tests work.

In the early days of the US pandemic, Slev’s team ran COVID-19 tests to determine if people with symptoms were actually infected with SARS-CoV-2. Those tests, usually based on a technology called polymerase chain reaction, look for viral RNA in a nose swab. Countries like Singapore and South Korea succeeded in reining in their initial outbreaks with widespread SARS-CoV-2 RNA testing, but in the US, the rollout of this kind of testing was slow. In many places, it was nearly impossible to get tested until several months into the outbreak, mainly because of supply and distribution problems and initially stringent requirements about who could be tested. So, to better understand the spread of COVID-19, it soon became important to see not only who was immediately infected, but who might have been missed because they weren’t tested while they were infected. By March, several tests were available in the US claiming to be able to detect SARS-CoV-2 antibodies, which seem to reach high levels a couple of weeks after infection.

Slev, an immunologist by training, explains that antibody tests detect proteins that our bodies create to fight off infections. When our immune system encounters a pathogen like SARS-CoV-2, it produces Y-shaped antibodies that glom onto small sections of viral proteins. Some of those stick to the virus itself, while others target viral proteins on the surface of infected cells, leading to destruction of the virus or of infected cells.

After an infection is cleared, some of those antibodies stick around in our blood, making them easy to track in what’s called a serological test. Depending on the manufacturer, COVID-19 serological tests measure a few different classes of antibodies: IgM, a five-piece antibody made early in infection that helps stimulate other parts of the immune system; IgG, a single Y-shaped antibody produced later in infection that is very specific for its infectious target; and IgA, an antibody made in large quantities in the mucus of our respiratory tracts, the place that SARS-CoV-2 first causes disease. Some tests measure combinations of these antibodies to give a broader picture of the antibody response.

The tests vary, but the basic design involves a tube of human blood that’s been centrifuged to separate the plasma from the red blood cells, and a well plate or another tube containing bits of viral protein. If the sample contains antibodies that recognize the viral protein in the other tube, they will stick to the protein. To visualize that sticking, a chemical reaction is built into the test—if antibodies are present, the sample gives off a signal, usually light or a color change, that a machine can measure.

So, after the phlebotomist drew my blood, he sent it with several other tubes to a central processing facility. He wasn’t sure what brand of antibody test I was going to get—Quest uses antibody tests from Abbott Laboratories and from Euroimmun. But after a quick call to their customer service, I learned I’d been given Abbott’s IgG test.

The Abbott test looks for antibodies to the SARS-CoV-2 nucleocapsid protein which, among other things, helps package up the genome when the virus has replicated. The viral protein used in the test is called an antigen; it’s what the antibody binds to.

“Antigens differ from assay to assay, but they all represent different proteins or parts of the virus. They are not standardized,” Slev says of the different antibody tests. “That’s one of the reasons why you want to validate these assays and determine the accuracy.”

Scientists developing these assays look for SARS-CoV-2 proteins that, like the nucleocapsid, are known or suspected to cause a strong antibody response in people. Viruses feature many such proteins, some of which are on the surface of the virus, while others are inside the virus and only become visible to the immune system when they end up on the surface of the infected cell.

Slev says that when developing these tests so quickly, researchers often will turn to proteins they can either easily make or buy. In some cases, the test contains just a fragment of an immune-stimulating viral protein rather than the whole viral protein. To figure out which protein, or perhaps, which section of a protein will best bind human antibodies in a test, teams will prototype several versions of their assay, using different antigens.

For example, when Ortho Clinical Diagnostics developed its assay, Chockalingam Palaniappan, the company’s chief innovation officer, says it tested the nucleocapsid, the spike protein, and the SARS-CoV-2 envelope protein. The envelope protein is important in infection, packaging, and budding from an infected human cell. Its scientists found that the most robust and reliable response came from the spike, a glycoprotein the virus uses to get inside cells. It focused its efforts on developing a commercial test that could detect the antibodies that would recognize it.

Palaniappan says Ortho also wanted to use the spike protein as a way to eventually detect neutralizing antibodies to SARS-CoV-2—ones that stick to the virus and keep it from infecting a human cell. The spike protein is a critical marker in both vaccine development and the use of convalescent plasma. This fraction of human plasma contains the antibodies from someone who has survived COVID-19, and is being tested as treatment in someone who is newly infected.

Roche Diagnostics went through a similar process when winnowing down viral protein candidates for its antibody test. Laura Parnas, a clinical chemist and senior scientist at the company, says Roche’s team created several prototypes of the test using four structural proteins from SARS-CoV-2 that they expected to trigger the production of human antibodies. The one that came out on top used a fragment of the nucleocapsid.

Like snowflakes, no two tests are alike

Tests that detect antibodies that bind to the same protein or protein fragment still might work differently. For example, Abbott and Roche both look for nucleocapsid antigens, but the way they measure them differs.

With Abbott’s test, the serum from my blood (the yellowish stuff at the top of the tube after it’s been sitting around for a bit) would have been mixed into a solution with tiny magnetic beads that are coated with the nucleocapsid protein. If I had been infected with SARS-CoV-2 and created any IgG antibodies to bind to the nucleocapsid, they would stick to the beads.

Then, the beads are washed to remove anything not bound to them. The machine adds a second antibody, designed to only react with any IgG in the sample. If my antibodies are attached to the nucleocapsid-coated beads, those IgG-recognizing antibodies, which are tagged with a fluorescent molecule, will bind to mine. Then, after another washing step, what the company calls a “trigger solution” is added. This causes those fluorescent molecules to light up, a signal measured by a machine that Abbott has designed for the diagnostic tests that many health-care providers already have. A positive test is one where the light emitted from the sample is greater than that emitted by the control sample, which could still have some lingering fluorescent antibody after the washing step.

By contrast, Roche’s test uses a light-based read out to detect the presence of antibodies to the SARS-CoV-2 nucleocapsid. A portion of the nucleocapsid pieces in the test are attached to ruthenium, and the rest to biotin. Antibodies are Y-shaped, and have two “arms.” Each arm of an antibody from a patient sample binds the nucleocapsid pieces. Magnetic particles are added to the mix to isolate the bound antibodies, bringing biotin and the ruthenium close together in what Parnas calls a “bridge” that is required for the reaction that produces light. The whole concoction gives off light when electricity is added to it. It’s called a double-antigen sandwich assay.

The tests are nearly all qualitative, Slev says. All they can tell me is whether or not I was previously infected with SARS-CoV-2. My results, if positive, wouldn’t be quantitative—they can’t measure how much antibody is floating in my blood. And none of them answer one of the most critical questions when considering using these tests to guide the reopening of society: Do any of those antibodies keep the virus from infecting me again?

The other thing that the SARS-CoV-2 antibody tests can’t tell us is how long your immunity lasts. For some infectious diseases, antibodies wane over time—this is why some vaccines require booster shots that prompt your body to start producing those antibodies again.

Greninger, the virologist at the University of Washington, says if he had to bet, he’d put his money on humans being able to produce enough antibodies to be protected for a long time. He points out that researchers and some clinics do capture that quantitative information with sensitive tests used in the lab—for things like developing vaccines and to determine if a person who has recovered from COVID-19 makes enough of the antibodies needed to be a convalescent plasma donor. But only time, and a whole bunch of testing, will tell.

“I’ll put my nickel down. There’s not going to be reinfection with this virus in the short term,” Greninger says, noting that early studies indicate that people infected with COVID-19 eventually generate large amounts of antibodies. And the virus mutates very slowly, meaning those antibodies should be effective for some time. He hedges. He could be wrong, but based on what he’s seen with protective antibody responses to other viruses, he thinks SARS-CoV-2 will be the same. “You don’t get infected with the same virus twice.”

Weeding out the bad tests

Since early April, the US Food and Drug Administration has granted several companies’ tests emergency use authorization, allowing their antibody tests to be used amid the pandemic. Like Abbott’s test, many are designed to be fast, mostly automated, and to be performed on machines that hospitals, health departments, and clinical labs might already have.

But as Slev says, no two tests are the same, and each one needs to be validated on patient samples.

To validate its antibody test, Roche tested about 5,200 samples of blood taken from people before SARS-CoV-2 began to spread. This is a common way of determining how specific the test is—in other words, how many false positives you might get. Roche’s test produced 10 false positives, giving the test a specificity rate of 99.8%. Roche also looked at clinical samples of people who had tested positive for active SARS-CoV-2 infection using the RNA-detection assay, known as a molecular test. By examining blood drawn from patients on a regular basis—every day or so—Roche scientists found that their antibody test was about 88% sensitive when administered between days 7 and 13 after a positive RNA test, so at those time points, there might be false negatives. But by day 14, the point at which many diagnostics and immunology experts say most people will have generated enough antibodies to be consistently detected, their antibody test was 100% sensitive.

“This assay design was the one that performed the best,” Roche’s Parnas says, adding that it had the least number of false positives compared to Roche’s other prototypes.

Ortho Clinical Diagnostics offers two antibody tests, one that measures IgG and another that measures IgG, IgM, and IgA. Palaniappan says that the company validated their IgG test on 400 blood samples drawn from life before SARS-CoV-2. There were no false positives. The test is 87.5% sensitive, and the data the company submitted to the FDA for its emergency use authorization were done on about 50 human samples from people with SARS-CoV-2.

Greninger, who also does clinical diagnostics for the University of Washington, independently tested the Abbott antibody test. In 1,020 blood samples from people without COVID-19, they got one false positive—“that’s pretty good,” he says. His team then tested blood samples taken periodically from a group of 125 people with COVID-19 still in the hospital. The Abbott test was 100% sensitive on 689 blood samples from that group of people—it could detect antibodies at different levels on different days.

A test that is 99.9% specific to SARS-CoV-2 and 100% sensitive is “really good news,” he says, compared to some of the first-available antibody tests, which had specificities of just 90%, meaning that if you tested it on 1,000 samples, 100 would give false positives.

How these less-accurate tests got on the market is frustrating to many people in the diagnostic world. The FDA initially allowed test makers to sell their products without vetting their validation results, and it’s unclear how many of those tests were sold, and whether the people who took them can trust their results. The FDA changed course in early May, requiring companies to share their results and to apply for emergency use authorization. In late May, after giving companies a grace period to submit their validation results or voluntarily withdraw their tests, the FDA pulled several tests from the market. Many in the testing community praised the FDA’s decision to finally become more hands-on.

“This revised policy makes a lot of sense and should have been in place” in March, when the first tests came on the market, says Scott Becker, CEO of the Association of Public Health Laboratories, in a statement.

What we can, and can’t ask of tests

Greninger, the virologist, says that outside a few bad apples, the tests we now have are much more reliable.

“The narrative for the last three weeks was that the antibody tests are hot garbage,” he says, referring to the time before the FDA changed its tune on requiring validation. “Now the new narrative . . . is that the antibody tests could be alright.”

With more reliable antibody tests, epidemiologists can better study the extent of community spread. The next challenge, Greninger says, is to produce enough of them. Parnas and Palaniappan both tell me that they are scaling up significantly.

“We estimate we will be at the maximum production around the end of May or mid-June,” Parnas says of Roche’s test, telling me that means tens of millions of tests. “And this is production for the entire globe.”

Abbott has also indicated that it can scale up quickly. As evidence of increasing availability, I had no problem getting its test, which costs $119, without a doctor’s order.

But Parthasarathy, the public policy expert, says, at least in the US, making enough tests isn’t as much a concern as getting them to the doctors, hospitals, and even businesses and schools who want to use them.

“Sometimes the demand can’t talk to the supply,” she says. “Either a hospital has a contract with a specific laboratory, and therefore can’t use another laboratory’s test or electronic platforms can’t actually communicate with one other—this is the level of infrastructural problem that we have.”

She also points out the ethical dilemma that workplace testing could pose: people having to present evidence of immunity to work could create even more social stratification and disenfranchisement in an already stressed workforce.

And, echoing Greninger, she says the message is still out there that the tests aren’t very good, and that all the open scientific questions about whether people are actually immune and how long SARS-CoV-2 antibodies last mean that “we’re now putting a lot of burden on tests and expecting them to play a kind of social role that we’ve never expected tests to perform before.”

As I waited for my results in the waning days of the stay-at-home orders where I live, I wondered what they might mean for my behavior. Would I be less likely to observe social distancing if the test was positive? Would I double-down on staying at home if I was negative?

I clicked on the link in my email, logged into my testing account, and scrolled down. Negative. It’s an unsatisfying answer. Assuming my test is accurate (it’s the same brand Greninger independently validated), I still don’t know what my risk is of getting COVID-19. And had the test been positive, I still wouldn’t know whether I’d be immune to reinfection.

It’s for that reason that experts unequivocally say we need to be more careful about rushing to reopen global economies. A lot of people say that doing so will seed a second wave of infections later this summer or fall. Some countries that relaxed their social restrictions after successfully tamping down the virus have already seen infections begin to rebound. But, for much of the US, we are still in the first wave of our outbreak. Opening too fast could just prolong the top of our infection curve, as more people grapple with the same questions as me, and make personal decisions about how to reinsert themselves into a reopened economy.

“You have an infinite number of people making individual decisions, and they are all slightly different,” Parthasarathy says. “We don’t seem to be learning that that is not a good approach.”