Likely produced by vaccine too, at least the ones that used full-length spike protein. The rejected BioNTech candidate for example (BNT162b1) had only the RBD.

Likely produced by vaccine too, at least the ones that used full-length spike protein. The rejected BioNTech candidate for example (BNT162b1) had only the RBD. BNT162b2 is the one everyone knows now because it was the one picked for phase 3 trials, and it uses the full spike. Obviously if the vaccine were the RBD-only one, there'd be no antibodies generated for NTD or S2 by the vaccine. Here's a paper where the studied the antibodies: https://www.degruyter.com/document/doi/10.1515/cclm-2021-0339/html https://www.degruyter.com/document/doi/10.1515/cclm-2021-033... Their assays(tests) may not have been quite specific enough for these other target locations, but it appears to cause antibodies against multiple sites on the spike. In the end, the OP article from NIH only analyzed the paper about a small set of people who recovered from infection and says nothing about analyzing vaccine recipients, because that's not what the paper was looking at. Now what you should ask: What is the distribution of antibodies for a vaccine vs infection and what does that mean for variants. Because the paper and the article are concerned that the results of infection seem to cause a lot of anti-NTD antibodies and while they work, the NTD domain is mutating a lot in the variants of concern. There's two lines of discussion that seem to result: 1. Is too much focus on anti-NTD from infection bad against variants long term? Is this how we're getting "reinfections" from variants? 2. Should new vaccines target certain parts that don't change like S2 to get even better and more universal vaccines against all the variants of covid19?