Some thoughts on SARS-CoV-2 and heparan sulfate

After an interaction on Twitter with a colleague who is associated with #LongCovid and #TeamClots, he asked me for some references. I thought what to send, and then realised that references plus something a bit more than a Tweet might be useful, so here goes.

A brief recap

Our first foray into the interactions of SARS-CoV-2 with heparin, and so with cellular heparan sulfate (HS), was published just over 2 years ago in March 2020, written up on a GoogleDoc on the  Friday night and the following Saturday. This was not a blind stab, but based on previous work by our Italian colleagues which demonstrated an interaction SARS-CoV with heparin and informed by comparison the receptor binding domain (RBD) of the viruses. In the next months we firmed up the data, put two more preprints out and then at the end of the year the data were eventually published in a peer reviewed journal.

We had considered that the SARS-CoV-2 spike protein might use a dual receptor system analogous to that of the fibroblast growth factors and many other protein ligands that regulate cell function in development, homeostasis and disease. This transpires to be the case.  Others in the glycosaminoglycan community were active, exploring structure-function and how binding to heparan sulfate is a prerequisite for the Spike protein to load onto ACE-2. Thus, in this respect Spike looks like a classic endogenous heparan sulfate-dependent ligand, requiring a ternary complex of ligand, heparan sulfate co-receptor and transmembrane protein receptor.

Enter stage left neuropilin-1 (NRP-1). In our excitement we had forgotten about this. The Spike protein binds NRP-1, e.g., here. It likely that this is at least one of the mechanisms whereby it accesses the brain

e.g., here and here), through binding NRP-1 on olfactory neurones or via the blood-brain barrier (NRP1 remember is also important in controlling angiogenesis). The interaction with NRP-1 is predictable. NRP-1 binds heparin and is a heparin binding binding protein. That is its acidic domains are reasonably heparin mimetic and enable it to bind to some, but by no means all, heparin-binding proteins. To confound matters, NRP-1 is a facultative proteoglycan, so can carry a heparan sulfate (or chondroitin sulfate) chain. It is worth noting the anatomical symmetry of the nervous system and the vasculature extends to the molecular level, with NRP-1 being a prime example of the latter. What the is glycanation status of NRP-1 on the olfactory neurons? Nobody knows. An interesting aside here is the link between 3-O sulfation of heparan sulfate and NRP-1.

Frustrations

There are many, many loose ends

The RBD and the Spike protein are very difficult proteins to work with. Regardless of the source of the protein, it loses heparin-binding ability very easily, which correlates with a conformational change in the protein, detectable by circular dichroism – and this quite a long time before any decrease in ACE2 binding can be observed. There is also something unusual in the interaction of the RBD with heparin in vitro. You need SDS to dissociate the RBD (here and here). I have seen this some years ago with thrombospondin (TSH), where we had to use urea to dissociate bound thrombospondin. This was never published, as the data were consequently rather messy. My explanation was that the binding reaction was a two step process

TSH + HS = TSH:HS > TSH*:HS 

Where TSH* is a conformation induced by the initial reversible binding event, which results in  complex that cannot dissociate. Something similar may be happening with Spike, and which would mean that heparanase may play an important role in mediating the loading of SARS-CoV-2 onto ACE2. At least one heparanase inhibitor is a potent in vitro inhibitor of SARS-CoV-2 infection and it may be that the inhibitory effects of a number of sulfated sugars are due in part to inhibiting heparanase.

Therapy

First to vent some frustration. The world spent millions on clinical trials based on fraudulent data (hydroxychloroquine,  ivermectin) while people were dying, and nothing on trials targetting the heparan sulfate-dependent mechanism of SARS-CoV-2. This despite the fact that (1) coagulopathy is treatable with heparins and (2) there are heparan sulfate inspired products that are excellent stimulators of wound repair in the context of chronic inflammation, e.g., lower limb diabetic ulcers from OTR3. A clinical trial of heparin in non-SARS-CoV-2 ARDS looked very promising (sadly behind a paywall). We now have a first study in SARS-CoV-2 and it is rather disappointing – no major effect. 

Why might that be? The most likely explanation is that heparin is not the right compound. Though widely used as an anticoagulant, it is a very specialised fraction of heparan sulfate, produced by the mast cell. Compared to heparan sulfate, heparin is very short, highly sulfated and lacks the extensive domain structure of heparan sulfate. The much longer heparan sulfate, with its N-acetyl and transition domains separating the sulfated domains has far greater reach and is able to bend quite sharply – this is seen with VEGF, which heparin binds rather poorly, since it has trouble engaging both binding sites on the VEGF dimer. The ability of HS to bridge several proteins is likely important in its functions (remember, in tissues and on cells there is only heparan sulfate, no heparin). These functions relate to binding over 800 extracellular proteins that regulate cell communication. So rather than heparin, a heparanase resistant mimetic of heparan sulfate like the OTR3 compounds may be the molecules of choice. There is as yet only anecdotal off label case reports, but these all suggest such compounds may be very effective both against SARS-CoV-2 and in reversing the molecular pathology of LongCovid, which include, but is not restricted to microclots.

Advertisement