Delyan Hristov explained nicely, but i think disilfide (S-S) must not be issue if you already have reduced the proteins with 5 mg/mL NaBH4. Yes protein conformation and thermodynamic stability can be the major cause. first proteins have fewer S at surface, second its conformation don't allow the protein to even bind these fewer S. I suggest also give a try using NaOH instead of NaBH4. I hope this will help. 

If the thiols form the S-S bond natively, targeting them is not the best strategy.

Try introducing new cysteine at exposed position.

There are also good gold-binding peptide tags, you can try to introduce them within

the loops, avoiding tricky thiol reactions all together.

Typically thiols in proteins are internal and very few. Since carboxylic and amine groups are on the surface of the protein, much more and also interact with gold they might be more likely to react with the surface. Naturally there is also the disulfite bridge issue. 

Thank you for your time! And your answers remind me that the main problem maybe that the real position of the thiol group (of the genetically modified cys residue), and the crystal structure is under exporing.

Thank you for the suggestion of GBP, and the related project is under way parallelly.

Thank you again for your suggestions.

Delyan Hristov explained nicely, but i think disilfide (S-S) must not be issue if you already have reduced the proteins with 5 mg/mL NaBH4. Yes protein conformation and thermodynamic stability can be the major cause. first proteins have fewer S at surface, second its conformation don't allow the protein to even bind these fewer S. I suggest also give a try using NaOH instead of NaBH4. I hope this will help.