For a small protein, disappearance can occur by different mechanism, like intermediate conformationnal chemical exchange or exchange with water (for not structured region, and therefore solvent accessible).

Try different temperature : low temperature to decrease water exchange and low/high to change exchange regime (leave intermediate regime that drastically decrease pic intensity).

In addition to Yoan's excellent suggestion to try different temperatures, you could also try different buffers (i.e., different buffering chemicals). Some proteins simply give terrible spectra in one buffer and excellent spectra in a different buffer. You already stated that you changed the pH, but did you actually try using different buffers? For example, many NMR-based structural genomics consortia routinely do HSQC screening of their protein of interest in different buffers in order to determine which buffer gives the best spectra.

See, for example, the following webpage: http://www.nmr2.buffalo.edu/nesg.wiki/Buffer_optimization

Best of luck!

Try dilution. Peak broadening could be because of exchange between monomer and dimer populations, where the missing peaks arise from the dimer interface. Also we are pretty sure that we have seen an example of this where the structure in the crystal is an asymmetric dimer, and what we see in solution is exchange between A:B and B:A homodimers leading to relatively large chemical shift differences and intermediate exchange.

If I may add my 2 cents...

Sometimes you will not be able to observe all the peaks. This can be for a number of reasons such as dimerisation (mentioned by Paul Driscoll) as well as intermediate exchange (mentioned by Yoan Monneau), but there can be other reasons too. With some enzymes with flexible active sites, you will quite often not be able to observe them without the addition of an inhibitor that helps lock the flexible regions in place. Also - if it is a metallo-enzyme or you have a paramagnetic contaminant, you can loose signals below the noise by PRE (paramagnetic relaxation enhancement).

Is there a published crystal structure? If you have started assignment, perhaps you can map the observable peaks onto the structure and see if you are missing peaks in a specific localised area or at random.

Are you sure you aren't seeing all the peaks? Have you done 3D experiments yet? It is not uncommon for a number of peaks to overlap. You won't be able to distinguish this without higher dimension experiments. Also - sometimes the N-terminus is not resolved due to intermediate exchange. Even so, missing 20% of your protein still seems a bit much. I do assume that you are not counting any proline residues - you do not expect to see them at all in an HSQC (no amide proton).

And finally - are you sure your protein hasn't been chopped up by a protease? I had a case where I was trying to find missing peaks from a protein until mass spec data told me that I was missing a whole segment of the protein, presumably due to digestion.

Hope this helps - Tj