We
study protein crystallization in solutions of human serum albumin
(HSA) exhibiting a metastable liquid–liquid phase separation
(LLPS) in the presence of trivalent salts. Specifically, we focus
on the effects of dense liquid phases (DLPs) on the crystallization
pathways. On the basis of the phase diagram, we choose two conditions
around the LLPS binodal: one condition is located close to, but outside
the LLPS region, resulting in protein clusters, but no macroscopic
LLPS. Yet, a surface-enhanced unstable DLP layer is observed at the
surface of the cuvette (wetting). The second condition, inside the
LLPS binodal, leads to a macroscopic metastable DLP. The crystallization
is followed by optical microscopy and small-angle X-ray and neutron
scattering (SAXS/SANS) as well as by ultraviolet–visible spectroscopy
to explore the role of LLPS. In no case evidence of nucleation inside
the DLP is observed. SAXS and SANS show a monotonous growth of the
crystals and a decrease of the overall material in the sample. We
thus conclude that the existence of a metastable LLPS is not a sufficient
condition for a two-step nucleation process. The DLP serves as a reservoir
and crystal growth can be described by the Bergeron process, i.e.,
crystals grow directly into the dilute phase at the expense of the
DLP. Furthermore, the crystallographic analysis of the resulting crystals
shows that crystals with different morphology grown under different
conditions share a similar crystal structure and that the metal ions
create two bridging contacts within the unit cell and stabilize it.