Thorpe Lab

oxidative protein folding

Quiescin-sulfhydryl oxidases and oxidative protein folding

Our group, in collaboration with Donald Coppock, uncovered a new enzyme family (abbreviated QSOX: Quiescin-sulfhydryl oxidase) that is capable of facile and direct  insertion of disulfide bonds into reduced unfolded client proteins.  Mispaired disulfide bonds are corrected by isomerization using a second enzyme, protein disulfide isomerase (PDI).

qsox and pdi


QSOX enzymes are present in most eukaryotes, but they are absent in yeast/fungi.  Intracellularly, they are found in the endoplasmic reticulum and the Golgi.  In addition, a significant fraction of QSOX appears to be secreted from cells.  QSOX immunoreactivity is particularly prominent in cells with a heavy secretory load [see Immunohistochemistry]. 

QSOXs can contain one or two thioredoxin (Trx) domains as shown:


Next comes a helix-rich domain (HRR), followed by the FAD binding domain (Erv/ALR) originally identified by Fass and coworkers for the yeast Erv2p protein.  The flow of reducing equivalents from a protein substrate (arrow 1), through the two redox-active disulfides (CxxC motifs; arrow 2), to the flavin ring (arrow 3) and then to molecular oxygen (arrow 4) is shown here:

We are currently studying representative examples of both one- and two-Trx QSOX enzymes to learn how these proteins are so effective at generating disulfide bonds in unfolded reduced proteins.


Augmenter of liver regeneration – an enigmatic flavoprotein

Human augmenter of liver regeneration (ALR, a.k.a. hepatopoietin, HSS, GFER) is the most enigmatic of those sulfhydryl oxidases that show homology with the flavin binding domain of QSOX.  Early studies showed that damaged liver releases a circulatory growth factor eventually found to be a small flavin-linked sulfhydryl oxidase.  ALR is found both extracellularly and intracellularly.  It is particularly abundant in the intermembrane space of the mitochondrion where it drives the retention of certain proteins via disulfide bond insertion:


Our work on ALR involves both the short (cytokine-like) and long (mitochondrial) form of the enzyme.  The figure shows the placement of the isoalloxazine ring of the FAD cofactor and the redox active (proximal) disulfide within the core flavin-binding domain.  The longer form (bar diagram) has an additional N-terminal extension that we have found is crucial for interaction with MIA40.




Recently we have worked with the laboratory of Dr. Brian Bahnson to obtain a crystal structure of the short form of human ALR  (PubMed).  


Oxidative Protein Folding - in vitro models

We are continuing our efforts to recapitulate oxidative protein folding in vitro:  in particular by developing efficient systems for folding proteins with complex disulfide connectivities.  The insights gained should help inform current debate concerning the way disulfide bonds are generated in the endoplasmic reticulum of higher eukaryotes.

We have found that riboflavin binding protein (RfBP) is a useful oxidative protein folding substrate.  RFBP has 9 disulfides (and consequently 33 million disulfide isomers for the fully oxidized protein).  A key advantage is that the native apoprotein binds riboflavin tightly and rapidly with complete quenching of the flavin fluorescence.  This allows us to follow oxidative protein folding continuously.



 We have shown that efficient oxidative refolding of RfBP can be achieved with nanomolar levels of QSOX, and concentrations of reduced PDI that are realistic for the endoplasmic reticulum.  Neither oxidized PDI nor glutathione are necessary for rapid folding.

In this model system we can also dispense with QSOX and use mixtures of reduced and oxidized PDI as a redox buffer without other redox components (in particular oxidized or reduced glutathione).  The fastest regain of riboflavin binding ability comes when the ratio of reduced/oxidized PDI is the largest (a little oxidized PDI is necessary for the stoichiometric oxidization of the 9 disulfides in native RfBP).  It seems likely that the optimal ratio of reduced/oxidized PDI may be dependent on the nature, and particularly disulfide complexity, of the client protein.  Nevertheless, in this case, refolding occurs most efficiently at the most reducing of PDI redox poises.

Since both reduced PDI is an extremely poor substrate of QSOX - and QSOX works best with unfolded reduced proteins - then QSOX can selectively insert disulfides into client proteins leaving reduced PDI to isomerize the mistakes it makes.  If this were to operate in the endoplasmic reticulum, collateral oxidation of the glutathione redox pool would also be avoided - since the primary oxidation of cliient proteins occurs via QSOX, and not oxidized PDI.


The extent to which QSOX and reduced PDI cooperate in vivo is currently unknown.

Inhibitors of Oxidative Protein Folding

Studies aimed towards the generation of inhibitors directed towards QSOX, and other enzymes with CxxC motifs, have led to the unexpected observation that arsenic(III) species can interfere with protein folding pathways by sequestration of reduced unfolded proteins.  This binding is dependent on the arsenic species and on the concentration of competing reduced glutathione.  Binding is of sufficient avidity to warrant consideration as one contributor to the physiological and pharmacological effects of arsenicals.

Arsenical are now being used for the treatment of acute promyelocytic leukemia and are being tested with other cancers.  In addition to specific protein targets of arsenicals, we have shown that As(III) species can have a more general effect by binding to protein thiols during protein folding.  This effect is likely to be most pronounced for cysteine-rich proteins – for example those found in secreted proteins prior to disulfide bond generation.

In the figure above, "A" is the unfolded reduced protein could yield the correctly-folded disulfide-bridged protein B.  Our experiments show that three representative reduced proteins (schematically represented by "A") bind a range of arsenicals with stoichiometries reflecting the nature of the arsenic(III) species and the number of free thiols available (see below).  This binding is sufficiently tight to remain significant in the presence of 5 mM reduced glutathione.  We hypothesize that some proteins may exhibit particularly avid dithiol or trithiol site for arsenicals and that binding could persist as disulfide bonds are inserted around these loci (C).  We have also shown that the monomethyl metabolite of arsenite (MMA, pink square) generates amyloid-like filaments with reduced RNase (E).  We have also suggested that arsenicals could also compromise some fraction of protein folding in the absence of disulfide bond generation.

The titration of reduced riboflavin bindiing protein with three arsenicals is shown below.  Binding is also followed by fluorescence and by the suppression of thiol reactivity after complexation by As(III) species.

These initial experiments suggest new strategies for the design of reagents that might selectively inhibit cells that carry a heavy secretory load.


Disulfide-rich Biomaterials

Disulfide-rich structural proteins that are assembled into matrices, fibers, and envelopes are important in a wide range of biological settings – from the keratin-associated protein matrices in differentiating keratinocytes, to disulfide-rich structures formed during spermatogenesis.  We are currently studying a disulfide-rich biomaterial that is both commonplace and cryptic.  Under the shell of a chicken egg is parchment-like film assembled from two layers of proteinaceous fibers (depicted schematically at the left, and in the confocal image at the right).


egg anatomy


Surprisingly, the major protein(s) from which these fibers are constructed have remained uncertain because lysine-derived crosslinks between the individual proteins render the material intractable to the standard tools of the protein chemist.  We have recently discovered a new disulfide-rich protein (CREMP: cysteine-rich eggshell membrane protein) in these membranes [Ref].  CREMP shows a striking repetition of disulfide-containing modules.  The regularity of cysteine spacing, and the remarkable conservation of intervening amino acids between modules, severely complicates the assembly of full-length sequences from genomic databases.

As an example, the figure below shows a small fragment of a translated avian EST sequence with an alternating pattern of a and b modules.   In addition to (ab)n patterns, we have also observed (abb)n and (a)n repeats in bird and reptile contigs.

These protein fragments bear a distant evolutionary relationship to spore coat proteins from slime molds.  However the CREMP fragments we have uncovered show a much more regular sequence and, consequently, display a much clearer modularity. We know that avian egg-shell membranes are highly ordered by solid state NMR, and that an abab 4-module segment, expressed in Escherichi coli, is also highly structured (both NMR approaches in collaboration with the Polenova laboratory at UD [Ref]).

Many outstanding questions remain.  Here are a few:

  •  what is the sequence of a full-length CREMP protein?
  •  do multiple CREMP proteins exist in birds and egg-laying reptiles?
  •  how are these novel disulfide-rich proteins incorporated in the fibers?
  •  while CREMP seem to be major constituents of eggshell membranes, what are the roles of the other protein components?
  •  and how does a chicken assemble these fibers over gloopy egg white anyway!?


You are here: Home Research