Thorpe Lab

oxidative protein folding

Putting the cart before the horse

When you know that a particular chemical reaction occurs in biology (e.g. glucose -> glucose-6-phosphate in glycolysis) you can use this activity to guide the isolation of the enzyme that catalyzes that transformation. During enzyme purification the aim is to winnow away the thousands of contaminating proteins leaving just the one you want. Traditionally, great emphasis has been placed on the need for purity – the late Efraim Racker opined "don’t waste clean thinking on dirty enzymes"!

While most enzymes known today were initially purified with prior knowledge of their likely substrates and products, and with a good idea of their place in metabolism, a few were isolated first as proteins. Knowledge of their enzymatic activity came later – putting the cart before the horse as it were. Our discovery of the QSOX family of enzymes came about in this back-handed way.  

A colleague, Hal White, told us of his finding that egg white contains FAD - a derivative of the vitamin riboflavin that is usually associated with redox enzymes. QSOX purification newWe jointly wondered “why is FAD in egg white” since no FAD-dependent enzyme was known from this source. Detractors of this type of non-hypothesis-driven research want an overarching hypothesis and an encirclement of specific aims to provide intellectual cover. We had none of this – but answering the “why is FAD in egg white” question has led to the discovery of two new enzyme families, to an intriguing new disulfide-rich biomaterial and to suggestions of a new strategy to slow the growth of human tumors. At the end of this blog post there are links to provide an entrée to some of these aspects.

So, out of curiosity, we purified the protein binding FAD in egg white. We did not know whether the protein would prove to be an enzyme – so we had no enzyme activity to follow. But FAD is yellow and that’s what we followed.   In crude egg white only about one protein molecule out of about 10,000 was the FAD-binding protein we sought – but we were able fish out a small amount of a pure bright yellow enzyme from a large volume of chicken egg white.

charge transfer newSo now we had the protein pure - what to do next? Since flavins are usually involved in redox reactions we tested a number of likely reducing substrates. We screened candidate substrates by adding them to the yellow protein one by one. We were looking for a color change because a real substrate of the enzyme would likely modify the visible spectrum of the enzyme – subtle changes would require a spectrophotometer to detect – major changes would be evident by eye. After trying a number of potential substrates that had no effect on the spectrum, we found that thiol-containing compounds induced dramatic changes in color of the flavin cofactor bound to the protein. In the figure you can see stills taken from a video showing the bright yellow color of the enzyme flash through a blue tint on its way to paler yellow as oxygen is depleted from the solution.

So this was the essential clue that led us to the discovery of a new family of sulfhydryl oxidase enzymes. The chemistry was typical of sulfhydryl oxidases that use that yellow FAD cofactor to catalyze the oxidation of thiols in the presence of molecular oxygen.

                        2 R-SH   +    O2        ->     R-S-S-R      +      H2O2

We then surveyed the range of thiol-containing substrates that the enzyme could process in pure solution. We found that the enzyme could oxidize a wide range of –SH groups found in both small and large molecules. Every protein we tested was a substrate of the enzyme providing that it was flexible enough to present two or more cysteine side chains in a flexible environment – so reduced unfolded proteins (prepared by breaking their native disulfide bonds prior to QSOX  cart before horse blog newassay) were excellent substrates of the enzyme. The enzyme worked at rates that were far greater than other candidates for disulfide bond generation in higher eukaryotes.

Fortunately the catalyst we had stumbled upon in egg white is not just for the birds! QSOX enzymes (the official abbreviation of the Quiescin-sulfhydryl oxidases) are found in almost all non-fungal eukaryotes - from small unicellular algae to humans. So what are these ancient enzymes doing? It is probably safe to say that the role of QSOX enzymes is to make disulfides (with the possible ancillary function of generating hydrogen peroxide) - but what disulfides exactly, and in what cellular or tissue locales?  

So we have done things backwards – we found an enzyme by accident and then constructed a story of a likely physiological role based on its prodigious ability to make disulfide bonds in protein substrates in the laboratory. Now researchers are searching for the molecular roles of QSOX in cells and tissues. For example QSOX enzymes have been shown by the Bulleid laboratory to be very efficiently secreted and the work of the Fass and Lake groups have provided intriguing evidence for a major extracellular role of QSOX in the extracellular matrix. Several studies now associate the up-regulation of QSOX with an adverse outcome in certain cancers.

Why, we wonder, is QSOX secreted into blood, sweat and tears! – is it an incidental product of the secretory machinery, or is QSOX doing something useful in biological fluids? Finally what is the role of QSOX in unicellular organisms - from benign marine algae to the pathogenic trypanosoma?

Lots of important questions – and we still don’t know why QSOX is secreted into egg white!    

Of possible interest ...

Discovery of QSOX – the real story ([caution: gripe] to say that QSOX was “isolated more than three decades ago” does not, in the writer's opinion, do justice to the facts of the case!).  For a history of the early developments see QSOX-ology.

QSOX and cancer

CREMP: an amazing disulfide-rich biomaterial

QSOX as a “bad actor” in (some forms of) breast cancer

The "bad actor" quote comes from a recent paper by Douglas Lake and collaborators: "Expression of Quiescin Sulfhydryl Oxidase 1 is associated with a highly invasive phenotype and correlates with a poor prognosis in luminal B breast cancer"

The enzyme: readers of earlier blogs may know that Quiescin-sulfhydryl oxidases (QSOXs) catalyze disulfide bond generation within conformationally flexible peptides and proteins. In the enzymological equivalent of a "bucket brigade", pairs of electrons are passed from the dithiol-containing substrate to molecular oxygen as shown in the Figure. In the crystal structure of the mouse QSOX1 enzyme the deep blue [Trx1] domain is handing over a pair of electrons to the green [ERV] domain prior to the final stages in catalysis (steps 3 and 4).

QSOX Blog breast c


Breast cancer: According to the American Cancer Society, women have a 1 in 8 lifetime probability of developing breast cancer – accounting for 30% of new cancer cases among women in 2012. There are multiple categories of breast cancer: the one that Lake and colleagues were studying was luminal B breast cancer – an estrogen receptor-positive subtype.

The first paper: Lake and colleagues found that QSOX1 protein is highly over-expressed in luminal B cancers and cell lines. Knockdowns of QSOX1 protein levels suppressed cell proliferation, and dramatically reduced the invasive potential of these tumor cells. Importantly, exogenously added QSOX1 enzyme restored the invasive potential of QSOX-depleted cells. Lake and coworkers suggest that elevated QSOX levels lead to increased activity of matrix metalloproteinase 9 with a consequent stimulation of degradation of ECM components and an increased propensity for metastasis.
Lake and coworkers used the GOBO database "Gene Expression Based Outcome for Breast Cancer Online" to make the case that elevated QSOX1 message levels are associated with a poor prognosis in both luminal A and B cancers.

In a second study " Elevated Transcription of the Gene QSOX1 Encoding Quiescin Q6 Sulfhydryl Oxidase 1 in Breast Cancer", Soloviev et al. analyzed expressed sequence tag (EST) and serial analysis of gene expression (SAGE) databases. They also concluded that QSOX1 is highly overexpressed in some breast cancers. They used quantitative PCR to analyze a series of breast tumors, and observed a strong correlation between QSOX1 expression levels and the progression of the disease.

Other cancers: Previous posts have reported that overexpression of QSOX1 correlates with the invasive potential of prostate and pancreatic cancer tumors. In addition to modulating the activity of MMP9, what other activities of QSOX1 contribute to the increased fitness of particular tumor cells? Are these effects confined to extracellular disulfide bond regeneration/remodeling – or is the other product of QSOX catalysis - hydrogen peroxide - a factor?

But is QSOX really a "bad actor"?  In a paper published in 2012, Pernodet et al. reached exactly the opposite conclusion to the studies reported above. They analyzed 217 invasive ductal carcinoma patients in their paper "High expression of QSOX1 reduces tumorogenesis, and is associated with a better outcome for breast cancer patients" Pernodet et al. found an inverse correlation between QSOX expression levels and the aggressiveness of the breast tumors they examined.

They report that "QSOX1 also reduces the invasive potential of cells by reducing cell migration and decreases the activity of the matrix metalloproteinase, MMP-2, involved in these mechanisms. Moreover, in vivo experiments show that QSOX1 drastically reduces the tumor development"

This chemist-contributor is now in a muddle! How to reconcile the differences between this work and the two studies reported at the beginning of this post – do they reflect differences in tumor populations studied? Hopefully the potential importance of these observations will encourage a more searching and comprehensive examination of the role of QSOX in tumorigenesis.

May 19th addendum to original May 8, 2013 post: for a more detailed comparison of the papers, see a recent Editorial in Breast Cancer Research by Postovit and colleagues "Illuminating luminal B: QSOX1 as a subtype-specific biomarker" [PubMed or Publisher]

Conduct Unbecoming

We have published a body of work on a small sulfhydryl oxidase – reasonably fundamental stuff I thought.  We found new substrates, new enzymatic activities, explored enzymatic mechanisms and, with colleagues, solved crystal structures.  So I was disconcerted that a speaker at a recent conference never mentioned any of our contributions in his talk on this same enzyme (I was there, and we have other witnesses!).   There was no acknowledgment in the introduction, nothing in the middle and nothing at the end.

So what is a conferee to do?  Should I have complained in the question-and-answer period and sounded like a whiner?  Should I raise this with the speaker privately and listen to his excuses? Should I send an aggrieved letter?  I did none of these things – maybe the speaker meant to credit our work but nerves overwhelmed him.  Or maybe he just forgot.  Give the fellow the benefit of the doubt.

My annoyance was rekindled recently when I read one of his papers and found that our contributions to the same enzyme were again essentially unacknowledged – an uninformed reader would conclude that the whole intellectual framework of the paper originated with the authors.  Yet many of the central ideas had been clearly identified years earlier.  Now we don’t need witnesses - the literature is our record!

This sort of behavior is, at best, ungracious.  What does he lose by citing the literature appropriately and by acknowledging in talks that others have played a role in the development of the field?  A failure to cite relevant work is also unethical – to quote from the Ethical Guidelines for Publication from the ACS:

“... An author is obligated to perform a literature search to find, and then cite, the original publications that describe closely related work….”

“Colin, just get over it – these things happen all the time” …. I was given this sage advice almost 20 years ago (for a different case entirely!).  The very next day my “turn-the-other–cheek” colleague heard his own work characterized with insufficient deference.  This time it was personal, and he went ballistic!

Maybe that’s what I’ll do next time!

Expanding the Biological NMR Toolbox

What’s your favorite element?  - for many scientists studying oxidative protein folding it would have to be sulfur.  Not only does the oxidation of cysteinyl sulfur generate the constellation of structural disulfide bonds found in secreted proteins, but cysteine residues are an essential catalytic ingredient in the enzymes that form and isomerize these very disulfide linkages.  While carbon, nitrogen and hydrogen can be routinely probed by nuclear magnetic resonance, the application of biological NMR to sulfur is impractical - there is no suitable sensitive NMR-active isotope for this critical player in oxidative protein folding.

Now Sharon Rozovsky and her colleagues have devised a simple way to substitute the NMR-active 77Se isotope for sulfur in recombinant proteins [PubMed].  Escherichia coli cells are grown in minimal media until they approach sulfur starvation, and then the medium is supplemented with ratios of Se (as selenite) to S (as sulfate) that are needed to achieve the desired incorporation ratio of Se/S.  Anyone who expresses recombinant proteins in minimal media should be able to make this method work following the protocol outlined in Schaefer et al. [PubMed].

One of the proteins they have studied is augmenter of liver regeneration (ALR) – an enzyme that drives disulfide bond generation in the intermembrane space of the mitochondrion.  The short form of ALR is a covalent homodimer (32 kDa) containing four structural disulfide bonds, two catalytic disulfides, and two FAD prosthetic groups.  Using the random selenium incorporation method of Rozovsky and coworkers, Se/S ratios of up to 90% are readily achievable for human ALR. The resulting Se-ALR enzyme is almost as stable as the wild type protein, and the substituted protein retains significant enzymatic activity.  Importantly, multiple selenocysteine and selenomethionine resonances can now be visualized by conventional solution NMR without the need for high field strengths or lengthy acquisition times.

But what are the structural consequences for such a wholesale substitution of sulfur for selenium in ALR?  Schaefer et al. were able to crystallize the 90% substituted protein and compare it to the coordinates of the human ALR protein already solved in the laboratory of Brian Bahnson [PubMed].  On the left you can see the structure of Se-ALR refined to a resolution of 1.5 Angstroms with an occupancy of 90% Se / 10% S [PDB link] .  The protein fold is essentially identical to that of native human ALR.


 The right panels show the reduction of the enzyme by the model substrate dithiothreitol (DTT) monitored through the changes in the flavin spectrum.  (Top panel is wild-type ALR; bottom panel is Se-rich ALR.  Note the new long-wavelength absorbance feature that intervenes before the accumulation of the blue semiquinone).

In sum, the work of Rozovsky and coworkers should help to put selenium into the biological NMR toolbox - as an environmental probe, and for the characterization of redox and ligation states. While attention has been drawn to the differences in reactivity between sulfur and selenium – they really do share lots in common.

So if you want to try an NMR-active surrogate for sulfur in your favorite protein … give 77Se a spin!


1)   The paper: Stephanie A. Schaefer, Ming Dong, Renee P. Rubenstein, Wayne A. Wilkie, Brian J. Bahnson, Colin Thorpe and Sharon Rozovsky. 77Se Enrichment of Proteins Expands the Biological NMR Toolbox. Journal of Molecular Biology (2012 [PubMed].

2)  Sharon Rozovsky's website is here

3)  Of course genetic incorporation of Se provides for site specificity - but to do this wholesale for a multicysteine-containing protein would generally require an unacceptable number of mutations to generate SECIS elements for each position.

4)  In terms of “disulfides” 90% selenium incorporation corresponds to  1% -S-S-, 18% -S-Se- (or -Se-S-) and 81% -Se-Se-.  So the majority of crosslinks in the selenium-ALR structure depicted in the Figure represent diselenides.

5)  The PDB coordinates for the Selenium-rich form of ALR are to be found here


You are here: Home Blog