Friday, March 27, 2015

Tilting at Ferric Windmills

As documented last week, I contacted a journal about very questionable methods used to conclude ferric iron could be detected by fluorescence methods in water. I was already aware that this was not the only occurrence of a research group failing to properly account for the aqueous solution chemistry of Fe3+ in their studies. With a reignited hyperawareness about this problem in the fluorescence sensing field, I noticed another study making similar mistakes while catching up on recent literature this week. I alerted the EIC of this journal to the problem as well, and got a completely different response with respect to both the scientific and peer review concerns. Since this is an ongoing dialogue, I will not comment any further until the issue is resolved. This second incident prompted me to conduct a more thorough investigation of the underlying prevalence of these protocols in Fe3+ sensing.

Counts of papers containing questionable ferric sensing methodologies.
*indexed in webofknowledge as of 3/26/205.
Using webofknowledge, I searched the combination of (ferric OR Fe(III) OR Fe3+) AND (sensor OR probe) AND (water OR aqueous OR buffer) AND fluorescence. To decrease the volume of references to analyze, I limited the results to those from journals published by the ACS, the RSC and Wiley since they have a reputation for publishing reliable sensor papers (reducing the number of results from ~500 to ~150). I then went through the search results looking for evidence that the experiments would be at risk for giving spurious results. I am not sure about the exact speciation/stability of Fe3+ in mixed organic/aqueous solvents; however, my experience suggests that any significant amount of water will be problematic. I excluded several studies using 1% aqueous content in solvents like THF and methanol, as that seemed to be a relatively safe protocol. Anything with 20% or more water made the "suspicious list". The overwhelming majority claimed to work in pure water, aqueous pH ~ 7 buffer or with 50% or less organic solvent added to water/buffer. There are many fewer (not counted) that describe protocols for working and handling Fe3+ in acidified aqueous solution, and many others using pure organic solvent (also excluded from consideration). Of the 150 results, 51 contained possible problems. The results broken down by publication year and journal name are shown in the figure above.

The results are quite informative. Before the Rurack paper in 2005, there were no Fe3+ sensing papers in ACS/RSC/Wiley family of journals using water (his paper has been cited >250 times). My paper showing the errors in Rurack's aqueous results was published in 2010. The yearly trends however, suggest that the acceptance of questionable (invalid) Fe3+ titration protocols are increasing rather than decreasing. Presumably, every published paper using similar methods provides unwarranted precedence for adoption in future studies. Whether there is a connection or not is unclear, but Inorg. Chem. where my paper was published, has not published a problematic paper in the Fe3+ sensing field that I can find.

Full disclosure, there are a few papers in this collection that are difficult to analyze, particularly those from the polymer/materials literature. A more thorough investigation would be required to fully evaluate the results in detail. Furthermore, some of the studies have ambiguous or nonexistent experimental protocols in the paper and/or the supporting information, which makes evaluation difficult or impossible. If procedures for measuring/adjusting the pH are not listed, I assume that this was not a consideration. A few papers don't even list the counter ion for the Fe3+. Also, I excluded papers that may have had questionable handling of Fe3+ solutions (selectivity studies), but the Fe3+ response was not a significant component of the paper's discussion/conclusions. The 51 papers all attempt to conclude something specific about Fe3+ detection in aqueous solution. No papers from Elsevier or Springer were examined.

There are some common issues in many of the 51 papers. For example, making stock solutions of FeCl3 in water by just mixing (i.e. without adjusting the pH to <4), and titrating Fe3+ into neutral aqueous solution. Some use phosphate buffer, which would generate Fe(PO4), a water insoluble salt (assuming all the Fe3+ wasn't already precipitated as insoluble Fe(OH)3). A cringe-worthy method used in more than one paper to confirm Fe3+ binds to the sensor in aqueous solution, is mass spectroscopic detection of the ferric complex prepared in methanol. I have no doubt that they detected the complex, but Fe3+ in methanol is completely different than in water. The MS data in methanol does not confirm anything about the species that are (not) present in aqueous solution. 

What does one do in such a situation? Even if I was somehow asked to referee every paper on Fe3+ sensing, there were more papers on the topic published last year than I could possibly handle, even if they were the only kind of requests I received/accepted. A broad search suggests an upper limit of 190 Fe3+ sensor papers were published in 2014. I have no desire to comb the literature and complain to editors/authors every time I find a problematic paper. I've already published a paper in a good journal that includes a cautionary tale about these issues, but it does not seem to have permeated the sensing field zeitgeist. As I mentioned in the previous post, it's discouraging to find out that some individuals need a reminder about the relevant/underlying undergraduate inorganic chemistry. I have no doubt that researchers in other fields can point to similar problems in the literature on other topics. How does one effectively get through to journals, peer reviewers and researchers without wasting time that should be spent on other aspects of academic science? It seems antiquated in the information age that such mistakes should persist and be perpetuated, but the traditional practice of publishing an opposing research study is the only clear recourse.

Friday, March 20, 2015

Metal sensing malarkey: (or the expected virtue of ignorance)

To paraphrase Stephen Colbert, "who's not citing me now?" It's a refrain that many researchers can identify with, but only a small part of why I was motivated to write this post on peer review, journal policy and bad science. A little background first. A couple of months ago, I attended a research presentation that covered several papers on fluorescent sensors. My experience with fluorescent sensors dates back to the late 1990s and early 2000s when I completed my Ph.D. thesis on zinc probes with Stephen Lippard at MIT. At the time we started, fluorescent sensors for metal ions were still a niche area in inorganic/bioinorganic chemistry. Now, it's a widespread topic of research. As many realize, your Ph.D. work will follow (haunt) you long after graduation, even if you no longer actively work in that area. My group published a couple of papers on fluorescent probes for ferric iron 5 years ago, but that was the last time I actively worked in the area. Despite moving on to other topics of inquiry, I am still inundated with referee requests on fluorescent sensor papers, so I remain familiar with the progress and problems in the field.

The presentation focused on two papers published in ACS Applied Materials & Interfaces on fluoride and iron sensing respectively. I questioned the lead author giving the presentation about the validity of the methods and data interpretation, but failed to make any headway or get any concession that there may have been problems with the protocols or conclusions. So-called "post-publication peer review" has become rather controversial over the last few years as social media has facilitated the community's ability to discuss published science in an open forum. The scientific community becomes incensed in cases of fraud or plagiarism, although journals have often reacted more negatively to those who exposed the problems than to those responsible for the actual infractions. After consulting with several other experts on the underlying science and scientific publishing, I sent an email to the EIC of the journal detailing the problems with the papers. To be clear, I did not, and I am not making any accusation of misconduct.* Serious mistakes were made in the research that in part, or in whole invalidate the conclusions of the studies. Furthermore, it is troublesome that these issues were not addressed during the peer review process.

The EIC responded after several days promising to address each paper in a separate message. His response to the fluoride paper was the only course of action would be to submit a peer reviewed comment (providing this as an example) where the author would have a chance to respond, since the paper had been in print for nearly 2 years. I did not find this to be a particularly satisfying response since any publishable comment would need to be supported with new information. Essentially collecting the data and conducting control studies that should have been requested by peer reviewers. With the current state of knowledge, a potential published comment could be summarized as "I think you're wrong" and speculation, which is not any better than what was done in the paper. Data-free speculation does not not usually hold up to peer review, so reluctantly I published my criticism on pubpeer with the hope that anyone seeking to follow-up or use this chemistry would be wary of the authors' conclusions. This is the greatest risk in a case of erroneous research, which makes it similar to some issue encountered in cases of data falsification. Any time spent trying to use flawed science, wastes time and resources that could go toward more productive efforts. Coincidentally, I came across this paper today on fluoride sensing that invokes attack on a positively changed ring (albeit a different heterocycle), which was my initial instinct for an alternative mechanism in the disputed study. I'll have to compare the data and see if it gives me any new insight.

As for the iron paper, I never received the promised response. About 6 weeks after the initial inquiry, I wrote again asking for an update. After receiving nothing during the last +2 weeks, I am once again in the position of not having any particular recourse other than forgetting about it, or providing public post-publication peer review. Of the 2 papers, the iron sensor annoyed me more, because it repeats the same mistake made by another group that my group investigated and published as part of a larger study.

In a 2005 JACS article, Rurack claimed his Fe3+ responsive sensor also worked in aqueous solution. The only significant difference between the Rurack sensor and the new one is the receptor for the metal ion; however, both rely on the same PeT signal transduction mechanism where coordination to the aniline nitrogen atom is the key event. Several years after the Rurack paper was published, my group attempted to use the same receptor for a different application. We spent many months working on the system and eventually had to re-evaluate the Rurack data because our observations did not match what had been reported. We ultimately demonstrated that while Rurack's sensor did bind and respond to Fe3+ in organic solvents that lacked alcohol or carbonyl functional groups, the fluorescence signal in water was due to protonation of the aniline nitrogen atom, which mimics an Fe3+ response. We were fortunate to still be able to develop an interesting story even though we started our investigation with a flawed premise from a published paper (in JACS!). In a similar manner, the new Fe3+ sensor paper provides the opportunity to examine both pre- and post-publication peer review in this post.

The error the authors made in new paper is essentially the same  you absolutely cannot titrate Fe3+ into aqueous solution unless you use use a complexing ligand (e.g. citrate, which has a log K > 10 for Fe3+ and would therefore bind the metal ion more tightly than the sensor) or work at pH ≤ 3. In aqueous solution, Fe3+ rapidly hydrolyzes to Fe(OH)3 plus three equivalents of H+. The pKas for the 1st three deprotonation events are 2.2, 2.9 and ~6. This neglects multi-Fe processes that are also possible and have similarly acidic pKas. This is not an obscure fact, but textbook chemistry that would be covered in a standard undergraduate course. It's a fundamental concept in bioinorganic chemistry used to discuss the importance of siderophores in the acquisition of iron by microorganisms and the acquisition/transport/storage of iron by higher organisms using proteins.

The authors perform their fluorescence assays in 9:1 H2O-CH3CN. The water is not buffered and the authors do not report the pH at the beginning or the end of the fluorescence titration. A back of the envelope calculation shows that at ~150 equiv of Fe3+ added (~1000 μM Fe3+), the pH of the solution should reach ~3 (the approximate point at which Fe3+ hydrolysis no longer occurs and Fe3+ persists in solution). This is also the point at which the fluorescence response levels off for the sensor. The calculation is based on the use of deionized water without accounting for dissolved CO2, not that it would make a huge difference. CH3CN can act as a metal-binding ligand, but CH3CN does very little to stabilize Fe3+ in water. The authors go on to claim that reversibility of the fluorescence response with TPEN (tetrapyridylethylenediamine), a strong metal chelator, demonstrates that the fluorescence response is Fe3+-induced. This is also a flawed conclusion because the pKa of a protonated pyridine is about the same as a protonated aniline (both pKa ~ 5) and the aliphatic amines of TPEN are even more basic. Coordination (protonation) of the aniline nitrogen atom drives the PeT process responsible for the fluorescence response. TPEN can act as a base as well as a chelator, and the authors use a huge excesses of TPEN to demonstrate reversibility (60 equiv or 360 equiv of N bases). There may be some modest equilibrium between [Fe(Sensor)]3+ and [H(Sensor)]+ at pH<3, but the authors are observing a H+-, not an Fe3+-based fluorescence response in their titration. The argument is further supported by the receptor being used. Ethers are notoriously poor metal binding ligands in water. It would be very surprising if a cryptand ligand, even one with as many donor groups as the one in this sensor, could stable an Fe3+ complex with in water. The low affinity for the purported "[Fe(Sensor)]3+ complex" based on the fluorescence data supports this conclusion.

Returning to the Rurack report, the erroneous aqueous chemistry accounted for perhaps 25% of the discussion/conclusions in the paper. The rest of the Rurack paper is correctly interpreted (and an interesting counter-intuitive inorganic story). The flawed methodology accounts for 100% of the discussion/conclusions in the new paper. In my opinion this paper should be withdrawn; however, the EIC's lack of a response suggests that this isn't going to happen. I'd also like to contrast the response I got from the editor at Inorganic Chemistry to the one I got from ACS Applied Materials and Interfaces. Granted, I was trying to published a completed study in Inorg. Chem., but it should not be necessary to conduct months worth of work to prove something that only requires an understanding of basic inorganic chemistry. The Inorg. Chem. editor carefully handled the situation and facilitated a review of the conflicting data by the original author. I recall that I might have sent the EIC a note afterwards commending the editor who handled the situation. To his credit, Rurack completely supported the publication of our paper. The only unfortunate thing is that there is no indication anywhere that the original JACS paper contains a flawed set of experiments 
except in our paper. I still see this paper cited periodically as evidence that Fe3+ sensing can be done in aqueous solution. 

As an aside, our Inorg. Chem. paper was kind of an "end of the innocence" moment for me in scientific publishing. As a recent news story on pubpeer indicates, you can't always believe what you read in scientific journals. As someone who looks at sensor papers regularly, fluorescence might be the most misinterpreted spectroscopic assay used in chemistry. Fluorescence is an easy to execute and readily available technique, but there is a great tendency to interpret the results without acquiring additional supporting data (e.g. absorption spectroscopy, product analysis, etc.). As part of the Inorg. Chem. paper, my postdoc was trying to obtain a crystal structure of the Fe3+ complex of the ligand. After many failed attempts, he tried with Cu2+, and to our surprise got a structure containing only Cu1+. Another colleague in the fluorescent sensing field, pointed us in the direction of a synthetic paper where Cu2+ in CH3CN is used as a 1 electron oxidant (by forming [Cu(CH3CN)4]+). This was clearly what happened in our crystallization as we used CH3CN as the solvent. Despite our attempts to educate the fluorescent sensor community about the dangers of using Cu2+ in CH3CN, this is still a common practice. There are dozens of ring-opening spirolactam probes for Cu2+. These probes are usually based on rhodamine fluorophores, which have 2 embedded aniline groups, and give a terrific fluorescence response in CH3CN; however, everyone attributes the signal transduction mechanism to coordination/Lewis acidity of Cu2+, and ignores the possible contributions of redox chemistry.

There are clearly problems with peer review. In going through my significant backlog of TOC alerts today, I realized that RSC Advances moved to a 100 issues/year publishing schedule in 2014. This is a tremendous number of papers for a single journal, which requires a correspondingly huge number of peer reviewers. Inevitably, the same people who make mistakes in their research will be asked to review papers on similar science. This lets more errors slip through to publication and propagates errors in the literature. So what should editors do when they are alerted to a potentially serious problem? "Nothing" does not seem like an appropriate answer. Is pubpeer the the right pathway? It appears that some journals are very resistant to changing the status quo, and unless the pubpeer comments are linked to the articles, post-publication peer review may go unnoticed. It also remains to be seen if flawed protocols/conclusions will generate the kind of impassioned response that accompanies cases of scientific fraud; however, anyone who has wasted time with someone else's scientific mistakes surely has an opinion they will share without much prompting.

Update 3/22/15:
A note added in proof: the same group has a similar K+ sensor using a very similar receptor. The amides from the Fe3+ system are replaced by anilines, which would make metal interactions stronger. This sensor does not respond to Fe3+; however, the authors used buffered water, which means that protons from the are prevented from interacting with the aniline fluorescence switch. There is almost certainly an impact of having 3 "basic" aniline nitrogen atoms in the receptor as the amount of buffer (5 mM) is low compared to the added metal ions.

*The PI has since move to NJIT as a Dean, which would suggest the research group at UCF is no longer active. He is and was listed as a member of the editorial advisory board for ACS Applied Materials and Interfaces. While it is not uncommon and should not be problematic for board members to publish in the journal for which they serve, this may make some readers more suspicious about the impartiality of the peer review process.