A Perennial Northern Blot

This post was earlier cross-posted at Leonid Schneider's site, hence the nonfrivolity and Explaining Voice. The version there is improved by Leonid's editing and frame-story.

The title of this post refers to the famously picaresque Western blot belonging to a Brazilian diabetes researcher. In its protean versatility, Saad's pentadecaplicating blot could transform itself into any protein -- tubulin, actin, GLUT4, IRS1 -- from any combination of source conditions. It thereby appeared in at least 15 versions, spread across 10 papers in "an intricate publishing web", serving as the loading control in that many different experiments (that is, as a measure of the total level of extracted protein, for normalising the measurements of the protein of interest). In my imagination it spoke with the voice of Robin Williams. This site forwarded a report on the Wandering Western... the ensuing saga included editorial Expressions of Concern, lawsuits, an investigation by Saad's university that saw no evidence of misconduct, and 13 retractions so far (RetractionWatch are keeping score).

Baldanders: the spirit animal of shapeshifting blots

The present case also concerns re-use of a loading control, but this time featuring a Northern blot. The compass-point tradition for naming gel-electrophoresis techniques began with Sir Edwin Southern, pioneer of Southern blotting, for this is how humor works in molecular biology. It has been explained to me that Northern blots do not directly measure the popularty of a protein in the cellular economy; instead, mRNA (encoding for a protein) is the chemical species, extracted from various sources (lanes), and spread out into bands according to molecular weight. Then transferred (blotted) from the electrophoresis gel to a filter for stability, and detected by inducing the mRNA to bind to a matching and radiotracing DNA probe.

So in this case, a team of researchers have a bank of 28 "cell smoothies": two sets of eight tissue types, one set of eight cancer-cell lines, and four fetal-tissue samples. In a series of papers published over a decade, the team have characterised numerous proteins from within the self-organising complexity of the human cell -- sequencing the DNA for each protein and specifying its chromosomal location, describing its role within that complexity, and checking which tissues express it (which depends on which genes remained active in each lineage of cells that differentiated and specialised and became a tissue). That is to say, the Northern Blots were just one aspect of the papers, and they are all outside my comfort grade and above my pay zone.

Each study took a few drops from the stored samples, blotted it ("Filters containing about 2 μg of polyadenylated RNAs from the indicated human tissues"), and probed for the mRNA of choice. But there are limits to the precision that a pipette can provide -- even in the hands of a trained gene-modified laboratory monkey -- so the final stage is to wash the probe DNA out of the filter and probe it again for Actin (a background "housekeeping" protein, required by cells to maintain their architecture, unless they are dead) to correct for the actual aliquots that were used. Thus papers in this sequence typically include a phrase along these lines:

Filters were subsequently hybridized with a human actin probe to ascertain differences in RNA loading.

It is conceivable, however, that this phrase was repeated from the first paper, along with the loading blot itself. Comparing 23 papers, there appears to be one original blot for each bank: four blots, which are variously compressed and clipped according to the exigencies of publication, and varying also in exposure, rather than a separate measurement after each separate exercise in tissue localisation. The sources are 'Zebedee', commenting on threads at this site; anonymous contributors to Pubpeer threads; Elizabeth Bik; and myself.

This comes to our notice because a 23-fold replication beats the 15-fold record of the Brazilian wanderer. Crucially, though the possible copies are consistently identified as Actin, and the authors have tried to label the sources of the lanes consistently. The reuse of a 'loading library' is deprecated, but this does not begin to approach the problematic level of the Brazilian Western: there was no attempt to mislead (other than the claim that the control in each study was specific to it, made subsequently to the data to be controlled). It is a perennial blot, always in the same place, rather than a wanderer or vagrant.

Regrettably, the labelling of lanes was not as consistent as was intended. In a 2003 appearance of the fetal-tissue blot, it was flipped horizontally relative to the lane labels, as marked with a red box in the Figures. Note that in some publications the lanes are listed in reverse order -- from Leucocytes to Heart rather than vice versa -- and in the Figures I have flipped each band and labels in such cases, to keep a single sequence of tissue types (hence the mirror-image text in places).

Red boxes were also necessary in some cases where the cancer-cell blot was flipped relative to its lane labels, and for the #2 array of tissue cells in the 1999 paper. In addition, that blot was rotated through 180° from 2001 onwards (so that the Actin background for Thymus cells becomes that of Colon cells, and vice versa, while Testes and Ovary change places, and Spleen with Leucocytes). This is marked with orange boxes. One can only hope that these pictorial labelling issues did not extend into the measurements of Actin from the blots, as used in the quantitative results.

Finally, two blue arrows mark the omission of 'Pancreas' and 'Skeletal muscle' from one study each, with the loading band spliced to remove that lane.

I am going to play 'good cop' here, and propose that the corner-cutting absence of study-specific controls probably made little difference to the results. Corrigenda to the paper acknowledging the use of archival controls would be appropriate (along with correction of any flipped and rotated bands). Other issues have been raised about other figures in some of the papers, but I do not address those here.

Details of the 23 publications follow. We are still hopeful of finding a few more examples of the Perennial Northern Blot in order to raise the number to a round two dozen.

1 1994. "Human cathepsin O. Molecular cloning from a breast carcinoma, production of the active enzyme in Escherichia coli, and expression analysis in human tissues", Velasco et al; J Biol Chem., 269(43):27136-42.

2 1995. "Cloning and expression analysis of a novel human serine hydrolase with sequence similarity to prokaryotic enzymes involved in the degradation of aromatic compounds", Puente & López-Otín; Journal of Biological Chemistry 270, 12926-12932. DOI 10.1074/jbc.270.21.12926 Figure 5.

3 1996. "Cloning and Expression Analysis of Human Bleomycin Hydrolase, a Cysteine Proteinase Involved in Chemotherapy Resistance", Ferrando et al.; Cancer Research 56: 1746-1750. PMID: 8620487

4 1996. "Molecular Cloning of a Novel Membrane-type Matrix Metalloproteinase from a Human Breast Carcinoma", Puente et al; Cancer Research 56:944-949.

5 1997. "Identification and characterization of a novel human matrix metalloproteinase with unique structural characteristics, chromosomal location, and tissue distribution", Pendás et al; J Biol Chem. 272(7):4281-6. doi: 10.1074/jbc.272.7.4281 Figure 7.

6 1998. "Cathepsin L2, a Novel Human Cysteine Proteinase Produced by Breast and Colorectal Carcinomas", Santamaría et al; Cancer Res. 58(8):1624-30.

7 1998. "Cathepsin Z, a novel human cysteine proteinase with a short propeptide domain and a unique chromosomal location", Santamaría et al; J Biol Chem. 273(27):16816-23. doi: 10.1074/jbc.273.27.16816 Figure 5.

8 1999. "Cloning and characterization of human MMP-23, a new matrix metalloproteinase predominantly expressed in reproductive tissues and lacking conserved domains in other family members", Velasco et al; J Biol Chem. 274(8):4570-6. doi: 10.1074/jbc.274.8.4570 Figure 6.

9 1999. "Molecular cloning and structural and functional characterization of human cathepsin F, a new cysteine proteinase of the papain family with a long propeptide domain", Santamaría et al; J Biol Chem. 274(20):13800-9. doi: 10.1074/jbc.274.20.13800 Figure 6.

10 1999. "Identification and Chromosomal Location of Two Human Genes Encoding Enzymes Potentially Involved in Proteolytic Maturation of Farnesylated Proteins", Freije et al; Genomics 58, 270–280. DOI: 10.1006/geno.1999.5834

11 2000. "Human MT6-matrix metalloproteinase: identification, progelatinase A activation, and expression in brain tumors", Velasco et al; Cancer Research 60, 877–882. pubmed: 10706098

12 2001. "Identification, Characterization, and Intracellular Processing of ADAM-TS12, a Novel Human Disintegrin with a Complex Structural Organization Involving Multiple Thrombospondin-1 Repeats", Cal et al; Journal of Biological Chemistry 276, 17932-17940. doi: 10.1074/jbc.M100534200 Figure 5.

13 2002. "Matriptase-2, a Membrane-bound Mosaic Serine Proteinase Predominantly Expressed in Human Liver and Showing Degrading Activity against Extracellular Matrix Proteins", Velasco et al; J Biol Chem. 277(40):37637-46. doi: 10.1074/jbc.M203007200 Figure 8.

14 2002 "Cloning, expression analysis, and structural characterization of seven novel human ADAMTSs, a family of metalloproteinases with disintegrin and thrombospondin-1 domains", Cal et al; Gene 283 49-62. doi: 10.1016/S0378-1119(01)00861-7

15 2003. "Polyserase-I, a human polyprotease with the ability to generate independent serine protease domains from a single translation product", Cal et al; PNAS 100(16): 9185–9190. doi: 10.1073/pnas.1633392100 Figure 4.

16 2003. "Human Autophagins, a Family of Cysteine Proteinases Potentially Implicated in Cell Degradation by Autophagy", Mariño et al; Journal of Biological Chemistry 278, 3671-3678. doi: 10.1074/jbc.M208247200 Figure 3.

17 2003. "Identification and Characterization of ADAMTS-20 Defines a Novel Subfamily of Metalloproteinases-Disintegrins with Multiple Thrombospondin-1 Repeats and a Unique GON Domain", Llamazares et al; Journal of Biological Chemistry 278(15):13382-13389. doi: 10.1074/jbc.M211900200 Figure 4.

18 2004. "Identification and Characterization of Human and Mouse Ovastacin", Quesada et al; JBC 279 (25) 26627-26634. doi: 10.1074/jbc.M401588200 Figure 3.

19 2004. "Cloning and enzymatic analysis of 22 novel human ubiquitin-specific proteases", Queseda et al; Biochemical and Biophysical Research Communications 314, 54-62. doi: 10.1016/j.bbrc.2003.12.050

20 2005. "Identification of Human Aminopeptidase O, a Novel Metalloprotease with Structural Similarity to Aminopeptidase B and Leukotriene A4 Hydrolase", Díaz-Perales et al; Journal of Biological Chemistry 280, 14310-14317. doi: 10.1074/jbc.M413222200 Figure 4.

21 2005. "Human Polyserase-2, a Novel Enzyme with Three Tandem Serine Protease Domains in a Single Polypeptide Chain", Cal et al; JBC 280, 1953-1961. doi: 10.1074/jbc.M409139200 Figure 3.

22 2005. "Identification and Characterization of Human Archaemetzincin-1 and -2, Two Novel Members of a Family of Metalloproteases Widely Distributed in Archaea", Diaz-Perales et al; JBC 280(34):30367-30375. doi: 10.1074/jbc.M504533200 Figure 4.

23 2006. "Identification and characterization of human polyserase-3, a novel protein with tandem serine-protease domains in the same polypeptide chain", Cal et al; BMC Biochemistry. doi: 10.1186/1471-2091-7-9 Figure 7.