The Editors-in-Chief of Molecular Cancer, a prominent open-access journal covering advances in cancer biology, have retracted a 2014 study that claimed to reveal how human epididymis protein 4 (HE4), working alongside the membrane protein Annexin II, drives the invasion and metastatic spread of ovarian cancer cells. The retraction note, published on 29 August 2026, closes a twelve-year saga for a paper that became embedded in the scientific literature on a protein many oncologists know best from blood tests used to monitor the disease. According to the notice, the decision followed post-publication concerns about highly similar images inside the article and striking parallels with figures in a second study written by a set of overlapping authors. With the authors declining to answer editorial queries, and two of them unreachable, the journal concluded it could no longer trust the data behind the paper’s central claims.

The retracted paper, “Human epididymis protein 4 in association with Annexin II promotes invasion and metastasis of ovarian cancer cells,” appeared in Molecular Cancer on 1 November 2014 as article 243 in the journal’s thirteenth volume. Its author list—Huiyu Zhuang, Mingzi Tan, Juanjuan Liu, Zhenhua Hu, Dawo Liu, Jian Gao, Liancheng Zhu and corresponding author Bei Lin—came from the Department of Obstetrics and Gynecology at Shengjing Hospital of China Medical University in Shenyang, in China’s northeastern Liaoning Province. The study argued that HE4 does not merely mark the presence of epithelial ovarian cancer in the bloodstream but actively participates in the disease’s progression, associating with Annexin II to boost the capacity of tumor cells to invade surrounding tissue and colonize distant sites. That framing mattered, because a biomarker that drives malignancy is not just a passive signal of disease but a potential therapeutic target, and the paper was read as evidence for exactly that possibility.

HE4, known formally as WFDC2, is a secreted glycoprotein belonging to the whey acidic protein (WAP) four-disulfide core domain family, a group of small proteins defined by a conserved motif locked into shape by four disulfide bonds. First characterized in the epithelium of the distal epididymis, the protein was later found to be strongly overexpressed in epithelial ovarian carcinoma, particularly the serous and endometrioid subtypes, and readily detectable in serum. Immunoassays for HE4 gained regulatory clearance in the late 2000s for monitoring women with epithelial ovarian cancer, and the marker entered everyday practice through composite tools such as the Risk of Ovarian Malignancy Algorithm, which combines HE4 with the older marker CA-125 to help evaluate pelvic masses. The protein’s normal physiological function remains debated, with proposed roles ranging from protease inhibition to antimicrobial defense along the reproductive tract. Precisely because HE4 is easy to measure and tracks closely with tumor burden, laboratories worldwide have asked whether it also acts inside the tumor itself—driving proliferation, adhesion, invasion and treatment resistance—rather than simply mirroring disease from a distance.

Annexin II, more commonly styled Annexin A2 or ANXA2 in the current literature, is a 36-kilodalton member of the annexin family of calcium-dependent phospholipid-binding proteins. It shuttles between the cytoplasm, nucleus and cell surface, where it frequently pairs with the S100A10 protein to form a heterotetrameric complex that organizes membrane trafficking, cytoskeletal remodeling and the assembly of plasmin-generating complexes on the outer face of the plasma membrane. Those functions place ANXA2 squarely within the machinery that cells deploy to migrate, degrade extracellular matrix and invade neighboring tissue—the processes that convert a localized tumor into metastatic disease. A reported physical association between HE4 and ANXA2 therefore offered an attractive, testable explanation for how an abundantly secreted biomarker could translate into invasive behavior. The 2014 paper built its case on experiments in which HE4 expression was manipulated in ovarian cancer cell lines, including Caov-3 cells, with figures presenting western blots, invasion readouts and time-lapse-style migration panels that appeared to show sharp differences between control and HE4-expressing cells.

Trouble surfaced long after publication. According to the retraction note, concerns were raised about highly similar images within the article and between it and another publication from shared authors. That second paper, published in Oncotarget in 2016 by Zhu and colleagues—several of whom also appeared on the retracted study—reported that overexpression of HE4 enhances the proliferation, invasion and metastasis of ovarian cancer. The retraction notice itemizes the correspondences: Figure 3D panels from the Molecular Cancer paper showing ANXA2 and HE4 immunoblot results in mock-transfected Caov-3 cells appear highly similar to Figure 1D panels in the Oncotarget paper showing HE4-labeled and mock-transfected Caov-3 samples. In plain terms, images that were supposed to document distinct experiments—different constructs, different conditions, in two different journals—look, to the editors, like the same underlying pictures recycled with new labels, a form of duplication that undermines the independence of both datasets.

The internal duplications are equally consequential. The notice states that in Figure 4D of the retracted paper, lane 1 at the zero-hour time point appears highly similar to lane 7 at the same time point, and that in Figure 4E, panel 1 at zero hours appears highly similar to panel 3. Figure 4F, which contains Caov-3 panels, shows pairs—1 and 4, 1 and 5, and again 1 and 4—that partially overlap. The zero-hour notation is the signature of wound-healing or transwell-style migration assays, in which researchers photograph the starting state of the assay and then track how quickly cells close a scratch or traverse a membrane over subsequent hours. Each condition in such an experiment begins with its own scratch in its own monolayer, so even the baseline photographs of different samples are expected to differ in texture, cell density and scratch geometry. When supposedly independent starting images turn out to be nearly identical, or two images partially overlap, the comparisons built on top of them lose their evidentiary footing: the apparent gap between treated and untreated cells, or between HE4-overexpressing and control cells, may be an artifact of reused photographs rather than genuine biology.

Under the publication-ethics guidelines that govern journals such as Molecular Cancer, editors are obliged to give authors the opportunity to respond to such concerns, whether by producing original, unmanipulated data or by explaining the similarities. The retraction note records that the authors were given that opportunity and did not respond to the editors’ correspondence about the retraction. The publisher could not contact Liancheng Zhu or Juanjuan Liu at all, while all other authors, including corresponding author Bei Lin, did not respond to the publisher regarding the retraction. Faced with that silence and with what the notice describes as a significant number of errors, the Editor-in-Chief concluded that confidence in the presented data could no longer be maintained, and the article was formally retracted more than a decade after it first appeared. The retraction note itself, published as Volume 25, article 207 of the journal, is open access and is now linked to the original paper in the journal’s record.

The case illustrates how modern research-integrity scrutiny operates. Image forensics—checks for duplicated panels, pixel-level overlay comparisons, contrast and brightness analyses, and screening tools that flag reused western blots across thousands of manuscripts—has become routine for large publishers, and readers increasingly report suspicious figures directly to journals. Editors can act years or even decades after publication because the images remain part of the permanent scholarly record, and retraction, far from erasing a paper, formally warns every future reader that its conclusions cannot be trusted. In this instance, the cross-paper duplication was especially damaging: overlapping figures in two venues from a shared author group suggested a pattern spanning manuscripts and journals rather than a one-off production error. The retraction notice identifies the 2016 Oncotarget paper by its own DOI, but the formal action concerns the 2014 Molecular Cancer article, whose record now carries the retraction through the journal’s Crossmark article-versioning service, ensuring that readers who land on the original page will see the notice attached.

For clinicians and researchers, the retraction draws a bright line between two distinct strands of HE4 science. The protein’s standing as a serum biomarker rests on large, independent clinical studies and regulatory evaluations that do not depend on the retracted mechanistic work, and the retraction does not touch that literature. What is withdrawn is the specific claim, as presented in the 2014 paper, that HE4’s association with Annexin II promotes invasion and metastasis of ovarian cancer cells. Research groups will now have to weigh how much of the mechanistic edifice they constructed around that finding needs re-examination, and authors who cited the paper—many of whom did so to support HE4’s proposed pro-invasive role—will need to update their reference lists. Retractions of this kind also serve a quieter function: they signal to early-career scientists that image reuse is detectable, that it tends to surface eventually, and that silence in the face of editorial queries usually ends in retraction rather than clarification.

The retraction note is deliberately spare: seven flagged image pairs, two unreachable authors, a decision that the data can no longer be trusted, and no author rebuttal on the record. A paper published at the height of the biomarker-discovery era, from a hospital department deeply engaged in ovarian cancer research, fell to a pattern of duplicated images that survived peer review, production and years of citation. The episode is a reminder that in modern science publication is not the end of scrutiny but the beginning of it, and that the credibility of a biomarker’s mechanistic story depends on the integrity of every panel in every figure. For readers who encounter HE4 in the clinic or the laboratory, the protein itself remains a legitimate object of study; the cautionary tale now sits, permanently attached, in the bibliographic record of Molecular Cancer.