Our diabetes research has asked more than how to lower blood glucose: why is diabetes difficult to reverse, and are there cells that sustain the disease?

The current research concept of diabetes stem cells (DSCs) grew out of studies of diabetes and bone marrow-derived cells dating back to 2003. The term DSCs describes a research concept, not an established clinical diagnosis.

Kojima H, Fujimiya M, Matsumura K, et al.
Nature Medicine. 2003;9:596–603.
DOI: 10.1038/nm867

Investigating islet formation in the liver and diabetes remission

This study used gene therapy with NeuroD and betacellulin to generate insulin-producing cells and islet-like structures in the livers of STZ-induced diabetic mice, improving their diabetic condition.

During the work, unusual proinsulin-positive cells were also observed in tissues from untreated diabetic mice.

That observation led to subsequent research into abnormal bone marrow-derived cells and diabetes and became one of the starting points for today’s diabetes stem cell research.

Katagi M, Terashima T, Ohashi N, et al.
Communications Biology. 2021;4:575.
DOI: 10.1038/s42003-021-02082-5

Tracing diabetic neuropathy to abnormal hematopoietic stem cells

This study reported that CD106 (VCAM-1)-positive short-term hematopoietic stem cells (ST-HSCs) are involved in diabetic neuropathy in mouse models.

When HSCs or CD106-positive ST-HSCs from diabetic mice were transplanted into nondiabetic mice, the abnormal cells fused with neurons and caused nerve dysfunction. Transient hyperglycemia also induced similar pathological cell fusion and neuropathy.

The findings support the idea that “abnormal cells supplied by the bone marrow, as well as damaged nerves themselves, contribute to the disease process”—a key finding supporting current diabetes stem cell research.

Kojima H, Katagi M, Okano J, et al.
Communications Biology. 2023;6:637.
DOI: 10.1038/s42003-023-05010-x

Normal blood glucose persisted after treatment ended in diabetic mice

This study combined insulin-based glucose control with a course of the HDAC inhibitor givinostat in STZ-induced diabetic mice.

The paper reported sustained normal blood glucose and a recovery of endogenous insulin secretion after treatment ended, along with reduced cell fusion associated with abnormal bone marrow-derived cells. Removing the thymus abolished the protective effect of treatment, providing evidence for a disease model linking bone marrow, pancreatic beta cells, and immune and thymic function.

The findings underpin the current pipeline:

• Diabetes Remission Therapy Candidate
• Diabetes Stem Cell Biomarker
• Biozipcode™ Cell-Targeted Diabetes Therapy Candidate

This study is a core preclinical foundation for these programs.

Itsuko Miyazawa, Hideto Kojima et al.
Practical Laboratory Medicine. 2026;51:e00548.
DOI: 10.1016/j.plabm.2026.e00548

From diabetes stem cell research to a peripheral blood biomarker in humans.

Earlier studies in diabetic mouse models investigated whether abnormal hematopoietic stem and progenitor cell populations expressing CD106, TNF-α, proinsulin, and other markers may contribute to diabetes and its complications.

This study extended that research hypothesis to humans by analyzing CD34-positive cells in peripheral blood from people with type 2 diabetes. Comparing 12 people with type 2 diabetes and 10 healthy participants, the researchers reported significantly higher proportions of CD106⁺, TNFα⁺, proinsulin⁺, CD34⁺ cells in the diabetes group.

The proportion of these cells also showed a significant positive correlation with HbA1c, suggesting a link with longer-term blood glucose levels.

The findings point to the potential to assess abnormal cell states associated with diabetes using peripheral blood, offering a perspective beyond conventional blood glucose and HbA1c measurements.

They provide part of the scientific basis for the Diabetes Stem Cell Biomarker program being developed by Biozipcode, Inc. and for future companion diagnostic research.

The paper also notes that establishing the clinical utility of these cells will require further validation in larger patient cohorts.

The scientific basis for Biozipcode™ grew from the discovery of homing peptides that bind to specific cells and tissues, followed by research into their use for drug and gene delivery.

Much of the following work predates the name Biozipcode™, but it directly underpins the present concept of using peptides to recognize target cells and deliver treatment.

Oi J, Terashima T, Kojima H, et al.
Neuroscience Letters. 2008;434:266–272.
DOI: 10.1016/j.neulet.2008.01.062

Finding seven-amino-acid peptides that recognize specific neurons

Using an M13 phage seven-mer peptide library, this study sought peptides that bind to dorsal root ganglion (DRG) neurons in mice.

Several DRG-homing peptides were identified and shown to recognize and be taken up by DRG neurons both in vitro and in mice.

“Finding short peptides that recognize specific cells” is a central idea behind today’s Biozipcode™ research.

Terashima T, Oka K, Kritz AB, et al.
Journal of Clinical Investigation. 2009;119:2100–2112.
DOI: 10.1172/JCI39038

From finding a target to delivering a gene to it

This work investigated selective gene delivery to sensory neurons by incorporating DRG-homing peptides into the surface of adenoviral vectors.

In mice, the targeted helper-dependent adenovirus substantially increased gene delivery to DRG neurons. Delivery of a therapeutic gene in a Sandhoff disease model was reported to improve neurological function.

It demonstrated a strategy combining

Targeting Peptide
＋ Vector
＋ Therapeutic Gene

that informs current cell-targeted drug delivery research.

Terashima T, Ogawa N, Nakae Y, et al.
Molecular Therapy – Nucleic Acids. 2018;11:203–215.
DOI: 10.1016/j.omtn.2018.02.007

Applying cell-targeting peptides to siRNA delivery

Using phage display, this study identified homing peptides that bind to astrocytes and microglia.

Combining the microglia-targeting peptide MG1 with siRNA against IRF5 and administering it in a neuropathic pain model was reported to reduce pain through gene modulation in the target cells.

It is an example of the application Biozipcode™ aims to pursue: recognizing specific cells and delivering nucleic acid medicines to them.

Ogawa N, Terashima T, Oka K, et al.
PAIN Reports. 2018;3:e695.
DOI: 10.1097/PR9.0000000000000695

Treating pain with DRG-targeted vectors

This study incorporated DRG-homing peptides into a helper-dependent adenovirus carrying the GAD67 gene and evaluated its effects in a neuropathic pain model.

The DRG-targeted vector produced efficient gene expression at a lower dose than an untargeted vector and reduced allodynia in mice.

Wada A, Terashima T, Kageyama S, et al.
Molecular Therapy – Oncolytics. 2019;12:138–146.
DOI: 10.1016/j.omto.2019.01.001

Identifying peptides that bind to prostate cancer while sparing healthy tissue

Using in vivo phage display, this study sought peptides that bind selectively to LNCaP prostate cancer cells while reducing binding to healthy tissue.

Candidate peptide LN1 bound to tumor tissue, while LN1-KLA, formed by linking LN1 to a cytotoxic peptide, suppressed tumor growth in vitro and in tumor-bearing mice.

This work provides an important scientific proof of concept for the current cancer cell-targeted therapy pipeline.

Terashima T, Kojima H, et al.
Molecular Therapy – Methods & Clinical Development. 2019;13:474–483.
DOI: 10.1016/j.omtm.2019.04.008

Delivering genes to the spinal cord through systemic circulation

In vivo biopanning identified two candidate peptides, SP1 and SP2, that target the spinal cord.

When the peptides were combined with plasmid vectors and given intravenously, gene expression was observed in the spinal cord while expression in the brain, liver, and kidneys was limited. IL-4 gene delivery with SP1 also showed an effect in a pain model.

The findings inform Biozipcode™’s future drug delivery concept of delivering treatment to target organs through systemic circulation, rather than by local administration alone.

Okano J, Kojima H, et al.
bioRxiv. 2026.
DOI: 10.64898/2026.04.16.719084

Revealing patterns of peptide selection hidden in whole-organ analysis

Conventional in vivo phage display has often involved recovering phage from an entire organ and identifying the sequences enriched there.

This study combines laser capture microdissection (LCM) and high-throughput sequencing to analyze peptide sequences separately in vascular compartments and parenchymal tissue.

The results suggest that peptide selection is shaped by structural constraints associated with organ microenvironments, rather than occurring at random. They may help reveal targeting principles that whole-organ analyses can miss.

We view this research as advancing Biozipcode™ from “finding sequences that bind to an organ” to “analyzing which compartments and cellular environments within an organ they bind to”.

This paper was posted on bioRxiv as a preprint (not peer reviewed).