Non-invasive imaging of CD4+ T lymphocytes holds great promise for enhancing our understanding of the critical role this immune cell subset plays in disease (e.g., HIV/AIDS, inflammatory diseases) or in response to therapy, such as vaccine development, transplantation/graft-versus-host disease, and immunotherapy for cancer. The increased interest in visualizing the spatiotemporal in vivo biodistribution of T cells, including trafficking, persistence, or expansion at tumor sites and activation or exhaustion phenotype, is reflected in the plethora of imaging modalities and targets being explored in preclinical and clinical trials, as reviewed in [21]. A humanized antibody fragment targeting human CD4+ T cells (IAB41 minibody, ImaginAb) is currently being explored for imaging in humanized mice and clinical translation [22, 23]. Another approach utilizes anti-human CD4 single-domain antibodies and shows the accumulation of 64Cu-radiolabeled CD4-Nb1 in lymphoid tissues of a human CD4 knock-in mouse model [24]. However, for preclinical cancer immunology research and evaluation of response to immunotherapies, syngeneic tumors transplanted into immunocompetent mice remain readily available, reproducible, and widely used models that require mouse-specific imaging tracers [25, 26].

Previous proof-of-principle studies using the anti-CD4 cys-diabody GK1.5 cDb for 89Zr-immunoPET imaging confirmed the ability to detect CD4 + T cells in the lymphoid tissues (spleen and lymph nodes) of immunocompetent mice, without significant impact on the function or proliferation of the target cells [13]. Furthermore, 89Zr-GK1.5 cDb was successfully used to visualize the reconstitution of CD4+ T cells after hematopoietic stem cell transplantation and in a model of inflammatory bowel disease [11, 13, 15]. Wider use of anti-CD4 immunoPET imaging could be facilitated by overcoming the very low expression levels of the GK1.5 cDb.

GK1.5 cDb was originally cloned based on the genetic information encoding the variable domains obtained from a hybridoma (ATCC TIB-207, GK1.5, rat IgG2b,K) and assembled into a functional cys-diabody (dimer of scFv). However, the use of degenerate primers can introduce mutations at the start and end of the sequence, which might impact expression and folding and result in low protein yields [27]. While CDR-grafting is more commonly used to humanize antibodies [28], we hypothesized that transplanting the antigen binding site onto a rodent framework with known higher expression levels could overcome the unsatisfactory protein expression yields of GK1.5 cDb.

The six CDR loops of GK1.5, based on the contact definition (i.e., residues that take part in interactions with antigen and are part of the canonical sequence for loop structure), were grafted onto the variable domain framework of the rat anti-CD8 hybridoma YTS169 [11, 20]. CDR-grafting resulted in a rodent cys-diabody (GK1.5 FR cDb) with retained specificity and affinity for murine CD4+ T cells and significantly improved expression yields.

One unusual sequence feature was identified in the variable light chain domain framework (residue 80, Kabat L81N), constituting a potential N-glycosylation site. Deglycosylation using PNGase F confirmed that the recombinant protein, produced in a mammalian expression system, is indeed glycosylated. N-linked glycans in the variable domains, so-called Fab glycans, are a product of somatic hypermutation and antibody diversification. Fab N-glycosylation function and effects are less understood than Fc-glycosylation, but variable domain glycans are reported to impact antigen–antibody binding, specificity, affinity, and antibody stability [29,30,31,32]. Furthermore, glycan-binding receptors in the liver can also significantly affect blood clearance and catabolism of glycoproteins [33, 34]. The glycosylated GK1.5 FR cDb exhibited unexpectedly rapid clearance from the blood and accumulation in the liver that is nontypical for a 50 kDa cys-diabody. Investigating the Fab glycan composition and the impact of specific liver glycan receptors on the clearance exceeded the scope of this study, but could inform future antibody modification. Genetic aglycosylation resulted in the variant GK1.5 N80D cDb and restored renal clearance and plasma half-life similar to that reported for other diabodies (2–5 h) [35,36,37]. The [89Zr]Zr-DFO is site-specifically conjugated to the C-terminal cysteine, which is likely reabsorbed in the kidney’s proximal tubules, leading to retention of the residualizing radiometal. These results corroborate the previous observation that the anti-CD8 minibody ([64Cu]Cu-NOTA-YTS169 Mb), which contains the N-glycosylation, exhibited rapid clearance from the blood. At the same time, the aglycosylated diabody ([89Zr]Zr-malDFO-169 N85D cDb) showed renal clearance at the expected rate [17, 18].

The novel CDR-grafted anti-CD4 cDb variants retained low-nanomolar affinity (3.0 ± 1.0 nM) comparable to the previously published GK1.5 cDb (2.7 ± 0.2 nM) and the parental GK1.5 IgG (1.1 ± 0.06 nM) [13]. Furthermore, flow cytometry analysis of mouse splenocytes showed that GK1.5 N80D cDb stained the same fraction of CD4+ T cells as the parental GK1.5 IgG. Importantly, these findings confirmed that CDR grafting did not affect affinity and specificity.

Monitoring CD4+ T cells over time would benefit from the capability to conduct serial immunoPET imaging in the same animal. We have not tested the immunogenicity of the rat variable domain framework in mice or its impact on the CD4 T cell population; however, further ‘’murinization” may be required. Another limitation of the anti-CD4 cys-diabody is its inability to distinguish between CD4+ T cell subsets (e.g., immunosuppressive Tregs or proinflammatory Th1 T cells) and other immune cell subsets that can express CD4, such as dendritic cells [38]. Combinations of distinct biomarkers could be explored for a more comprehensive in vivo phenotyping of the immune tumor microenvironment.

The 89Zr-labeled GK1.5 N80D cDb, with its rapid blood clearance at low protein doses, yielded high-contrast PET images as early as 2–4 h p.i. In this model (non-tumor bearing C57BL/6J), 89Zr-GK1.5 N80D cDb reached higher uptake in lymphoid tissues than previously reported for the mCD4-Mb minibody (IAB46M2-18) [26]. The low background in normal tissues, except for the kidney (organ of clearance), could facilitate the detection of tumor-infiltrating CD4 + T lymphocytes in tissues that often exhibit a higher background, such as the liver. Furthermore, the cys-diabody format would be suitable for radiolabeling with shorter-lived radionuclides, such as fluorine-18 (18F). We have previously shown that cys-diabodies can be site-specifically or randomly 18F-radiolabeled for same-day imaging (4 to 8 h p.i.) without impacting their biodistribution or pharmacokinetics [39, 40]. An 18F-labeled anti-CD4 cDb would further reduce the absorbed radiation dose, which could be crucial for serial imaging of radiosensitive CD4+ T cells [41]. Quantitative image analysis could provide a more comprehensive profile of the pharmacokinetics and biodistribution of immunoPET tracers, especially at earlier time points. However, limitations such as partial volume effects and the proximity of tissues with high signal retention need to be considered[42]. The integration of artificial intelligence and deep learning for PET image reconstruction and analysis holds promise for significant improvements[43].

In this study, we have re-engineered an anti-mouse CD4-specific cys-diabody (GK1.5 N80D cDb) that can be produced with sufficient yields for imaging studies in murine models of disease and response to therapy, enabling rapid and non-invasive monitoring and quantification of CD4 + T Cells. The cys-diabody format allows site-specific radiolabeling, provides a good compromise between rapid clearance, effective targeting, and reduced radiation dose, and is biologically inert (no Fc region). 89Zr-GK1.5 N80D cDb immunoPET imaging in preclinical disease models has the potential to guide the development and translation of anti-CD4 immunoPET imaging and other CD4-targeted immunotherapies.