Proof of Concept

Content

Our project, "Yeast Medics," aims to develop an intelligent living dressing for chronic diabetic wounds. This report provides systematic proof-of-concept across four core aspects: chassis organism robustness, therapeutic system functionality, intelligent regulation, and encapsulation delivery. The experimental results confirm the successful construction of engineered Yarrowia lipolytica strains capable of expressing key therapeutic factors (Pexiganan, IL-4, VEGF) with demonstrated biological activity. Furthermore, the smart L-HBC hydrogel achieved efficient yeast encapsulation and exhibited designed responsiveness, collectively verifying the feasibility of our integrated therapeutic strategy.

1. Validation of Chassis Organism Robustness

We selected the Yarrowia lipolytica Po1h strain as our chassis organism. Certified as GRAS (Generally Recognized as Safe), this yeast exhibits exceptional tolerance, allowing it to thrive in the high-sugar, high-oxidative-stress microenvironment typical of diabetic wounds (Park & Ledesma-Amaro, 2023; Sanya & Onésime, 2022). Notably, it utilizes glucose from wound exudate as a carbon source, thereby ensuring in-situ survival and functionality. Its protease-deficient genotype (ΔAEP, ΔAXP) minimizes degradation of heterologous therapeutic proteins, providing a solid foundation for functional expression (Jaafar & Zueco, 2004).

The successful genomic integration of all four target genes—Pexiganan, VEGF, IL-4, and IL-10—was confirmed by PCR and agarose gel electrophoresis (Img.1). Despite the failure of the IL-10 strain to express detectable protein, which led to its exclusion from functional studies, the construction of the other three engineered strains was successful.

2. Modular Functional Validation of the Therapy System

We adopted a modular approach to validate the three core therapeutic functions step by step.

2.1 Constitutive Antimicrobial Activity via Surface Display:

The antimicrobial peptide Pexiganan was successfully designed for cell surface display. The Oxford cup assay showed that the engineered yeast culture produced a clear inhibition zone of approximately 1.6 cm in diameter against both Staphylococcus aureus and Escherichia coli (Img.2). This confirms the functional surface display of Pexiganan, providing a foundational, constitutively active antimicrobial barrier.

2.2 Glucose-Inducible Anti-inflammatory Activity:

The strain engineered for IL-4 secretion was successfully constructed. IL-4’s potent bioactivity was unequivocally demonstrated in a macrophage polarization assay. Flow cytometric analysis revealed that the crude protein secreted by the engineered yeast significantly promoted the polarization of RAW264.7 macrophages towards the anti-inflammatory M2 phenotype, reducing the M1/M2 ratio to below 1%—an effect comparable to commercial IL-4 (Img.3). This validates the potential of this module to dynamically alleviate wound inflammation in response to high glucose.

2.3 Heat-Inducible Pro-angiogenic Activity:

For the VEGF module, while direct protein detection was challenging, its functionality was confirmed through a cell proliferation assay. Treatment of Human Umbilical Vein Endothelial Cells (HUVECs) with the yeast-secreted crude protein resulted in a significantly higher cell viability (OD450) compared to the blank control group (p < 0.01) (Img.4), indicating the secretion of bioactive VEGF with pro-proliferative activity.

3. Validation of Intelligent Regulation and Encapsulation

3.1 Signal-Responsive Promoter Screening:

To achieve dynamic control, we identified native promoters responsive to key wound signals. qRT-PCR analysis demonstrated that the PGK promoter was most strongly induced under high glucose (2.3-fold upregulation, p < 0.05; Img.5), while the HSP90 promoter showed exceptional thermal responsiveness, with induction exceeding 700-fold at 39°C (Img.6). The functionality of both promoters was further confirmed by driving GFP expression, visually demonstrating their potential for programming therapeutic responses.

3.2 Visual Confirmation of Promoter Activity via GFP Reporter

To visually confirm the functionality of the selected PGK and HSP90 promoters beyond transcriptional-level analysis, we constructed corresponding GFP reporter cassettes (pPGK-GFP and pHSP90-GFP). Fluorescence microscopy analysis provided clear and direct evidence of promoter activity. Under high-glucose conditions, engineered yeast harboring the pPGK-GFP plasmid exhibited intense green fluorescence, which was absent in the wild-type control and significantly stronger than in glycerol-cultured counterparts (Img.7). Similarly, upon thermal induction at 39°C, transformants with the pHSP90-GFP plasmid displayed a strong fluorescent signal, starkly contrasting with the negligible background of the non-induced wild-type control (Img. 8). These results offer visual and spatial confirmation that both promoters are effectively activated by their respective stimuli and are capable of driving robust heterologous gene expression, thereby validating their core role in our designed intelligent genetic circuits.

3.3 Efficient and Safe Hydrogel Encapsulation:

We developed a thermosensitive L-DOPA-modified hydroxybutyl chitosan (L-HBC) hydrogel. Rheological characterization determined its gelation temperature to be 20.3°C, ensuring a rapid liquid-to-gel transition upon application to wounds (37°C) (Img.9). Encapsulation leakage assays established a safe therapeutic window of at least 24-48 hours, confirming excellent containment with zero leakage observed within the first 24 hours and minimal leakage (4 colonies) by 48 hours (Img.10). This result ensures effective biocontainment for practical application.

4. Proof of Concept (PoC): Fine-Grained, Stable Manipulator Operation

4.1 Research Background & Objective

This study examines whether a compact, cost-constrained 6-DoF desktop robotic arm can achieve a steady, repeatable stop with millimeter-level end-point accuracy—satisfying use cases such as drug/gel dispensing that demand “stable touchdown and minimal vibration.” We compare two controller configurations: the legacy setup (T104) and a new setup (T102/T02) featuring time-parameterization and IK-continuity optimization. Test data include paired symmetric targets for end-point accuracy logging and high-rate Cartesian/Joint telemetry for smoothness analysis. Unless otherwise noted, distances are in millimeters (mm) and time in seconds (s).

4.2 Data Processing & Evaluation Perspective

Rather than reporting error statistics alone, we focus on how the motion process itself generates error and jitter. From measured trajectories, we reconstruct velocity, acceleration, and jerk (the third derivative of position), and quantify residual motion within the trailing 0.5-s window before the controller declares “motion complete.” Pairing these kinematic quantities with end-point error enables a fuller judgment of whether the new method lowers both the numerical end-point error and the dynamic intensity that causes oscillation.

4.3 Hardware & Test Scene

We use a simple setup: a 6-axis desktop manipulator operating in an open workspace with a fixed tool frame. Prior to testing, we verify mounting and emergency-stop reachability and warm up the joints to stabilize friction. If vision is involved, we first complete camera–robot extrinsic calibration; we then perform TCP calibration via the 3-point method or hole-centering and record the worktable plane to ensure coordinate consistency across runs. Logged signals include timestamps, Cartesian x/y/z, joint angles, and load/torque, sampled at the controller’s native rate.

4.4 Data Acquisition & Sampling Characteristics

Target pairs are laterally symmetric, so the arm traverses the workspace and returns along mirrored paths. The legacy method provides 9 trajectory pairs; the new method provides 3 pairs. Their intersection {1, 2, 3} is used for pairwise comparison. Sampling characteristics are similar: the legacy method has a median sampling rate of ~17.21 Hz and a typical per-pair duration of ~11 s; the new method is ~16.96 Hz with ~7.1 s duration. Each segment clearly exhibits three phases—launch, cruise, and profiled (shaped) deceleration—allowing stable computation of kinematic norms without heavy re-filtering.

4.5 Positioning Accuracy Results

We compute the Euclidean distance between commanded and actual end-points. The new method’s distribution is sharply concentrated: MAE ≈ 1.62 mm, RMSE ≈ 1.70 mm, 95th percentile ≈ 2.31 mm, maximum ≈ 2.35 mm; 100% of samples fall within ≤ 5 mm (hence also within ≤ 10 mm). By contrast, the legacy method yields MAE ≈ 4.56 mm, RMSE ≈ 5.14 mm, 95th percentile ≈ 8.73 mm, maximum ≈ 10.47 mm; ≈ 70% within ≤ 5 mm and ≈ 90% within ≤ 10 mm. Box-and-whisker plots (not shown here) illustrate the new method’s “dual compression” of both the interquartile band and the tails.

4.6 Motion Smoothness Results

In line with the smoothness narrative, we reconstructed derivatives using central differences to obtain the vector magnitudes of velocity, acceleration, and jerk. For each motion segment, we characterized the process by peak speed, RMS acceleration, RMS jerk, and the standard deviation of position during the dwell (stop) interval. Metrics were averaged over comparable point pairs.

Key findings (new method vs. old):

• Peak speed: ~752 mm/s (old: ~1,581 mm/s)
• RMS acceleration: 1,174 mm/s² (down from 1,833 mm/s²)
• RMS jerk: 17,766 mm/s³ (down from 21,975 mm/s³)
• Dwell jitter (position STD): ~0.00 mm (old: ~6.83 mm)

Moreover, pairwise jerk comparisons across corresponding points clearly reinforce this trend in favor of the new method.

4.7 Mechanisms of Improvement

Two mechanisms underlie these gains. First, the time-parameterized Cartesian profile imposes explicit constraints on peak velocity and peak acceleration, which inherently suppresses peak jerk and reduces excitation of structural modes. Second, continuity-preserving IK keeps the joint configuration as close as possible to the previous pose, avoiding sudden flips in singular neighborhoods and thereby curbing late-stage aggressive motions. Together, these effects make the arm “gentler” as it approaches the target and virtually eliminate “ringing” at stop, consistent with the observed near-zero jitter.

4.8 Limitations and Risks

This dataset still has boundaries. The new method did not cover all nine trajectory pairs; pairwise comparisons were limited to {1, 2, 3}. Numerical differentiation is noise-sensitive; we intentionally avoided filtering to remain faithful to the raw dynamics, though a first-order low-pass or Savitzky–Golay smoother can be introduced later to verify robustness across estimators. In addition, this round of tests used no end-effector mass, so grasp–release transients and load inertia must be addressed in follow-up evaluations.

Conclusion

Instruction book