This study compared manual palpation with trans-abdominal echocardiography (echo) in assessing graft function in a heterotopic heart transplant mouse model. The study consisted of three sequential phases: Phase 1 defined intra- and inter-operator differences in graft rejection determination when using manual palpation or echocardiography. Phase 2 assessed the diagnostic accuracy of each method using direct graft visualization as the gold standard. Phase 3 tested the correlation between echocardiographic findings and the magnitude of T cell alloimmune responses. Experiments in all phases were repeated once for confirmation.

WT C57BL/6J and WT BALB/c mice were bred at the Icahn School of Medicine at Mount Sinai (ISMMS) or purchased from the Jackson Laboratory (Bar Harbor, ME). Male and female mice were used in equal numbers as both recipients and donors. All animals were housed at the Icahn School of Medicine at Mount Sinai and cared for as per Institutional Animal Care and Use Committee (IACUC) and Association for Assessment and Accreditation of Laboratory Animal Care (AALAC) guidelines.

TONBO Bioscience: FC shield (cat. no. 70-0161-U500), CD4-PECy7 (cat. no. 60-0041-U100). BioLegend: MHC II–BV605 (cat. no. 107639), CD8–FITC (cat. no. 100706), H2-Kᵇ–PerCP-Cy5.5 (cat. no. 116516). Invitrogen: Viability–APC-Cy7 (cat. no. 65-0865-14), IFNγ–PerCP-Cy5.5 (cat. no. 45-7311-82), H2-Kᵈ–APC (cat. no. 17-5957-82), H2-Kᵇ–Pacific Blue (cat. no. 48-5958-82).

The microsurgery core at the ISMMS performed the heterotopic heart transplantation procedures as previously described (11). WT B6 mice between the ages of 6-12 weeks were used as recipients and treated with a single dose of 250ug CTLA4Ig (Abatacept) via intraperitoneal injection on post-transplant day 2. Both male and female mice were used as recipients. Buprenorphine (0.015 mg/mL) was administered subcutaneously (3.33uL/gm) for post-operative analgesia for five days post-transplant, in accordance with IACUC guidelines. All animals were housed under specific pathogen-free conditions.

Mice were assessed by both palpation and echocardiography on the same day but at independent times. All evaluators provided independent and blinded observations.

Mouse spleens were physically dissociated, passed through a 40 μm strainer (Fisherbrand™ Sterile Cell Strainers, Thermo Fisher Scientific, Waltham, MA, cat 22-363-547), and lysed with RBC lysis buffer (Thermo Fisher Scientific, Waltham, MA, cat A1049201). To reduce experimental variability, all recipient splenocytes were frozen (90% FBS/10% DMSO) for synchronized testing in mixed lymphocyte reactions.

Frozen recipient splenocytes (responders) were defrosted simultaneously on the day of the experiment. Responder cells (2 × 10⁶) were co-cultured with allogeneic BALB/c antigen presenting cells (APCs, 2 × 10⁶) in RPMI media (Gibco, Thermo Fisher Scientific; cat. no. 11875-101) with 10% FBS (Gibco; cat. no. A5256801) as previously described (12). APCs were enriched from naïve BALB/c splenocytes by taking the negative fraction of the magnetic separation kit as per manufacturer's instructions (EasySep™ Mouse CD90.2 Positive Selection Kit II, Cat 18951, STEMCELL Technologies, Cambridge, MA). Co-cultures were incubated at 37 °C with 5% CO₂ for 24 h, followed by a 4-hour treatment with GolgiPlug (BD Biosciences; cat. no. 555029). Cells were then stained for viability and surface markers, fixed and permeabilized (eBioscience™ Fixation/Permeabilization Kit, Thermo Fisher Scientific; cat. no. 00-5523-00), stained for IFNγ, and analyzed by flow cytometry.

All statistical procedures were carried out with GraphPad Prism®, version 10.5.0 (GraphPad Software, San Diego, CA, USA). Continuous variables are expressed as mean ± standard deviation (SD) unless otherwise stated. Two-sided p-values < 0.05 were considered statistically significant. Normality was tested with the Shapiro–Wilk test and homogeneity of variances with Levene's test. For comparison between groups, normally distributed data were compared with an unpaired two-tailed Student's t-test; non-normal data were analyzed with the Mann–Whitney U test. Categorical comparisons, including the proportion of mice experiencing one or more discordant “reversal” episodes between serial graft assessments, were evaluated with a two-tailed Fisher's exact test; odds ratios (OR) and 95% confidence intervals (CI) were derived from the corresponding 2 × 2 contingency tables. For agreement between assessment methods, linear relationships between quantitative assessment techniques were quantified with the Pearson correlation coefficient (r) and its 95% confidence interval (CI). Diagnostic performance (sensitivity, specificity, positive/negative predictive values) was calculated for each method against the reference standard. Kaplan–Meier curves were constructed to estimate cumulative graft survival. Differences between curves were evaluated with the log-rank test.

We first compared echocardiography and manual palpation in determining rejection of cardiac allografts in mice. We transplanted a set (n = 6/group) of heterotopic BALB/c hearts into MHC-mismatched WT B6 mice treated with CTLA4Ig immunosuppression to delay rejection, then enlisted two blinded echocardiographers (Echo1 and Echo2) and two independent blinded palpators (P1 and P2) to assess graft rejection. To explicitly address the role of operator experience, palpation was performed by investigators with distinct training levels: an experienced investigator and a junior postdoctoral fellow early in training. This design intentionally reflects the structure of many academic laboratories, where graft monitoring is frequently performed by trainees and where variability introduced at this step can propagate through downstream analyses. The timing of rejection was reported to an independent team member who also sacrificed the mice for visual confirmation of graft rejection once all four evaluators recorded cessation of beating (Figure 1A, schematic).

At the cohort level, graft survival curves generated by palpation and echocardiography were broadly similar. However, rejection timing at the individual-animal level differed substantially between methods (Figures 1B–D). We quantified the concordance between each evaluator using Pearson's coefficient and found that Echo1 and Echo2 had nearly perfect concordance (Echo1 vs. Echo2: r = 0.94, 95% CI 0.51–0.99), while P1 and P2 had significant variation (P1 vs. P2: r = 0.89, 95% CI 0.29–0.98) (Figure 1D). When we compared echo with palpation, Palpator 2 (principal investigator, PI) had higher concordance with echocardiography than P1 (post-doctoral fellow), as anticipated.

We next evaluated the intra-observer variability between ultrasound and manual palpation. Due to the relative mobility of the transplanted heart and potential for overlying intestine interfering with assessment, determination of graft rejection requires serial palpations to confirm cessation of beating. In practice, a graft flagged as possibly rejected is often determined to be beating on subsequent evaluation. Manual palpation led to 42.6% of mice having at least one such discordant episode. By contrast, echocardiography was almost universally consistent in delineating rejection, with this form of discordance occurring only once (odds ratio = 0.01, p = 0.038 relative to manual palpation).

To determine the accuracy of echo vs. palpation in determining graft rejection, we transplanted a second set of B6 mice with BALB/c hearts (n = 12 mice) and had the same two palpators and two echocardiographers evaluate the grafts for rejection. Here, we sacrificed the recipient when any evaluator determined graft rejection, and the heart was visually inspected after laparotomy for cessation of beating (gold standard) and submitted for additional histologic confirmation (Figures 2A,B). There were no differences in graft survival or M-mode trajectories when the recipients were stratified by sex (Supplementary Figure 1). Echocardiography achieved almost perfect diagnostic performance: sensitivity 92% (CI 64.4-98.5%), specificity 100% (CI 75.7%–100%), positive-predictive value (PPV) 100% (CI 74.1%–100%), negative-predictive value (NPV) 86% (CI 66.7–98.6%), and overall accuracy 96% (CI 79.8–99.3%) (average between the two investigators, Table 1). The composite sensitivity was derived from pooled data across both investigators: 11 true positives and 1 false negative across 12 total animals and the 95% confidence interval was calculated using the Wilson score method applied to the aggregate TP and FN counts. We show the individual investigator performance with separate confidence intervals in Table 1.

Manual palpation again showed significant operator-dependent variability as previously. Palpator 1 detected 2 of the 6 true rejections (sensitivity 40%) and misclassified 5 of 6 beating grafts as rejected (specificity 17%), yielding an accuracy of 27% and a positive likelihood ratio (LR⁺) of 0.48, values akin to random chance. Palpator 2 performed better (sensitivity 50%, specificity 83%), but still misclassified one-third of examinations and produced a modest LR⁺ of 3 (Table 1). Taken together, these results confirm that high-resolution echocardiography provides more accurate detection of heterotopic heart-graft failure and effectively reduces operator-dependent variability. Indeed, significant palpator experience (post doc v PI) is required to overcome the improved performance from even limited training with echocardiography.

In transplant immunology studies, the magnitude of the anti-donor T cell response is quantified at a pre-determined post-transplant time-point, typically before rejections are predicted to occur. However, as the graft survival curves show (Figure 1C), the median time to rejection has significant variability, 42.5 days (IQR 29–55.5), suggesting that at any given time point the anti-donor immune reaction is similarly variable. A benefit in using echocardiography-guided evaluation of the allograft is direct visualization and measurement of allograft beating parameters. The M-mode, or motion mode, on ultrasound measures the movement of structures, such as the heart wall (Figure 1B). During phases 1 and 2 above, the echocardiographers noted that there was a distinct reduction in the M-mode (decrease of > 200 bpm) that preceded graft rejection. We verified this finding in a separate set of B6 recipient mice (n = 6) and observed that this change occurred in all mice at 10.3 +/- 4.8 days prior to rejection (Figure 3A).

We postulated that the observed change in M-mode on ultrasound correlated with the evolving magnitude of the allogeneic T cell immunity (T cells mediate rejection in this model). To test this, we transplanted two sets (n = 4/group) of WT B6 animals with BALB/c hearts as previously. For one set of mice, we quantified anti-donor T cell immunity by mixed lymphocyte reaction (MLR) at 26d post-transplant (usual practice, fixed time point), which was late post-transplant but before any grafts rejected. For the second set of mice, we harvested splenocytes for MLR when decrease in the M-mode as described above was recorded (Echo guided, Figure 3B, schematic).

The echo guided group had spleens harvested at 36.25 +/- 8.44 days post-transplant. The overall proportion of alloreactive IFNγ- and TNFα-producing CD4 + and CD8+ T cells did not differ between the two groups, while the percentage of IL2-producing T cell subsets was greater in the fixed time point group. Importantly, in all conditions, the variance in the alloreactive subsets was significantly lower (Welch's) in the echo guided than time point guided group (Figures 3C,D). Levene's test for homogeneity of variances also revealed a trend toward reduced variance in echo-guided assessment compared to fixed time-point, though did not statistical significance due to the small sample size (n = 4). These findings establish a link between alloimmune responses and quantifiable changes in graft function.