Journal Logo

Original Clinical Science—General

Assessing the Relationship Between Molecular Rejection and Parenchymal Injury in Heart Transplant Biopsies

Madill-Thomsen, Katelynn S. PhD1; Reeve, Jeff PhD1; Aliabadi-Zuckermann, Arezu MD2; Cadeiras, Martin MD3; Crespo-Leiro, Marisa G. MD, PhD4; Depasquale, Eugene C. MD5; Deng, Mario MD6; Goekler, Johannes MD2; Kim, Daniel H. MD1; Kobashigawa, Jon MD7; Macdonald, Peter MD, PhD8; Potena, Luciano MD9; Shah, Keyur MD10; Stehlik, Josef MD, MPH11; Zuckermann, Andreas MD2; Halloran, Philip F. MD, PhD1

Author Information
Transplantation: August 15, 2022 - Volume - Issue - 10.1097/TP.0000000000004231
doi: 10.1097/TP.0000000000004231
  • Free
  • SDC
  • PAP
  • Infographic

Abstract

INTRODUCTION

Parenchymal injury occurs in every heart transplant, and the quality of the heart parenchyma determines function and, ultimately, outcome. Transplantation subjects heart tissue to unique stresses, including brain death, preservation-implantation, donor-derived changes,1 rejection, and infection. Cardiac myocytes are particularly susceptible to injury, which triggers inflammation/innate immunity.2 In addition, many late heart transplants show suboptimal function and outcomes3-11 and have abnormalities such as interstitial fibrosis and diastolic dysfunction reflecting parenchymal injury.12,13 This may be associated with coronary artery abnormalities—cardiac allograft vasculopathy (CAV)14—which is not unexpected because arteries are donor tissue and subject to all of these unique stresses.

A molecular assessment of parenchymal injury is therefore of interest. The Molecular Microscope Diagnostic System (MMDx)2,15-19 measures genome-wide gene expression from 49 495 probesets in endomyocardial biopsies (EMBs). MMDx previously used the expression of rejection-associated transcripts (RATs) plus archetypal analysis (AA) to molecularly define T cell–mediated rejection (TCMR) and antibody-mediated rejection (ABMR). Because some transcripts are shared between rejection and innate immunity, this analysis also detected some early inflamed biopsies with injury but with no rejection (NR-Early injury)20; however, this analysis of RATs distinguished “early injury-or-rejection” and therefore could not assess the extent of parenchymal injury in biopsies with rejection or explore the problem of late injury.

The present study of heart transplant EMBs aimed to define the extent of parenchymal injury in every biopsy, as well as its rejection state and to establish the relationships of TCMR and ABMR with parenchymal injury. We aimed to measure the expression of previously defined injury-related transcript sets to classify injury states using strategies similar to those that defined rejection states using RATs. The injury-related transcript sets included those with increased expression in mouse heart isografts compared with normal hearts—cardiac injury-repair transcripts (cIRITs)21—and heart transcripts (HTs) with high expression in normal human hearts.19 We included additional injury-induced transcript sets originally annotated in injured kidneys but which also increase in injured hearts with no rejection, correlating with the transcripts induced in injured hearts (cIRITs).2,20 We included macrophage transcripts to reflect the innate immune response. Finally, we included immunoglobulin transcripts as a potential marker for late parenchymal deterioration because they increase with time in other organ transplant biopsies22-25 and correlate with atrophy-fibrosis in kidneys. We hypothesized that assessing the parenchymal injury state in the heart in addition to diagnosing the rejection state would help us to understand dysfunction and risk of graft loss in biopsies with no rejection, and allow us to compare the relative impact of TCMR and ABMR on the heart parenchyma.

The research plan is summarized in Figure 1A and B. A summary of abbreviations is provided in Table S1, SDC, (https://links.lww.com/TP/C469).

F1
FIGURE 1.:
(A) CONSORT Diagram showing study enrollment and workflow in this study. (B) Study flowchart describing the sample and analysis strategy of this article. EMB, endomyocardial biopsy; LVEF, left ventricular ejection fraction; PCA, principal component analysis; TCMR, T cell–mediated rejection.

MATERIALS AND METHODS

Population

We used microarrays to analyze gene expression in 1320 EMBs from the prospective INTERHEART study, approved by the ethics review board of each local center (ClinicalTrials.gov #NCT02670408).17 Standard-of-care (SOC) biopsies for clinical indications or protocol from consenting patients at 13 centers were placed in RNAlater and shipped to the Alberta Transplant Applied Genomics Centre.26 Histology followed International Society for Heart and Lung Transplantation guidelines27,28 per local SOC, interpreted to permit histology-molecular comparisons as previously reported.2,18,19 Molecular diagnoses were assigned without knowledge of histology or donor-specific antibody. All biopsies adequate for molecular examination were included (~98%).

Microarray Analysis

As described,2,17,19 total RNA from EMBs was labeled with the 3′ IVT Plus kit (Affymetrix, SC) and hybridized to PrimeView microarrays (Affymetrix) using manufacturer protocols (www.affymetrix.com). CEL files are available on the Gene Expression Omnibus website (GSE150059).

Pathogenesis-based Transcript Sets

Transcript sets were previously annotated in cell lines, experimental models, and human transplant biopsies (https://www.ualberta.ca/medicine/institutes-centres-groups/atagc/research/gene-lists). Transcript set scores are the mean fold change across all probes within the set, using log2 raw data, compared with controls (371 biopsies with no molecular rejection >30 d posttransplant). Statistics and calculations were performed using the log of the scores.

We used 10 transcript sets listed in Table 1:

TABLE 1. - The 10 injury-related pathogenesis-based transcript setsa,b used for the injury-based principal component and archetypal analyses
Biological processes Transcript set Description of transcript set Detail
Expressed in macrophages QCMAT Quantitative constitutive macrophage associated High expression in human primary macrophages, not inducible by IFNG 46
AMAT Alternative macrophage activation Alternative activation-induced in mouse macrophages 46
Increased in recently injured hearts cIRIT Cardiac injury and repair induced Injury—induced in mouse cardiac isografts 47 compared with normal hearts
Other injury-induced transcript sets that correlate with cIRITs in human hearts 48 IRRAT Injury-repair response associated Induced in early human kidney transplant injury 49
IRITD3 Injury and repair induced Induced in mouse kidney isografts, peaking at day 3 posttransplant 50
IRITD5 Injury and rejection induced Induced in mouse kidney isografts, peaking at day 5 posttransplant 50
DAMP Damage-associated molecular pattern Literature-based damage-associated molecular pattern (DAMP) 51,52
Highly expressed in normal heart
(“Normalness”) HT1 Normal heart transcripts—set 1 (heart-selective compared with kidney) High expression in normal mouse heart compared with kidney (no solute carriers) 53
HT2 Normal heart transcripts—set 2 (heart-selective compared with kidney) High expression in normal mouse heart compared with kidney (solute carriers) 53
Increased in late transplants IGT Immunoglobulin transcripts Increased by time and associated with atrophy-fibrosis, reflecting plasma cells 54
bThe transcript sets were empirically derived in human cell lines, human transplants, and mouse models. They reflect biological processes relevant to rejection and injury.
AMAT, alternative macrophage-associated transcripts; cIRIT, cardiac injury-repair induced transcripts; DAMP, damage-associated molecular pattern transcripts; HT1, heart transcripts set 1; HT2, heart transcripts 2; IGT, immunoglobulin transcripts; IRITD3, injury-repair induced transcripts day 3; IRITD5, injury-repair induced transcripts day 5; IRRAT, AKI transcripts; QCMAT, quantitative constitutive macrophage-associated transcripts.

  1. 5 previously annotated as induced by recent injury2,21;
  2. 2 reflecting macrophage infiltration, which is triggered by heart injury2;
  3. 2 highly expressed in normal hearts19;
  4. the immunoglobulin transcripts, which increase with time in kidney and lung transplants and correlate with atrophy-fibrosis in kidney transplants.22,23

Dimensionality reduction, clustering, and data visualization. Principal component analysis (PCA) and AA were described previously.2,19 PCA is used to reduce the dimensionality of datasets with large numbers of variables with minimal loss of information, facilitating analysis and visualization. AA assigns a user-defined number of archetypes (idealized extreme phenotypes) to a dataset. Each sample is assigned scores representing its proportional relationship to each archetype, summing to 1.0. The highest score for each biopsy assigned the biopsy to an archetype group.

Rejection PCA and AA were based on RAT expression.20 Injury PCA and AA were based on injury-related transcript set scores. We used the “FactoMineR”29 and “archetypes”30 packages in R, version 3.6.2.31

Rejection

Molecular rejection sign-out categories have been described.20 Each biopsy was assigned to 1 of the 8 modified sign-out categories: ABMR, pABMR, TCMR, pTCMR, mixed, and No rejection (NR), which was subdivided into NR-Minor, NR-Normal, and NR-Early injury.20

In addition, the published 5 rejection archetype model20 was used for some analyses: NR, TCMR, ABMR, Minor-injury, and Early-injury. Archetypes are automatically assigned and avoid subjectivity. Rejection PCA and AA used a 1320 (biopsy) × 437 (RATs) dataset as input.

Injury

PCA and AA for injury were based on the expression of 10 injury transcript sets (Table 1).

We used 1320 (biopsy) × 10 (injury transcript set scores) dataset as input. The injury class assignments were independent of the rejection class assignments.

A 5-archetype injury model was selected by inspecting models with 2to6 groups and choosing the one with the best trade-off between biological interpretability and diversity: No-injury, Minor-injury, Moderate-injury, Severe-injury, and Late-injury.

Rolling Averages

Data were first ordered by the variable on the x-axis. Then a sliding window of the size indicated was used to plot mean y versus mean x values, for example, with a window size of 200 the means of samples 1–200 is plotted, then samples 2–201, 3–202, etc.

Survival Analyses

Analyses were based on death-censored survival 3 y postbiopsy, using 1 random biopsy per transplant. Patients with grafts surviving longer were censored at 3 y. In this set, median follow-up was 343 d, and 52 transplants failed within 3 y of biopsy. Kaplan-Meier estimates and plots used the R “survival” package.32

Logistic Regression

The R “rms” package33 was used for logistic regression. Because of collinearity, multivariable logistic regression using the rejection and injury archetype scores excluded the “normal” scores for No-rejection and No-injury.

Splines

Restricted cubic splines were used to show nonlinear relationships between variables. Three “knots” were selected and smooth curves fit based on within-knot data constrained so that curves between segments are joined. Overfitting is minimized in restricted cubic splines by using only linear trend lines for segments beyond the left- and right-most knots, reducing the influence of the tails of the distributions where fewer data points are available. Splines were generated using the R package “rms.”33 The threshold selected for each plot was based on visual clarity (0.4 in the ABMR biopsies, 0.35 in the TCMR biopsies).

Left Ventricular Ejection Fraction

We selected a cutoff of 55 for left ventricular ejection fraction (LVEF) when binary groups were needed for analyses (high LVEF as >55, low LVEF as ≤55). This was based on recommendations from the clinical investigators in the INTERHEART study group and on literature supporting this threshold.34-36 We additionally used multiple cutoffs to show relationships with LVEF in more detail when needed (LVEF <30, 30–45, and >45).

RESULTS

Population and Demographics

To understand the relationships between rejection and parenchymal injury, we examined the same 1320 EMBs from 645 patients used for the previous rejection analysis20 (Figure 1A). Table S2, SDC, (https://links.lww.com/TP/C469), shows population demographics,20 and Table S3, SDC, (https://links.lww.com/TP/C469), shows histologic and molecular diagnoses. Molecular rejection sign-outs20 classified 853 biopsies as NR, with 3 “NR” subclasses: NR-Normal (N = 462), NR-Minor (N = 359), and NR-Early injury (N = 32). Rejection-related sign-out classes were assigned to 467 biopsies: ABMR-related (ABMR = 179 and possible ABMR [pABMR] = 161), and TCMR-related including Mixed (TCMR = 76, pTCMR = 38, and Mixed = 13). We grouped Mixed with TCMR-related for some analyses because TCMR rapidly produces parenchymal injury.2,21,37

LVEF was decreased in many late hearts and slightly increased in many hearts early posttransplant (likely from donation-implantation injury; Figure S1A and S1B, SDC, https://links.lww.com/TP/C469). In a ttest comparing biopsies before and after 1-y posttransplant, mean LVEF was lower in later biopsies (P = 0.02).

Relationships Between Injury-induced Transcript Sets and LVEF

We examined the average expression of each injury-associated transcript set in EMBs from hearts with high LVEF >45, intermediate LVEF 30 to 45, or low LVEF ≤30 (Table 2). Except for IRITD5 and DAMPs, injury-related transcript sets were significantly different between LVEF groups. In hearts with low LVEF <30, expression of the injury-increased transcript sets (cIRIT, IRRAT, and IRITD3) and the macrophage transcript sets was higher, and the normal HTs (HT1 and HT2) were lower than hearts with intermediate (30–45) or high LVEF (>45). This confirmed that the injury-associated transcript sets reflect the injury status of the heart parenchyma.

TABLE 2. - Expression of injury-related pathogenesis-based transcript setsa,b in hearts with LVEF>55 vs LVEF≤55
Biological processes Injury-related transcript set LVEF >55 LVEF ≤55 P value for LVEF >55 vs LVEF ≤55
Expressed in macrophages QCMAT 0.30 0.54 1.8 × 10−07
AMAT 0.40 0.68 1.5 × 10−07
Increased in recently injured hearts cIRIT 0.10 0.17 2.4 × 10−05
Other injury-induced transcript sets that correlate with cIRITs IRRAT 0.21 0.32 0.001
IRITD3 0.07 0.10 0.007
IRITD5 0.11 0.10 0.83
DAMP 0.06 0.08 0.28
Highly expressed in normal heart HT1 −0.08 −0.18 1.2 × 10−09
HT2 −0.12 −0.29 1.1 × 10−10
Increased in time IGT 0.30 0.80 5.8 × 10−09
bThe transcript sets were empirically derived in human cell lines, human transplants, and mouse models. They reflect biological processes relevant to rejection and injury.
AMAT, alternative macrophage associated transcripts; cIRIT, cardiac injury-repair induced transcripts; DAMP, damage-associated molecular pattern transcripts; HT1, heart transcripts set 1; HT2, heart transcripts 2; IGT, immunoglobulin transcripts; IRITD3, injury-repair induced transcripts day 3; IRITD5, injury-repair induced transcripts day 5; IRRAT, AKI transcripts; LVEF, left ventricular fraction; QCMAT, quantitative constitutive macrophage-associated transcripts.

Visualizing Parenchymal Injury Groups (Archetypal Clustering)

Figure 2A shows the correlation of each of the 10 injury transcript sets with injury PC1 and PC2. PC1 accounted for 71% and PC2 for 12% of the variance in the data set (Figure 2B). PC1 correlated positively with all transcript set scores increased by injury and negatively with normal heart parenchymal transcripts. Thus, increasing PC1 indicates increasing parenchymal injury and dedifferentiation.

F2
FIGURE 2.:
PCA and AA using the 1320 biopsy population. PCA and AA representing injury was done with 10 injury-associated transcript set scores as input variables. (A) Correlations between the input variables and PCs 1 and 2. Transcript set scores: AMAT, cIRITs, DAMP, HT1, HT2, IGT, IRITD3, IRITD5, IRRAT, and QCMAT. (B) Principal component scores from the same analysis. Each small triangle is a biopsy, colored by its injury archetype group. AA, archetypal analysis; AMAT, alternative macrophage-associated transcripts; cIRIT, cardiac injury and repair induced transcript; DAMP, damage-associated molecular pattern transcripts; HT1, heart transcripts set 1; HT2, heart transcripts set 2; IGT, immunoglobulin transcripts; IRITD3, injury and rejection-induced transcripts (intermediate); IRITD5, injury and rejection-induced transcripts (late); IRRAT, injury/repair-associated transcripts (acute kidney injury); LVEF, left ventricular ejection fraction; PCA, principal component analysis; QCMAT, quantitative constitutive macrophage-associated transcripts.

Injury PC2 correlated positively with immunoglobulin transcript sets (IGTs), which strongly correlate with time-dependent parenchymal deterioration (atrophy-fibrosis) in kidneys.22,25

As shown by their vectors in Figure 2A, LVEF decreased as PC1 increased and as PC2 increased. Time posttransplant and immunoglobulin transcripts increased with PC2.

Figure 2B shows the biopsies plotted by their injury PC scores. The location of each biopsy is determined by scores for the injury-related transcript sets and the vectors in Figure 2A. Each biopsy is colored by its injury archetype group assignment. We named the injury groups for the molecular and clinical features most characteristic of the group: No-injury, N = 376; Mild-injury, N = 526; Moderate-injury, N = 110; and Severe-injury, N = 87. The archetype group with high PC2 was called Late-injury (N = 221).

Figure 2B shows a progression of injury severity corresponding with increasing PC1: No-injury to Mild-injury to Moderate-injury to Severe-injury. The Late-injury group had high PC2 scores, which correlate with time.

Characteristics of the Injury Phenotype States

Table 3 shows the mean (median) time posttransplant (d) and injury transcript set scores across the 5 injury archetype states. Moderate-injury was earliest (mean 218 d), followed by Mild-injury (408 d) and then No-injury (1065 d), suggesting recovery from universal donation-implantation injury. Severe-injury was intermediate (548 d), perhaps because it sometimes reflects TCMR-induced injury—see below. Late-injury had the longest mean time posttransplant (1430 d).

TABLE 3. - Mean time posttransplant and transcript set scores in the 5 parenchymal injury states
Injury variables assessed Parenchymal injury states
No-injury (N = 376) Mild-injury (N = 526) Moderate-injury (N = 110) Severe-injury (N = 87) Late-injury (N = 221)
Mean days posttransplant (median) 1065 (329) 408 (126) 218 (65) 548 (85) 1430 (712)
Biological processes Transcript sets Mean transcript set scores in each parenchymal injury state
Expressed in macrophages QCMAT a 1.05 1.17 1.45 2.80 1.54
AMAT a 1.08 1.24 1.67 3.28 1.78
Increased in recently injured hearts cIRIT a 1.00 1.05 1.22 1.47 1.15
Other injury-induced transcript sets that correlate with cIRITs IRRAT a 0.99 1.15 1.61 2.16 1.26
IRITD3 a 0.99 1.04 1.19 1.26 1.08
IRITD5 a 0.99 1.07 1.35 1.40 1.10
DAMP a 0.92 1.13 1.02 1.41 1.03
Highly expressed in normal heart HT1 a 0.98 0.98 0.86 0.68 0.88
HT2 a 0.97 0.99 0.79 0.54 0.83
Increased in late transplants IGT a 1.03 0.99 1.03 1.79 3.19
Bold mark the highest expression groups in each row; underlining indicates the lowest.
aThese were the 10 transcript sets used in the principal component and archetypal analyses. AMAT, alternative macrophage associated transcripts; cIRIT, cardiac injury-repair induced transcripts; DAMP, damage-associated molecular pattern transcripts; HT1, heart transcripts set 1; HT2, heart transcripts 2; IGT, immunoglobulin transcripts; IRITD3, injury-repair induced transcripts day 3; IRITD5, injury-repair induced transcripts day 5; IRRAT, AKI transcripts; QCMAT, quantitative constitutive macrophage-associated transcripts.

Injury scores increased on a gradient from No- to Mild- to Moderate- to Severe-injury, with correspondingly progressively decreasing normal HT scores.

The main feature of Late-injury biopsies was high expression of the immunoglobulin-associated transcript set, compatible with a plasma cell infiltrate as described in late kidney transplants with atrophy-fibrosis,22 in late lung transplants,23 and in late failing heart allografts.38,39

Although the INTERHEART study only includes SOC biopsies and is not allowed to deliberately perform biopsies for time series analyses, there were some patients with multiple biopsies. Most Mild-injury phenotypes either stayed Mild-injury or became No-injury in later biopsies; most Moderate-injury phenotypes improved to Mild- or No-injury or stayed Moderate-injury (data not shown).

We previously showed that early cardiac injury was associated with loss of myosin and tropomyosin transcripts.2 We confirmed that myosin and tropomyosin genes, which were highly expressed in normal EMBs, declined with increasing scores for PC1, Severe-injury, and Late-injury (Table 4), compatible with injury-induced myocyte dedifferentiation.

TABLE 4. - Relationship of myosin and tropomyosin expression to parenchymal injury statesa
Affymetrix designation Gene symbol Gene name Spearman correlation of gene expression with PC1 and Severe-/Late-injury archetype scores in 1320 biopsies
Injury PC1 score Severe-injury score Late-injury score
11718277_a_at MYL2 Myosin light chain 2 −0.28 −0.30 −0.14
11719790_a_at MYL3 Myosin light chain 3 −0.51 −0.54 −0.87
AVERAGE 0.40 0.42 1.01
11740313_s_at MYH6, MYH7 Myosin, heavy chain 6, alpha; myosin, heavy chain 7, beta −0.41 −0.48 −0.15
11717570_s_at MYH7 Myosin, heavy chain 7, beta −0.42 −0.43 −0.17
AVERAGE 0.415 0.46 0.16
11738892_a_at TPM1 Tropomyosin 1 (alpha) −0.22 −0.26 −0.07
11738893_s_at TPM1 Tropomyosin 1 (alpha) −0.36 −0.40 −0.10
11742308_s_at TPM1 Tropomyosin 1 (alpha) −0.34 −0.39 −0.09
11742309_x_at TPM1 Tropomyosin 1 (alpha) −0.25 −0.35 −0.11
AVERAGE 0.29 0.35 0.09
aProbesets for myosin light chains and heavy chains and for tropomyosin genes were selected only on the bases of high expression values (>10 000) in normal biopsies.

Figure 3 illustrates some trends in this biopsy population. A striking rise in the IGT and Late-injury scores over log time was accompanied by the decline of the Severe-injury score, the recent injury transcript set scores (cIRIT and IRRAT), and the macrophage transcript set scores (Figure 3A). The close relationship between the cIRIT and Severe-injury score is shown in Figure 3B, and likewise between Late-injury scores and IGT scores in Figure 3C.

F3
FIGURE 3.:
Relationships between injury-induced transcript set scores, time posttransplant, and archetypal injury states. (A) Rolling means showing the relationships between Late- and Severe-injury scores, QCMATs, IRRATs, cIRITs, and IGTs and time posttransplant (window size = 400). (B) A scatterplot showing the relationship between the Severe-injury score and the cIRIT score (biopsies assigned to the Severe-injury group are colored red). (C) A scatterplot showing the relationship between the Late-injury score and the IGT score (biopsies assigned to the Late-injury group are colored cyan). cIRIT, cardiac injury-repair induced transcripts; IGT, immunoglobulin transcripts; IRRAT, AKI transcripts; QCMAT, quantitative constitutive macrophage-associated transcripts.

Relating Injury Archetype Groups to Rejection

Table 5 distributes the biopsies called TCMR and ABMR by molecular rejection sign-outs into their parenchymal injury groups. To avoid excessive subgroups, we grouped the TCMR-related biopsies (TCMR, pTCMR, and Mixed rejection) and the ABMR-related biopsies (ABMR and pABMR). Mixed was grouped with TCMR because TCMR rapidly induces parenchymal injury.37

TABLE 5. - Parenchymal injury states in hearts with TCMR and ABMR
Modified rejection sign-outs Parenchymal injury states Row Total
No-injury (N = 376) Mild-injury (N = 526) Moderate-injury (N = 110) Severe-injury (N = 87) Late-injury (N = 221)
TCMR-related, including mixed and pTCMR 1 (1%) 14 (11%) 4 (3%) 44 (35%) 64 (50%) 127
Mean days posttransplant 164 d 134 d 144 d 885 d 1064 d
ABMR-related including pABMR (excluding mixed) 66 (19%) 105 (31%) 47 (14%) 21 (6%) 101 (30%) 340
Mean days posttransplant 970 d 510 d 350 d 342 d 1729 d
Principal injury groups in each RAT sign-out group (>15% of row total) are bolded and shaded. ABMR, antibody-mediated rejection; RAT, rejection-associated transcript; pABMR, possible ABMR; TCMR, T cell–mediated rejection.

Biopsies with TCMR almost always (85%) had extensive parenchymal damage (assigned to Severe- or Late-injury phenotypes) and virtually never had No-injury.

In contrast, ABMR biopsies seldom had Severe-injury, and 19% had No-injury. Early ABMR-related biopsies usually had the mild parenchymal injury characteristic of all biopsies in the early posttransplant period: Moderate-injury (350 d), Mild-injury (510 d), and No-injury (970 d); however, 30% of ABMR-related biopsies were classified Late-injury at a much later time (1729 d). ABMR was detected in many hearts (46%) with Late-injury.

Figure 4 visualizes these associations using splines. We plotted the Severe-injury archetype score against time posttransplant and colored each biopsy by its rejection sign-out category (Figure 4A). We summarized each rejection group by a spline line. TCMR biopsies (red) consistently had high Severe-injury scores, but ABMR biopsies (blue) did not.

F4
FIGURE 4.:
Splines showing the evolution of Severe- and Late-injury scores over time within the modified rejection sign-out categories. Solid lines (restricted cubic splines) represent nonlinear trends in the scores over time, using logistic regression predicting (A) injury scores >0.4 and (B) >0.35. NR, no rejection; TCMR, T cell–mediated rejection.

We similarly examined the Late-injury archetype scores versus time and colored the rejection groups (Figure 4B). TCMR was consistently associated with elevated Late-injury scores. At early times, ABMR biopsies did not have elevated Late-injury scores; however, ABMR at late times (>3 y) often had Late-injury (101 of 340 or 30%), much more than late biopsies with no rejection (15 of 462 or 3%). NR-Minor biopsies (which have mild ABMR-like molecular changes despite being considered no rejection20) also showed rising Late-injury scores over time. Biopsies with NR (NR-Normal) showed no substantial increase in Late-injury scores at late times.

In summary, TCMR is strongly associated with extensive parenchymal injury, but ABMR in the early years posttransplant is not. Nevertheless, ABMR in the long term is highly associated with Late-injury.

Associations with Short-term Postbiopsy Graft Failure

As a prospective cross-sectional study, INTERHEART does not have extensive long-term follow-up but does permit estimates of short-term loss. We randomly selected 1 biopsy per patient with available follow-up information (543 hearts) and compared survival between groups of biopsies defined by their injury archetype groups.

Late- and Severe-injury biopsies showed increased risk for failure (Figure 5A, P = 0.002), even when all TCMR and ABMR rejection sign-outs were excluded (Figure 5B, P = 0.04).

F5
FIGURE 5.:
Survival probability in various biopsy groups by modified rejection sign-out categories, histology diagnoses, or injury archetypes. Kaplan-Meier plots showing 3-y postbiopsy survival stratified by (A) injury archetype groups and (B) injury archetype groups after removing all samples with molecular rejection (TCMR and ABMR). (C) Survival probability in only TCMR biopsies and (D) in only ABMR biopsies by modified rejection sign-out categories, comparing Late- and Severe-injury biopsies to those assigned to No-, Mild-, and Moderate-injury groups. All biopsy counts (N) per panel represent 1 random biopsy per patient within the selected population. ABMR, antibody-mediated rejection; TCMR, T cell–mediated rejection.

We studied biopsies with TCMR or ABMR by molecular rejection sign-outs. In biopsies with TCMR (Figure 5C), hearts with Severe- and Late-injury states showed increased graft loss. In biopsies with ABMR (Figure 5D), ABMR with Severe- and Late-injury states showed increased short-term graft loss compared with ABMR with less injury (P = 0.02ss).

Further Analyses

In multivariable Cox regression, both injury states and rejection contributed to predictions of short-term graft loss within 3 y. Injury archetype scores added predictive value to rejection archetype scores alone (P = 3.8 × 10−5; Table S4, SDC,https://links.lww.com/TP/C469). Rejection archetype scores also added predictive value to injury archetype scores alone (P = 2.6 × 10−5).

We assessed the injury states and LVEF within biopsies called ABMR and TCMR by molecular rejection sign-outs in Table 6. Within both ABMR and TCMR groups, LVEF was lower and the number of losses was increased when these hearts also had extensive injury.

TABLE 6. - Relating parenchymal injury state to LVEF and graft loss in ABMR and TCMR
Modified rejection sign-outs Injury archetype group assignment N per group Mean time posttransplant Median time posttransplant Mean LVEF Number of graft failures within 3 y
ABMR+pABMR a No-injury 66 970 366.5 62.79 3
Mild-injury 105 510 179.5 66.38 3
Moderate-injury 47 350 95 61.96 0
Severe-injury 21 342 144 56.86 2
Late-injury 101 1729 927 55.85 12
TCMR+pTCMR+Mixed b No-injury 1 164 164 55 0
Mild-injury 14 134 115 63.71 0
Moderate-injury 4 144 92 55.00 0
Severe-injury 44 885 123 53.11 8
Late-injury 64 1064 702.5 52.48 8
aWithin ABMR biopsies, a ttest comparing LVEF in biopsies with Late-injury+Severe-injury versus those with No-injury+Mild-injury+Moderate-injury was significantly different (P = 1.43 × 10−7).
bWithin TCMR biopsies, a ttest comparing LVEF in biopsies with Late-injury+Severe-injury versus those with No-injury+Mild-injury+Moderate-injury was significantly different (P = 0.007). ABMR, antibody-mediated rejection; LVEF, left ventricular ejection fraction; TCMR, T cell–mediated rejection.
Bold highlights the Severe- and Late-injury rows for comparison.

DISCUSSION

This study used the existing genome-wide microarray measurements to assign parenchymal injury states to EMBs previously classified for rejection20 and assessed the relationship between parenchymal injury and low LVEF, short-term graft loss, and rejection phenotypes. We previously found inflamed injured biopsies in our rejection analyses, but these “early injury-or-rejection” analyses could not assess the degree of parenchymal injury in biopsies with rejection or elucidate the common late dysfunction problem. The present study used archetype clustering methods based on expression of previously-annotated injury-related transcript sets. To study time-dependent late changes, we included immunoglobulin transcripts, which correlate with time-dependent parenchymal deterioration in kidney and lung transplants and which showed a striking increase with time in heart transplants. Injury transcript sets were significantly associated with low LVEF ≤30, validating their relationship to the state of the parenchyma. AA identified group of hearts with Severe-injury and a large group of hearts with Late-injury. TCMR was almost always accompanied by extensive parenchymal injury, even early posttransplant. In contrast, in the early years posttransplant, ABMR had minimal parenchymal injury (No-injury or Mild-injury) beyond that expected in all early hearts and rarely displayed Severe-injury; however, in later years posttransplant, ABMR was associated with the Late-injury state, much more than in biopsies with no rejection. Injury states emerged as strong predictors of graft dysfunction (decreased LVEF) and graft loss within 3 y postbiopsy in TCMR and ABMR groups, even in the many hearts that had no rejection. We conclude that assessing parenchymal injury states in addition to rejection states enhances our understanding of dysfunction and short-term outcomes, reveals the parenchymal impact of rejection, and highlights a large group of hearts with Late-injury, many associated with ABMR.

These results underscore the major impact of TCMR on the parenchyma compared with ABMR, explaining our earlier findings that TCMR is associated with impaired short-term survival, whereas ABMR is not.20 All failures in TCMR-related biopsies were those assigned to Severe- or Late-injury states. As clinicians, we often consider TCMR to be “treatable,” but because it often has severe effects on the parenchyma, it adversely affects short-term survival, much more so than ABMR, which appears to spare the parenchyma for prolonged periods of time.20 A majority (85%) of TCMR biopsies have extensive parenchymal injury, which we suspect will persist even after TCMR activity has been sterilized by treatment. Moreover, because we do not actually know whether our usual treatments actually sterilize TCMR molecular activity, existing parenchymal injury at the time of diagnosis may be exacerbated by new injury induced by smoldering TCMR activity long term.

These results show that although in the early years ABMR has minimal parenchymal injury beyond the universal injury of donation-implantation and comparatively little short-term graft loss,20 ABMR is not benign and may slowly develop a progressive Late-injury state with its attendant dysfunction and graft loss. The Late-injury state in kidney transplants manifests as atrophy-fibrosis, and we are currently assessing the histology of late EMBs to characterize the corresponding state in hearts. Of interest, even the subtle “Minor” ABMR-like molecular changes in biopsies usually considered to have no rejection are associated with the Late-injury state. ABMR begins as a pure microcirculation disease that spares the parenchyma but becomes associated with Late-injury over time, and most short-term failures in ABMR were in the Late-injury group (a concept supported by other recent studies40,41).

The contributions of TCMR and ABMR to graft loss may be underestimated if hearts damaged by rejection are biopsied after rejection has abated. Extensive injury states can persist even when the rejection state is sterilized by treatment or subsides spontaneously because of factors in the natural history of the immune response such as T cell checkpoints. Both TCMR and ABMR impact heart transplant function and survival through the parenchymal injury states they induce, added to which is the possibility of new injury from persisting smoldering rejection after treatment. That is why both rejection and injury states contribute to predictions of future graft loss (The processes of brain death, preservation, and implantation are universal stresses that induce recent injury transcripts in every heart to various degrees and are the main driver of elevated means for the recent injury transcripts in the population, driving the recent injury scores. These scores regress over a long time2 before reaching “normal” levels after 1 y; however, individual hearts also experience new injuries, e.g., TCMR or virus infection.).

The Late-injury state is a major problem in heart transplants and is present in 17% of all biopsies and associated with impaired function and with impaired survival even in hearts with no rejection. ABMR is present in 101 of 221 (46%) of Late-injury biopsies and is, therefore, an important contributor to this state. But even this may underestimate the impact of ABMR because ABMR activity can “burn out” as the parenchyma deteriorates. In kidney transplants, molecular ABMR activity often becomes attenuated before graft failure (Late-stage ABMR), with the persistence of atrophy-fibrosis and of the characteristic time-dependent histologic lesion, glomerular double-contours.25 In hearts, it would be very useful to have a similar time-dependent feature like kidney double contours that identifies late-stage injury induced by earlier ABMR. ABMR activity may evolve over the years as the Late-injury state emerges—from ABMR to pABMR to NR-Minor. It would also be useful to have long-term serial observations in individual patients, but this is impossible in INTERHEART, which cannot collect serial observations in patients without SOC indications. Institutional review board (IRBs) are understandably reluctant to approve time-series biopsies in patients without indications; however, using noninvasive methods such as donor-derived cell-free DNA measurements may be useful to understanding this natural history of ABMR activity.

Future studies will be necessary to determine the extent to which Severe- or Late-injury without rejection represent an evolution from previously active TCMR or ABMR, compared with the impact of factors such as stress from donation-implantation, virus infection, hypertension, and coronary artery disease. Heart transplant clinicians have long been aware of a group of late heart transplants with compromised function and increased risk of failure without clear evidence of active rejection, and the MMDx Late-injury state provides an objective classification of this state and quantifies the molecular changes. The immunoglobulin transcripts suggest that Late-injury is characterized by increased expression of low-grade inflammatory infiltrate typical of atrophy-fibrosis in kidneys,22 which also have impaired function25,42 and increased risk of failure.

Determining the relationship between Late-injury and CAV6,8,9,14 is of great interest for future studies but cannot be estimated in INTERHEART. As an IRB-approved SOC study, INTERHEART could not request coronary artery studies at the time of biopsy unless they were SOC, and indeed these were seldom done around the time of biopsy; however, we hypothesize that the Late-injury state is closely correlated to the CAV state as seen in the parenchyma. Both states are late and have impaired LVEF and survival, and the immunoglobulin transcripts in the Late-injury state recall the association of CAV with B cells.43 Many in the Late-injury group had ABMR, which has been associated with CAV.44 But the overlap between CAV and Late-injury does not necessarily mean that arterial changes are always the “cause” of Late-injury parenchymal changes. The arteries and the parenchyma are subject to many of the same stresses and may simply deteriorate in parallel with the cumulative burden of shared injuries, for example, ABMR. Some arterial narrowing in late organ transplants may reflect the loss of the parenchymal metabolic activity: in kidney transplants, fibrous intimal thickening of small arteries is a universal feature of advanced atrophy-scarring.45

Some additional limitations of this study are imposed by the IRB-approved prospective multicenter design, which in most centers limited the number of pieces available for molecular analysis to 1. The protocol for INTERHEART that the centers agreed to only allowed them to provide certain data and did not agree to share biopsy images for central histology review, and follow-up was relatively short (median 1 y). The study was based on intact RNA to avoid the irreversible damage from formalin fixation, which reduces the quality of the extracted RNA, but this does require additional tissue beyond that taken for histology. It would be useful to have the conclusions of the molecular injury studies validated outside of the INTERHEART study. The effect of treatment on molecular rejection and parenchymal injury is of great interest in the ongoing MMDx studies, but we get limited information because follow-up biopsies after treatment are not SOC. This issue will require a dedicated study with follow-up biopsies after treatment.

In conclusion, genome-wide microarray measurement of gene expression being performed to diagnose rejection also offers an opportunity for assessing parenchymal injury in heart transplant EMBs by analyzing injury-associated transcripts already measured by the microarray. This new information is “free”: no additional tissue or expense is required, only software. Expression of injury transcripts correlates with dysfunction and outcomes, and future studies can define the relationship between parenchymal injury states and CAV. The identification of injury-induced changes raises the hope that interventions can eventually be directed specifically at healing injury in damaged heart transplants to improve function and prevent graft loss.

ACKNOWLEDGMENTS

Some biopsies were provided by Dr Alexandre Loupy, Paris, France.

REFERENCES

1. Bergenfeldt H, Lund LH, Stehlik J, et al. Time-dependent prognostic effects of recipient and donor age in adult heart transplantation. J Heart Lung Transplant. 2019;38:174–183.
2. Halloran PF, Reeve J, Aliabadi AZ, et al. Exploring the cardiac response to injury in heart transplant biopsies. JCI Insight. 2018;3:123674.
3. Kobashigawa JA. The search for a gold standard to detect rejection in heart transplant patients: are we there yet? Circulation. 2017;135:936–938.
4. Wilhelm MJ. Long-term outcome following heart transplantation: current perspective. J Thorac Dis. 2015;7:549–551.
5. Khush KK, Cherikh WS, Chambers DC, et al.; International Society for Heart and Lung Transplantation. The International Thoracic Organ Transplant Registry of the International Society for Heart and Lung Transplantation: thirty-fifth Adult Heart Transplantation Report-2018; Focus Theme: Multiorgan Transplantation. J Heart Lung Transplant. 2018;37:1155–1168.
6. Stehlik J, Kobashigawa J, Hunt SA, et al. Honoring 50 years of clinical heart transplantation in circulation: in-depth state-of-the-art review. Circulation. 2018;137:71–87.
7. Aleksova N, Alba AC, Molinero VM, et al. Risk prediction models for survival after heart transplantation: a systematic review. Am J Transplant. 2020;20:1137–1151.
8. Ramzy D, Rao V, Brahm J, et al. Cardiac allograft vasculopathy: a review. Can J Surg. 2005;48:319–327.
9. Langstraat M, Musters KJS, Manintveld O, et al. Coronary artery disease in heart transplantation: new concepts for an old disease. Transpl Int. 2018;31:787–827.
10. Shah KS, Kittleson MM, Kobashigawa JA. Updates on heart transplantation. Curr Heart Fail Rep. 2019;16:150–156.
11. Kim IC, Youn JC, Kobashigawa JA. The past, present and future of heart transplantation. Korean Circ J. 2018;48:565–590.
12. Daud A, Xu D, Revelo MP, et al. Microvascular loss and diastolic dysfunction in severe symptomatic cardiac allograft vasculopathy. Circ Heart Fail. 2018;11:e004759.
13. Cheng R, Kransdorf EP, Wei J, et al. Angiogenesis on coronary angiography is a marker for accelerated cardiac allograft vasculopathy as assessed by intravascular ultrasound. Clin Transplant. 2017;31 [Epub October 16, 2017]. doi: 10.1111/ctr.13069.
14. Nikolova AP, Kobashigawa JA. Cardiac allograft vasculopathy: the enduring enemy of cardiac transplantation. Transplantation. 2019;103:1338–1348.
15. Reeve J, Böhmig GA, Eskandary F, et al.; MMDx-Kidney study group. Assessing rejection-related disease in kidney transplant biopsies based on archetypal analysis of molecular phenotypes. JCI Insight. 2017;2:94197.
16. Halloran KM, Parkes MD, Chang J, et al. Molecular assessment of rejection and injury in lung transplant biopsies. J Heart Lung Transplant. 2019;38:504–513.
17. Halloran PF, Potena L, Van Huyen JD, et al. Building a tissue-based molecular diagnostic system in heart transplant rejection: the heart Molecular Microscope Diagnostic (MMDx) System. J Heart Lung Transplant. 2017;36:1192–1200.
18. Loupy A, Duong Van Huyen JP, Hidalgo L, et al. Gene expression profiling for the identification and classification of antibody-mediated heart rejection. Circulation. 2017;135:917–935.
19. Parkes MD, Aliabadi AZ, Cadeiras M, et al. An integrated molecular diagnostic report for heart transplant biopsies using an ensemble of diagnostic algorithms. J Heart Lung Transplant. 2019;38:636–646.
20. Halloran PF, Madill-Thomsen K, Aliabadi-Zuckermann AZ, et al. Many heart transplant biopsies currently diagnosed as no rejection have mild molecular antibody-mediated rejection-related changes. J Heart Lung Transplant. 2022;41:334–344.
21. Mengel M, Sis B, Kim D, et al. The molecular phenotype of heart transplant biopsies: relationship to histopathological and clinical variables. Am J Transplant. 2010;10:2105–2115.
22. Einecke G, Reeve J, Mengel M, et al. Expression of B cell and immunoglobulin transcripts is a feature of inflammation in late allografts. Am J Transplant. 2008;8:1434–1443.
23. Parkes MD, Halloran K, Hirji A, et al. Transcripts associated with chronic lung allograft dysfunction in transbronchial biopsies of lung transplants. Am J Transplant. 2022;22:1054–1072.
24. Madill-Thomsen KS, Halloran PF, Grp IS. Molecular assessment of fibrosis and steatohepatitis in the INTERLIVER Study. Am J Transplant. 2021;21:472–473.
25. Venner JM, Famulski KS, Reeve J, et al. Relationships among injury, fibrosis, and time in human kidney transplants. JCI Insight. 2016;1:e85323.
26. Halloran PF, Reeve J, Akalin E, et al. Real time central assessment of kidney transplant indication biopsies by microarrays: the INTERCOMEX Study. Am J Transplant. 2017;17:2851–2862.
27. Berry GJ, Burke MM, Andersen C, et al. The 2013 International Society for Heart and Lung Transplantation Working Formulation for the standardization of nomenclature in the pathologic diagnosis of antibody-mediated rejection in heart transplantation. J Heart Lung Transplant. 2013;32:1147–1162.
28. Stewart S, Winters GL, Fishbein MC, et al. Revision of the 1990 working formulation for the standardization of nomenclature in the diagnosis of heart rejection. J Heart Lung Transplant. 2005;24:1710–1720.
29. Lê S, Josse J, Husson F. FactoMineR: an R Package for multivariate analysis. J Stat Software. 2008;25:18.
30. Eugster MJA, Leisch F. From spider-man to hero - archetypal analysis in R. J Stat Software. 2009;30:1–23.
31. RCT. R: A languageand environment for statistical computing. 2019. Available at http://www.r-project.org/. Accessed February 3, 2022.
32. Therneau T. A package for survival analysis in R_. 2020. Available at https://CRAN.R-project.org/package=survival Accessed March 1, 2022.
33. rms: Regression Modeling Strategies. R package version 6.0-0. 2020. Available at https://CRAN.R-project.org/package=rms. Accessed February 3, 2022.
34. Ueda T, Kawakami R, Nishida T, et al. Left ventricular ejection fraction (EF) of 55% as cutoff for late transition from heart failure (HF) with preserved EF to HF with mildly reduced EF. Circ J. 2015;79:2209–2215.
35. Tsao CW, Lyass A, Larson MG, et al. Prognosis of adults with borderline left ventricular ejection fraction. JACC Heart Fail. 2016;4:502–510.
36. Yeboah J, Rodriguez CJ, Qureshi W, et al. Prognosis of low normal left ventricular ejection fraction in an asymptomatic population-based adult cohort: the multiethnic study of atherosclerosis. J Card Fail. 2016;22:763–768.
37. Famulski KS, Broderick G, Einecke G, et al. Transcriptome analysis reveals heterogeneity in the injury response of kidney transplants. Am J Transplant. 2007;7:2483–2495.
38. Loupy A, Toquet C, Rouvier P, et al. Late failing heart allografts: pathology of cardiac allograft vasculopathy and association with antibody-mediated rejection. Am J Transplant. 2016;16:111–120.
39. Clerkin KJ, Restaino SW, Zorn E, et al. The effect of timing and graft dysfunction on survival and cardiac allograft vasculopathy in antibody-mediated rejection. J Heart Lung Transplant. 2016;35:1059–1066.
40. Loupy A, Coutance G, Bonnet G, et al. Identification and characterization of trajectories of cardiac allograft vasculopathy after heart transplantation: a population-based study. Circulation. 2020;141:1954–1967.
41. Coutance G, Zouhry I, Racape M, et al. Correlation between microvascular inflammation in endomyocardial biopsies and rejection transcripts, donor-specific antibodies, and graft dysfunction in antibody-mediated rejection. Transplantation. 2021. doi:10.1097/TP.0000000000004008
42. Halloran PF, Matas A, Kasiske BL, et al. Molecular phenotype of kidney transplant indication biopsies with inflammation in scarred areas. Am J Transplant. 2019;19:1356–1370.
43. Zorn E. Effector B cells in cardiac allograft vasculopathy. Curr Opin Organ Transplant. 2019;24:31–36.
44. Loupy A, Cazes A, Guillemain R, et al. Very late heart transplant rejection is associated with microvascular injury, complement deposition and progression to cardiac allograft vasculopathy. Am J Transplant. 2011;11:1478–1487.
45. Sis B, Einecke G, Chang J, et al. Cluster analysis of lesions in nonselected kidney transplant biopsies: microcirculation changes, tubulointerstitial inflammation and scarring. Am J Transplant. 2010;10:421–430.
46. Famulski KS, Einecke G, Sis B, et al. Defining the canonical form of T cell-mediated rejection in human kidney transplants. Am J Transplant. 2010;10:810–820.
47. Mengel M, Sis B, Kim D, et al. The molecular phenotype of heart transplant biopsies: relationship to histopathological and clinical variables. Am J Transplant. 2010;10:2105–2115.
48. Halloran PF, Reeve J, Aliabadi AZ, et al. Exploring the cardiac response to injury in heart transplant biopsies. JCI Insight.2018;3:e123674.
49. Famulski KS, de Freitas DG, Kreepala C, . Molecular phenotypes of acute kidney injury in human kidney transplants. J Am Soc Nephrol. 2012;23:948–958.
50. Famulski KS, Broderick G, Einecke G, et al. Transcriptome analysis reveals heterogeneity in the injury response of kidney transplants. Am J Transplant. 2007;7:2483–2495.
51. Land WG, Agostinis P, Gasser S, . Transplantation and damage-associated molecular patterns (DAMPs). Am J Transplant. 2016;16:3338–3361.
52. Heil M, Land WG. Danger signals—damaged-self recognition across the tree of life. Front Plant Sci. 2014;5:578.
53. Parkes MD, Aliabadi AZ, Cadeiras M, et al. An integrated molecular diagnostic report for heart transplant biopsies using an ensemble of diagnostic algorithms. J Heart Lung Transplant. 2019;38:636–646.
54. Einecke G, Reeve J, Mengel M, et al. Expression of B cell and immunoglobulin transcripts is a feature of inflammation in late allografts. Am J Transplant. 2008;8:1434–1443.

Supplemental Digital Content

Copyright © 2022 Wolters Kluwer Health, Inc. All rights reserved.