The Pivotal Role of Immune Functional Assays in Deciphering Immune Function Alterations

3.1 Study population

3.2 Sample collection

For each study, at the time of sampling, heparinized-whole blood was collected and incubated in TruCulture^® tubes (Myriad Rbm, Austin, TX, USA) pre-filled with either LPS or SEB, as previously described (4,11). Unstimulated whole blood samples were also collected in PaxGene^™ Blood RNA tubes (PreAnalytiX/QIAGEN Inc., Valencia CA, USA) and stored at -80°C.

3.3 Isolation of RNA from PaxGene whole-blood

Total RNA was extracted from whole-blood samples using the PaxGene™ Blood RNA Kit (PreAnalytiX/QIAGEN Inc., Valencia CA, USA). After thawing, blood RNA tubes were centrifuged at 3,000g for 10 min at room temperature (RT). The supernatant was decanted, followed by the addition of 4 ml RNase-free water and redissolving of the pellet. Next, the tube was centrifuged again at 3,000 g for 10 min at RT, and the supernatant was discarded. The pellet was resuspended in 350 µl of buffer BR1 and transferred to a microcentrifuge tubeNext, 300 µl of buffer BR2 and 40 µl of proteinase K were added, mixed, and incubated for 10 min at 55°C using a shaker-incubator at 450 rpm. This lysate was transferred to a PaxGene shredder (Qiagen, Valencia, CA) spin column and centrifuged at 20,000 g for 3 min at RT. The supernatant of the flow-through fraction was transferred to a fresh microcentrifuge tube, and 350 µl of 100% ethanol was added and mixed. Then 700 µl of this sample was transferred to a PaxGene RNA spin column and centrifuged at 10,000g for 1 min at RT. The flow-through was discarded, and this step was repeated with the remaining sample from the previous step.

Next, 350 µl of buffer BR3 was added onto the RNA spin column, which was centrifuged at 10,000g for 1 min at RT. The flow-through was discarded, and 10 µL of DNase 1 was mixed with 70 µl of buffer RDD in a separate tube. This was pipetted onto the spin column membrane and incubated at 20–30°C for 15 min. Then 350 µl of buffer BR3 was added onto the RNA spin column, which was centrifuged at 10,000 g for 1 min at RT. The flow-through was then discarded. Next, 500 µl of buffer BR4 was added onto the RNA spin column, which was centrifuged at 10,000 g for 1 min at RT. The flow-through was discarded, and 500 µl of buffer BR4 was added again onto the RNA spin column and centrifuged at 10,000 g for 3 min at RT. The flow-through was discarded. The RNA spin column was transferred to a 1.5-ml microcentrifuge tube, and 40 µl of buffer BR5 was pipetted onto the column to elute the membrane-bound RNA. The column was then centrifuged at 10,000 g for 1 min at RT. This step was repeated with the same 40-µl buffer BR5, and the eluate was incubated at 65°C for 5 min, then chilled on ice. RNA concentration was estimated using the Nanodrop One spectrophotometer (Thermo Scientific, Swedesboro, NJ) according to the manufacturer’s instruction.

3.4 Gene expression analysis

Gene expression was evaluated through a 89-(11) and 144-gene panel (4) using the NanoString^® technology for the sepsis and allo-HSCT study, respectively. Briefly, 300 ng of RNA were hybridized to the probes at 67 °C for 18 h using a thermocycler (Biometra, Tprofesssional TRIO, Analytik Jena AG, Jena, Germany). After removal of excessive probes, samples were loaded into the nCounter Prep Station (NanoString^® Technologies, Seattle, WA, USA) for purification and immobilization onto the internal surface of a sample cartridge for 2–3 h. The sample cartridge was then transferred and imaged on the nCounter Digital Analyzer (NanoString^® Technologies) where color codes were counted and tabulated for each panel of genes. Data treatment and normalization were performed on nSolver^® analysis software (version 4.0, NanoString^® Technology) using internal controls and 3 housekeeping genes (detailed in Table S1 and S2). Of note, for both cohorts, analyses were conducted using genes with expressions exceeding the background noise threshold in more than 75% of individuals, as previously described (4,11).

3.5 Statistical Analysis

Normality testing was performed using the Shapiro-Wilk normality test. The distribution of quantitative data was expressed as mean (range) or median (interquartile range, [IQR]) where appropriate. Cluster analyses were performed using the PAM method with correlation and average distance, as previously described by Albert-Vega et al. (11). The alluvial plot was obtained via http://www.bioinformatics.com.cn, a free online platform for data analysis and visualization. Principal component analysis (PCA) was carried out using Partek Genomics Suite software (version 7.0; Partek Inc., St. Louis, MO), and Euclidean distances were calculated as previously described by Mouton et al. (4). Differences in Euclidean distances between groups were calculated using a non-parametric unpaired Wilcoxon test with Benjamini correction. Differences in standard deviations (SD) were calculated using paired Pitman-Morgan test. Statistical analyses were conducted using GraphPad Prism software (version 8; GraphPad software, La Jolla, CA, USA) and R (version R.2.2). P values and adjusted P values (P adj) < 0.05 were considered significant.

4.1 Immune function assessment appears as an essential tool to obtain clinically relevant clustering in the context of sepsis

Overall, 28 out of the 30 patients with sepsis as well as 10 HVs initially included in the Albert-Vega et al. study were analyzed herein, as 2 unstimulated samples had not been collected. Among these patients, 7 developed a hospital-acquired infection (HAI) during ICU stay, and 3 had died at day 28 (Table S3). Following the data control and normalization steps described above, 81 out of 86 genes initially analyzed were kept for analyses through PaxGene™ tubes. To evaluate the added value of the stimulation in identifying immune functional alterations in a sepsis population, statistical analyses and clustering must be conducted using an equal number of patients and genes, whose expression was measured post-stimulation or from PaxGene™ tubes without stimulation. To that end, we first aimed to validate that the conclusions derived from the restricted post-stimulation dataset were consistent with those obtained from the complete dataset published in the Albert-Vega et al. study.

We thus conducted again the multivariate clustering analysis from gene expression levels post-SEB stimulation using 81 genes, 28 patients with sepsis, and 10 HVs. The analysis found 3 clusters with the same composition as previously described by Albert-Vega et al. (Figure 1 left, Table S4). The first cluster (n=11) grouped together all the HVs and one patient with sepsis, constituting the healthier cluster, gathering immunocompetent individuals. The second cluster (n=13) included all non-survivors, hence designated as the severe cluster. It was characterized by a diminished immune responsiveness upon in vitro SEB stimulation and a specific modulation of genes previously described to be associated with mortality, such as MDC1 and IFIT44L. The third cluster (n=14) included 86% (6/7) of the patients with sepsis who developed a HAI, forming the intermediate cluster. As demonstrated by Albert-Vega et al., these patients exhibited, among other, an upregulation of the HLA family and interferon-related genes, suggesting a potential for immune recovery, implying that patients identified in this cluster may benefit from immunostimulatory therapy.

• Download figure
• Open in new tab

FIGURE 1.

Having confirmed the consistency of the conclusions that can be drawn from the restricted dataset, we conducted the same analyses on PaxGene™ tubes (Figure S1.A). Using the dataset obtained from PaxGene™ tubes, we identified an equivalent of the healthier cluster comprising all HVs, consistent with observations from the post-stimulation gene dataset. However, in the absence of stimulation, this healthier cluster included 4 of 9 patients with sepsis who had been classified within the severe cluster identified post-SEB stimulation, including patients who exhibited an evident and profound alteration of immune function post-stimulation (Figure 1 right). This discrepancy demonstrates the loss of ability to distinguish immunocompetent individuals among a sepsis population based on gene expression analysis from unstimulated whole blood.

Altogether, these results highlight a notable discrepancy in patient stratification according to immune alteration profiles when determined with or without in vitro stimulation, and suggest that, during the course of sepsis, immune function assessment can reveal distinct immune profiles that are coherent with clinical characteristics.

4.2 Immune function assessment remains necessary in the context of allo-HSCT to uncover immune function alterations

We then evaluated the added value of the stimulation to reveal immune functional alterations in another clinical context, e.g. during the immune reconstitution after allo-HSCT. For this study, 59 of the 60 allo-HSCT recipients and 5 of the 10 HVs initially included were analyzed, as one patient sample did not pass quality control, and no PaxGene™ tubes were available for 5 HVs. Regarding the hematological- and transplant-related characteristics of allo-HSCT recipients included at a median [IQR] of 6.5 [5.8-8.3] months after transplantation, 52.5% had been transplanted due to acute myeloid leukemia, 35.6% had active graft-versus-host disease (GvHD) at inclusion, and 32.2% were undergoing immunosuppressive treatment at inclusion (Table S5).

Following the data control and normalization steps, 121 out of 138 and 134 genes initially analyzed post-SEB and -LPS stimulation, respectively, were kept for analyses through PaxGene™ tubes. As for the sepsis study, we first aimed to confirm whether the same conclusions obtained from the entire post-stimulation transcriptomic dataset used in the Mouton et al. study could be replicated using the new restricted dataset.

We thus conducted again the PCA from gene expression levels post-stimulation with either LPS or SEB using 121 genes, 59 allo-HSCT recipients, and 5 HVs. Similarly, post-stimulation gene expression analysis revealed a strong homogeneity among HVs, while allo-HSCT recipients represented a more heterogeneous population (Figure S2). For quantitative purposes, we once again calculated, using the PCA projection from the restricted dataset, the Euclidean distance of each allo-HSCT recipient to the centroid of the HV population, which serves as a reference value for a functional immune response. We hypothesized that a greater distance corresponds to a more impaired immune response and assessed the association between these distances and clinical data. Thus, using this restricted post-stimulation dataset, we were able to confirm an increase in Euclidean distance associated with an ongoing immunosuppressive treatment (median [IQR] 9.71 [4.91-12.6] versus 13.77 [11.34-18.43] P adj < 0.047, Table S6) (4).

We then proceeded to conduct the same analyses on PaxGene™ tubes (Figure S1.B). Whether obtained post-stimulation or under unstimulated conditions, the immune profiles projected onto the first 2 principal components of the PCA revealed similar overall variability, explaining 36.6% and 36.7% of the overall variance, respectively (Figure 2.A). Nevertheless, as illustrated in the PCA overlay, the immune functional heterogeneity of allo-HSCT recipients, captured using gene expression analysis through unstimulated whole blood, was less pronounced. This was supported by two indicators, the inertia of the point cloud formed by the allo-HSCT immune profiles and the SD of the Euclidean distance (Figure 2.B), which were both highly reduced in the unstimulated condition compared to the stimulated one (inertia: 39.4 vs. 87.5 and SD: 3.814 vs. 6.977, p-value <0.001). In addition, neither ongoing immunosuppressive treatment (Figure 2.C) nor any other parameter (Table S7) could account for the significant increase in the Euclidean distance using the gene expression dataset obtained without whole blood stimulation, in contrast to that obtained post stimulation. It appears that in the context of allo-HSCT, immune function assessment through IFA uncovers a treatment-dependent immune response, which could help improve the monitoring of post-transplantation immune reconstitution by complementing the classical markers currently used.

• Download figure
• Open in new tab

FIGURE 2.