Kinetics of DNases after hematopoietic stem cell transplant

Hematopoietic stem cell transplant (HSCT) involves lysis of the entire hematopoietic system, comprising ∼2 to 3 kg of cells, inside the vasculature, over a period of 8 to 10 days. Cell lysis leads to the release of toxic intracellular molecules in the circulation, including chromosomal DNA, proteins such as actin, and lipids originating in the cell membrane.^1-3 Moreover, at the time of neutrophil engraftment, nascent neutrophils release neutrophil extracellular traps (NETs) contributing to the cell-free DNA (cfDNA) burden in the circulation.^4-6 Free DNA in the circulation is recognized as danger-associated molecular patterns and causes activation of inflammation and endothelial dysfunction.^7-12 

Extracellular DNA is degraded by serum endonucleases including DNase I and DNase1L3.^13,14 DNase I is ubiquitously expressed, whereas the main source of DNase1L3 is myeloid cells.^15,16 Human serum contains inhibitors of DNases to limit their activity in a homeostatic physiologic state.¹⁷ Several studies have described decreased DNase levels and DNase activity in patients with autoimmune disorders such as systemic lupus erythematosus, inflammatory bowel disease, trauma, and COVID-19–related lung injury.^18-24 Of note, actin, which is released into the circulation during cell lysis, binds to DNase I and is a potent inhibitor of DNase I.^17,25-27 

In this study, we hypothesized that cfDNA is increased in circulation during the early transplant period, being released from lysed hematopoietic cells early during conditioning therapy, and also subsequently as NETs at the time of engraftment. We also hypothesized that DNase levels might decrease after HSCT, being consumed by the large load of cfDNA, and that those with reduced DNase levels will have higher risk of transplant-related toxicity related to increased exposure to cfDNA. Here, we present data from a large cohort of pediatric transplant recipients to address this hypothesis.

We analyzed plasma samples collected from 108 consecutive pediatric HSCT recipients who underwent allogeneic HSCT at Cincinnati Children’s Hospital Medical Center from September 2018 to September 2020 and consented to participate in an HSCT biorepository.

Medical records and clinical databases were reviewed for demographics and clinical outcomes. We reviewed supplemental oxygen requirement with particular attention to the periengraftment (days +7 to +21) period as a marker of early pulmonary injury. Supplemental oxygen requirement was defined as any supplemental oxygen requirement outside of a clinical procedure for >1 hour, including blowby oxygen because some young children receive supplemental oxygen for hypoxia by blowby because of poor tolerance of an oxygen mask. We diagnosed transplant-associated thrombotic microangiopathy (TA-TMA) according to previously published harmonized guidelines.²⁸ 

We measured DNase I (MyBioSource, MBS2515385) and DNase1L3 (MyBioSource, MBS925159) by enzyme-linked immunosorbent assay (ELISA) in plasma samples collected at baseline (before HSCT) and on days 0, 7, 14, 30, and 100 after HSCT. DNase I samples were diluted 1:500, and DNase1L3 was measured without dilution.

We quantified double-stranded cfDNA levels using the Quant-iT PicoGreen double-stranded DNA reagent and kit, as previously described.⁵ 

We measured histone H3.1 levels using the Diapharma H3.1 ELISA kit (1001-01-03) according to the manufacturer’s instructions in a subset of patients to confirm whether cfDNA at day 14 originated from NETs (n = 41). These samples were a random subset of the general cohort, chosen as consecutive patients enrolled in the bone marrow transplantation (BMT) repository without any selection criteria.

We used a plasmid DNA digestion assay adapted from Aramburu et al to assess DNase activity in plasma.²⁴ Briefly, 0.75 μg of a 4361-base-pair plasmid (pBR322) was incubated with 10% plasma in tris(hydroxymethyl)aminomethane (Tris)-buffer (10 mM Tris, 2.5 mM MgCl₂, and 2.5 mM CaCl₂; at pH 7.4) in a final volume of 20 μL for 2 hours at 37°C. Plasma proteins were then denaturized at 95°C for 10 minutes in the presence of 0.02% sodium dodecyl sulfate and 0.9% 2-mercaptoethanol. Plasmid degradation was visualized using 1% agarose gel electrophoresis for 40 minutes at a constant 120 V.

Median with interquartile range (IQR) was used to describe continuous variables, and frequencies (%) were used to describe categorical variables, respectively. Longitudinal levels were compared using paired Wilcoxon signed-rank test. Two unpaired groups were compared using nonparametric Wilcoxon rank-sum tests (equivalent of Mann-Whitney U test). Correlations were analyzed using Spearman correlation and are reported as r. Kruskal-Wallis tests were used for comparisons of groups of >2. Odds ratio was obtained from univariable logistic regression for each outcome separately. Adjusted odds ratio for cfDNA was obtained from a multivariable logistic regression model with dependent variables cfDNA, age, sex, and conditioning. P values were adjusted for multiplicity using Bonferroni correction. Analyses were performed with GraphPad Prism version 9.4.1.

Patient demographics are shown in Table 1. The study includes 108 children and young adults who underwent allogeneic HSCT. Median age was 7.7 years (range, 2.9-14); 56.5% of patients were male, and most of the patient population was White. The most common indications for HSCT were hematologic malignancy and bone marrow failure. Most patients received a myeloablative conditioning regimen, most frequently busulfan based. Most patients received calcineurin inhibitor–based graft-versus-host disease (GVHD) prophylaxis, and 30% (n = 32) also received abatacept. Overall, 42% of patients received a T-cell depleted graft without additional immune suppression (n = 45).

Thirty-one percent (n = 33) of patients developed a periengraftment oxygen requirement, 24% (n = 26) developed engraftment syndrome (ES), 28% (n = 30) developed TA-TMA, and 5% (n = 6) had acute GVHD grade ≥2. Moderate- to high-risk TA-TMA was diagnosed at median of 22 days (IQR, 12-43) after transplant.

cfDNA is released into the circulation under normal circumstances at low levels, with most coming from turnover of hematopoietic cells. cfDNA can be present in plasma in multiple forms including double-stranded DNA and histone-bound DNA.¹¹ We analyzed levels of cfDNA longitudinally to see whether conditioning chemotherapy has an association with an increase in cfDNA burden in the circulation before the expected release of NETs at time of engraftment. Our results show that levels of cfDNA remain similar to baseline at days 0 and 7 but peak at day 14 (the typical time for engraftment and release of NETs) then decrease but remain significantly above baseline at day 30 (Figure 1A). We measured histone H3 (H3.1) levels to assess circulating intact nucleosomes containing histone 3, as a marker of histone clipping and NETosis to determine whether this cfDNA came mostly from NETs.²⁹ Histone H3.1 levels showed an increase at day 14 (Figure 1B). The peak at day 14 aligns with the day 14 cfDNA peak.

We looked to see whether patient and transplant characteristics are associated with cfDNA levels. Overall, there was no significant difference, except for the source of the stem cells. Patients who received reduced-intensity conditioning had higher levels of cfDNA at baseline (supplemental Table 5). Patients who received peripheral blood stem cell grafts had higher levels of cfDNA at day 14 (median for peripheral blood stem cell: 137.1 ng/mL, bone marrow: 108 ng/mL, and cord blood: 105.6 ng/mL; P = .0008; supplemental Table 2). We also looked at the peak in cfDNA in association with clinical outcomes. Patients with ES and those who later developed TA-TMA had higher levels of cfDNA at day 14 (Table 2; Figure 1D).

We hypothesized that depletion of DNases might modify risk of later HSCT complications by exposing the endothelium to higher levels of cfDNA for longer times. We measured levels of DNase I and DNase1L3 longitudinally at baseline and days 0, 7, 14, 30, and 100 to test this hypothesis. Baseline DNase levels and levels of cfDNA varied according to diagnosis (Table 3). Patients with marrow failure had significantly lower cfDNA, DNase I, and DNase1L3 levels at baseline than those with other diagnoses. The lower level of cfDNA likely reflects reduced cellularity of the marrow leading to lower levels of baseline hematopoietic cell turnover and may contribute to the generally greater success of transplant for marrow failure compared with other diagnoses.³⁰ DNases are induced in response to cfDNA burden, likely explaining the reduced levels of DNase I and DNase1L3 in children with marrow failure as a response to the lower level of cfDNA.³¹ 

Longitudinal changes in levels of DNase I are shown in Figure 2A. Levels of DNase I are not significantly changed at day 0 (P = 1), decrease significantly at day 7 (P < .0001), then recover quickly to a level above baseline at day 14 (P < .0001). DNase I level continues to increase to a level significantly higher than baseline (before HSCT) at day 100 (P < .0001). The nadir at day 7 was present for most of the patients, regardless of patient and transplant characteristics such as diagnosis, conditioning, HLA match, HSCT source, and GVHD prophylaxis (supplemental Table 4). Patients who received myeloablative conditioning had lower levels of DNase I at day 14 (supplemental Table 5). DNase I levels were significantly but only moderately positively correlated with cfDNA levels on day 7 (r = 0.3; P = .004, Spearman correlation) but not on day 14 (r = 0.06; P = .5, Spearman correlation; Figure 2C-D). Longitudinal changes in levels of DNase1L3 are shown in Figure 2B. DNase1L3 levels also declined after HSCT, and, in contrast to DNase I, were significantly reduced as early as day 0 (P < .0001). Similar to DNase I, DNase1L3 levels reached a nadir at day 7 (P < .0001). Recovery of DNase1L3 after the nadir at day 7 was slower than recovery of DNase I, returning to baseline levels by day 30, and continuing to rise to a level significantly higher than baseline (before HSCT) at day 100 (P = .0041; Figure 2B). This slower recovery is likely related to the need for engraftment to replenish DNase1L3 because DNase1L3 is mostly expressed in myeloid cells. DNase1L3 levels were moderately positively correlated with cfDNA levels at day 7 (r = 0.2; P = .05) and day 14 (r = 0.3; P = .006; Figure 2E-F).

We considered that decrease in DNase I levels, likely due to consumption during clearance of high levels of released cfDNA during conditioning, might mean that low levels of DNase I would be associated with increased clinical evidence of endothelial injury. Surprisingly, our results showed the opposite. Patients who developed a periengraftment oxygen requirement had significantly higher DNase I levels at baseline and days 0 and 14. (Figure 3A). Patients who developed TA-TMA also had higher levels at baseline and day 14 (Figure 3B). DNase1L3 level was not associated with any HSCT complications (Figure 3C-D). To control for test multiplicity, we also performed focused analysis of day 14 DNase I levels. DNase I levels were significantly higher at day 14 in patients with periengraftment oxygen requirement and TA-TMA (Figure 4A-B). There was no significant difference in day 14 DNase I levels in patients with ES, acute GVHD grade ≥2, or those deceased within 1 year (Table 4; Figure 4).

We hypothesized that DNase I activity, as well as protein level measured by ELISA, is decreased immediately after transplant. We analyzed plasmid DNA degradation as a functional measure of plasma DNase I activity using patient samples from baseline and days 0, 7, and 14. In all patients (UPN 1041, 1043, and 1044), DNase I digestion activity, shown by the appearance of a smear of smaller DNA fragments is lowest at day 7, with a varying recovery at day 14 (Figure 5). These results are consistent with decreased DNase I activity in the early posttransplant period, raising the possibility of DNase I activity inhibition in addition to decreased protein levels.¹⁷ 

The early posttransplant period is a critical time for patients undergoing allogeneic HSCT and a time when the stage is set for later HSCT complications, as a consequence of widespread endothelial injury. In this paper, we show, to our knowledge, for the first time, that the endonucleases DNase I and DNase1L3 are depleted in the first 7 days of transplant. The kinetics of depletion and recovery differed between the 2 endonucleases. DNase I levels were sustained until day 0, dropped significantly by day 7, and had recovered to a level similar to baseline by day 14. We hypothesize that DNase I is depleted by clearing circulating cfDNA from lysis of hematopoiesis by conditioning therapy. However, as NETosis appears during engraftment, around day 14, the cfDNA burden increases in the circulation and DNase I levels and activity are not enough for cfDNA clearance, leading to a spike in cfDNA. We also hypothesize that this early depletion in the DNase I and its inhibition by actin from ongoing cell lysis and NETosis, causes increase in DNase I levels after day 14 in response to increased need for cfDNA clearance. In contrast to DNase I, DNase1L3 is depleted by day 0 after HSCT and does not recover until day 30. This difference is likely because of the main cellular source of DNase1L3 being myeloid cells, so compensatory upregulation of production during conditioning is limited, and replacement of depleted levels only occurs when myeloid engraftment is established.

Interestingly, although DNase I levels are clearly depleted during the early days after HSCT, we found that it was patients with high levels of DNase I that had increased risk of clinical evidence of endothelial injury after transplant, for example, a periengraftment oxygen requirement and TA-TMA. This finding, which is a little counter-intuitive, may indicate that higher levels of DNase I are a marker for children with higher exposure to endothelium-toxic cfDNA, which, in turn, induces higher levels of DNase I. This hypothesis agrees with our finding that patients with TA-TMA have higher levels of cfDNA, and that cfDNA levels are positively (not negatively) associated with DNase I levels. DNase1L3 showed no association with clinical outcomes, likely because the absence of myeloid cells means that induction of increased levels is not possible in the absence of myeloid cells.²⁵ 

In previous work we have shown that levels of filamentous actin are increased in the circulation early after HSCT and are associated with increased risk of endothelial injury.¹ Actin is known to bind to DNase I and render it inactive, perhaps serving as a way for a cell to store DNase I.^25,32 We hypothesize that DNase I may have reduced activity after HSCT due to increased actin binding in addition to depletion due to consumption. The combination of protein depletion and actin binding of DNase I may contribute to the delayed clearance of NETs at day 14 that we have previously shown to be closely associated with later HSCT complications. Our overall interpretation of our data is summarized in Figure 6.

Our work extends the work of others. In a well-conducted study of patients undergoing HSCT, Cheng et al showed that dynamics of host-derived cfDNA in the posttransplant period is associated with posttransplant outcomes, but the role of endonucleases in these dynamics were not studied.² Decreased clearance of cfDNA because of a combination of low DNase I levels and activity has been shown to play a role in tissue injury in patients with burns, trauma, autoimmune disorders such as lupus, inflammatory bowel disease, and sepsis.^{9,21,22,24,33} Hakkim et al²¹ showed that patients with lupus who have DNase I deficiency, or decreased DNase I activity were more likely to have severe complications such as nephritis. It has also been shown that decreased degradation of NETs by DNase I is associated with thrombotic microangiopathy outside of the setting of HSCT.²² In a very elegant experiment, Aramburu et al confirmed that chromatin upregulation in the setting of sepsis-related cell death promoted upregulation of DNase I, and that decreased DNase I activity played a role in severe COVID-19 complications due to low chromatin clearance and NETosis.²⁴ 

Aramburu et al also showed that the magnitude of actin release had different effects on DNA clearance capacity. In their study, patients with sepsis were able to compensate for increased actin release by upregulating DNase I and a protein gelsolin essential for clearance of actin.²⁴ This finding supports the idea that DNA clearance capacity can be variable depending on the underlying cause of cfDNA release. Our results also show that kinetics of DNA clearance is likely different in patients undergoing HSCT compared with patients with infection and autoinflammation, because of unusually rapid and massive hematopoietic cell death during transplant conditioning therapy. Importantly, preparation for HSCT is a scheduled event and a more controlled environment than sepsis, and we have a chance to control the injury with appropriate preventative or early intervention.

Restoring DNase I levels and DNase I activity has previously been considered as a potential therapeutic intervention in other inflammatory settings. In a study of patients with burn injuries, Dinsdale et al showed that release of actin from damaged tissue leads to decrease in DNase activity and that patients who received fresh frozen plasma had higher levels of DNase activity, suggesting repletion of DNase by fresh frozen plasma.³³ Porter et al also showed decreased inflammation and DNA levels in patients with severe COVID-19 pneumonia when they were treated with inhaled dornase-α.³⁴ Weber et al showed that use of systemic DNase I in rat models of cardiopulmonary bypass was effective in decreasing endothelial dysfunction.⁹ Our data indicate higher levels of DNase I in those with complications, so it is unclear whether additional pharmacologic administration would be beneficial.

Our data have strengths and weaknesses. This is, to our knowledge, the first work showing depletion of DNases in an HSCT population. The size of our cohort allows us to show clinical associations, but our data are limited to a single center and our results reflect only pediatric patients. Further studies are needed to expand this work to the adult transplant population. All patients at our center undergo prospective evaluation for TA-TMA during transplant, so our clinical phenotyping is strong. Our data also indicate increased risk of early lung injury with periengraftment oxygen need, but this phenotype is less clear because multiple variables may contribute to oxygen need. Further prospective work in this area is needed.

Taken together, our data indicate that important changes in DNases are seen in patients undergoing allogeneic HSCT. Further studies are needed to assess their role in HSCT outcomes, and to explore their potential for novel prophylactic and therapeutic interventions.

The authors thank the patients and their families for their contributions to this work as participants of the Bone Marrow Transplant Repository.

Contribution: N.L., L.L., S.A., and L.S. performed experiments; A.I., N.L., S.M.D., A.L., and A.W. analyzed results and created figures; and A.I., N.L., S.M.D., K.C.M., and S.J. designed the study and wrote the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Azada Ibrahimova, Division of Bone Marrow Transplantation and Immune Deficiency, Cincinnati Children’s Hospital Medical Center, 3333 Burnet Ave, MLC 11027, Cincinnati, OH 45229; email: [email protected].