Circulating acetylated polyamines correlate with Covid-19 severity in cancer patients

Abstract

Cancer patients are particularly susceptible to the development of severe Covid-19, prompting us to investigate the serum metabolome of 204 cancer patients enrolled in the ONCOVID trial. We previously described that the immunosuppressive tryptophan/kynurenine metabolite anthranilic acid correlates with poor prognosis in non-cancer patients. In cancer patients, we observed an elevation of anthranilic acid at baseline (without Covid-19 diagnosis) and no further increase with mild or severe Covid-19. We found that, in cancer patients, Covid-19 severity was associated with the depletion of two bacterial metabolites, indole-3-proprionate and 3-phenylproprionate, that both positively correlated with the levels of several inflammatory cytokines. Most importantly, we observed that the levels of acetylated polyamines (in particular N₁-acetylspermidine, N₁,N₈-diacetylspermidine and N₁,N₁₂-diacetylspermine), alone or in aggregate, were elevated in severe Covid-19 cancer patients requiring hospitalization as compared to uninfected cancer patients or cancer patients with mild Covid-19. N₁-acetylspermidine and N₁,N₈-diacetylspermidine were also increased in patients exhibiting prolonged viral shedding (>40 days). An abundant literature indicates that such acetylated polyamines increase in the serum from patients with cancer, cardiovascular disease or neurodegeneration, associated with poor prognosis. Our present work supports the contention that acetylated polyamines are associated with severe Covid-19, both in the general population and in patients with malignant disease. Severe Covid-19 is characterized by a specific metabolomic signature suggestive of the overactivation of spermine/spermidine N₁-acetyl transferase-1 (SAT1), which catalyzes the first step of polyamine catabolism.

Keywords: COVID-19; cancer; metabolomics; microbiota; polyamines.

Figure 1

Heatmap representing the serum metabolome of each individual cancer patient clustered by clinical severity of Covid-19. Targeted metabolomic data on 211 serum samples from 204 patients were normalized areas of identified metabolites. Results are listed in Supplementary Table 1.

Figure 2

Identification of metabolites discriminating cancer patients by clinical severity of SARS-CoV-2. (A) Volcano plot comparing mild with moderate/severe Covid-19 patients, classified by chemical classes. X-axis: log² fold change of metabolites; Y-axis: fold change of –log¹⁰ P value determined by the Mann–Whitney test. (B) Random forest classification model based on metabolites altered (p < 0.05) between mild and moderate/severe Covid-19 cases. The downregulated and upregulated metabolites in mild compared to moderate/severe patients are marked in blue and red, respectively. (C) Indole-3-propionic acid and 3-phenylproprionic levels in patients. All boxes indicate the interquartile range Q1 to Q3 with Q2 (median levels) in the center. The range of outliers is depicted by whiskers. The black bar of the figure indicates the p-value.

Figure 3

Acetylated polyamine derivatives and associated metabolites in cancer patients with different levels of Covid-19 severity. Acetylated polyamine derivatives were identified by targeted metabolomics data (A). S-adenosylhomocysteine and S-adenosylmethionine are shown (B). All data represent the normalized areas of mass spectrometric peaks and were analyzed by non-parametric unpaired Wilcoxon test (Mann–Whitney) for each two-group comparison. The black bar of the figure indicates the p-value.

Figure 4

Aggregate analyses of acetylated polyamine derivatives in cancer patients with different levels of Covid-19 severity. For each patient, the sum of the normalized peak areas corresponding to N₁-acetylputrescin, N₁-acetylspermidine, N₁,N₁₂-diacetylspermine and N₁,N₈-diacetylspermidine were calculated and shown with black bars indicating p-values.

Figure 5

Identification of metabolites discriminating cancer patients according to the duration of SARS-CoV-2 shedding. Volcano plot comparing Covid-19 patients with short versus long viral shedding (determined by RT-PCT of nasopharyngeal PCRs, the threshold between short and long shedding being 40 days), classified by chemical classes. X-axis: log² fold change of metabolites; Y-axis: fold change of –log¹⁰ P value determined by the Mann–Whitney test.

Figure 6

Integration of metabolic with inflammatory markers in serum samples from cancer patients with different levels of Covid-19 severity. The correlation heatmap of data was generated by means of Pearson’s method, and clustered using the ward. D2 method. The red color indicates positive correlations with a FDR<0.05, and the green color marks negative correlations with FDR<0.05. Black indicates non-significant (FDR>0.05) associations. *p < 0.05, **p < 0.01, ***p < 0.001. TNFα: tumor necrosis factor alpha. GzB: Granzyme B. TNFα on GzB in CD8 is the ratio between soluble TNFα and GzB in CD8+ T cells (by flow cytometry) and IFNγ on GzB in CD8 is the ratio between soluble IFNγ GzB in CD8+ T cells (by flow cytometry).

Figure 7

Correlation among metabolites and inflammatory markers. Correlations of (A) 3−phenylpropionic acid and tumor necrosis factor alpha (TNFα), (B) 3−phenylpropionic acid and indole-3-phenylpropionic acid, (C) indol-3-propionic acid and N₁-acetylspermidine or (D) N₁,N₈-diacetylspermidine or (E) N₁,N₁₂-diacetylspermine.

Figure 8

Anthranilic acid levels in different patient cohorts. Barplots demonstrating the anthranilic acid area levels for each subject issued from the cancer-free and Covid-19-free control cohort (COCHIN Control), previously described by Danlos et al., [45], and our dataset of cancer patients without Covid-19 (ONCOVID controls) and the cancer patients with moderate or severe Covid-19 (ONCOVID Moderate/Severe). Error bars show standard errors of the mean.