Long‐term immune changes after COVID‐19 and the effect of BCG vaccination and latent infections on disease severity

Abstract

Objectives

Several years after the COVID‐19 pandemic, the impact of SARS‐CoV‐2 on immunity and the potential protective role of Bacillus Calmette–Guérin (BCG) vaccination through trained immunity remain a subject of investigation. This study aimed to determine the long‐term impact of SARS‐CoV‐2 on immune cells and the association between BCG vaccination, latent infections and COVID‐19 severity and sepsis progression.

Methods

We conducted a prospective analysis of patients who recovered from mild/severe/critical COVID‐19 (n = 97, 3–17 months after COVID‐19) and sepsis patients (n = 64). First, we assessed the impact of COVID‐19 and its severity on immune cell frequencies and expression of functional markers. Further, we analysed plasma titres of anti‐Toxoplasma gondii/cytomegalovirus/BCG antibodies and their association with COVID‐19 severity and sepsis outcome. To examine monocyte responses to secondary challenge, monocytes isolated from COVID‐19 convalescent patients, BCG vaccinated and unvaccinated volunteers were stimulated with SARS‐CoV‐2 and LPS.

Results

Post‐COVID‐19 patients showed immune dysregulation regardless of disease severity characterised by altered expression of activation and functional markers in myeloid (CD39, CD64, CD85d, CD11b) and lymphoid cells (CD39, CD57, TIGIT). Strikingly, post‐critical COVID‐19 patients showed elevated expression of CD57 in CD8⁺ T cells compared to other severity groups. A trend toward improved outcomes in BCG‐seropositive COVID‐19/sepsis patients was observed, although this may be confounded by age differences between groups. In contrast, the monocyte response to stimulation appeared unaffected by COVID‐19 severity.

Conclusion

These findings highlight the long‐term alterations of immune cells in post‐COVID‐19 patients, emphasising the substantial impact of COVID‐19 on immune function.

This study examined long‐term immune alterations in individuals who recovered from COVID‐19 and explored potential links between Bacillus Calmette–Guérin (BCG) vaccination, latent infections and disease outcomes. Persistent immune dysregulation was observed in post‐COVID‐19 patients 3–17 months after COVID‐19, characterised by altered expression of activation and functional markers in both myeloid and lymphoid cells, even in those with mild disease. While BCG seropositivity was associated with a trend toward improved outcomes in COVID‐19 and sepsis, no effect of previous BCG vaccination on monocyte responsiveness to SARS‐CoV‐2 infection was observed.

Introduction

The emergence of severe acute respiratory syndrome coronavirus 2 (SARS‐CoV‐2) and the subsequent global pandemic affected millions of people. Recent research shows that SARS‐CoV‐2 introduces long‐lasting changes to the immune system. Specifically, studies have shown that patients who recovered from COVID‐19 had a long‐term reduction of innate and adaptive immune cell numbers, ¹ changes in T and B cell subsets distribution, ² , ³ , ⁴ prolonged systemic inflammation ⁴ or inflammatory activation of monocytes. ⁵ All together potentially provide an immunological basis for the long‐term effects of COVID‐19 with impact on health and well‐being. Nevertheless, the knowledge of how varying degrees of COVID‐19 severity affect the immune system in the long term is limited.

The Bacillus Calmette–Guérin (BCG) vaccine attracted considerable attention for its potential to prevent severe COVID‐19 progression. Originally developed to protect against tuberculosis, BCG has been reported to provide off‐target heterologous protection against non‐mycobacterial infections ⁶ , ⁷ as a result of innate immune reprogramming, a phenomenon known as trained immunity (TI). ⁸ , ⁹ However, the BCG vaccination effectiveness in COVID‐19 remains controversial. For example, several early epidemiological studies found that inhabitants of countries lacking a BCG vaccination policy were more prone to SARS‐CoV‐2 infection and succumbing to COVID‐19. ¹⁰ , ¹¹ Although these studies suggested a potential protective effect of BCG vaccination, they cannot establish definitive causality because of several inherent biases. ¹² Meanwhile, several randomised clinical trials evaluating single‐dose or multiple‐dose BCG vaccination reported either protective effects ¹³ , ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ or no effect ¹⁸ , ¹⁹ , ²⁰ , ²¹ , ²² , ²³ on COVID‐19 outcomes in high‐risk patients. Finally, a post‐hoc analysis of several recent studies found a beneficial effect of BCG vaccination on the overall survival of COVID‐19 patients, ²⁴ with the emphasis on COVID‐19 severity rather than incidence of infection. ²⁵ Similarly, latent infections such as Toxoplasma gondii (T. gondii) and cytomegalovirus (CMV) were associated with an improved response of innate immune cells to unspecific pathogens in animal models ²⁶ , ²⁷ as well as in primary human monocytes. ²⁶ , ²⁸ While the molecular mechanisms are still being delineated, ²⁹ we know that TI is generally underpinned by the enhanced production of inflammatory cytokines (including TNF‐α, IL‐6, IL‐1β) following re‐exposure to unrelated pathogens. ³⁰ , ³¹

Aside from COVID‐19, the impact of BCG vaccination has also been considered in the context of other infections and even sepsis. Data suggest that the BCG vaccine elicits protection mainly against respiratory infections in the elderly ³² and neonatal sepsis in newborns of vaccinated mothers. ³² , ³³ , ³⁴ Similarly, the contribution of latent infections in the protection against infection and sepsis progression has been studied. Nevertheless, the outcomes are again mixed: latent CMV infection provided protection against bacterial pathogens Listeria monocytogenes and Yersinia pestis in a mouse model, ²⁷ and latent T. gondii infection had a long‐term impact on the phenotype and responsiveness to T. gondii re‐infection of primary human monocytes, ²⁶ which may have important implications for innate immune responses to unrelated pathogens. On the other hand, mice chronically infected with T. gondii were more susceptible to sepsis induced by caecal ligation and puncture, and patients with sepsis and T. gondii seropositivity have a high mortality rate. ²⁸ Therefore, further research is required to rule out the beneficial or deleterious role of latent infections in the outcome of sepsis or other infections.

To determine the long‐term impact of different severities of COVID‐19 on the immune cell frequencies and functionality, we established a prospective study, recruiting individuals who had recovered from COVID‐19 with different severities and performed deep immunophenotyping. Additionally, we were interested in whether previous BCG vaccination or latent infections are associated with COVID‐19 severity and sepsis survivorship. We also assessed the relationship between prior BCG vaccination and COVID‐19/sepsis progression to determine whether there indeed exists a link between BCG vaccination status, latent infections and COVID‐19/sepsis outcomes, including pneumonia severity and sepsis survivorship. Finally, we analysed isolated monocytes from COVID‐19 patients and BCG‐vaccinated/non‐vaccinated volunteers and tested their responses to lipopolysaccharide (LPS) and SARS‐CoV‐2 stimulation to investigate whether COVID‐19 or previous BCG vaccination induces enhanced pro‐inflammatory responses in monocytes.

Results

Immune dysregulation signs are present in myeloid and lymphoid immune cells of post‐COVID‐19 patients

To investigate the effect of COVID‐19 on the immune system, we performed whole blood immunophenotyping and monitored the changes in frequencies and functionality of immune cells according to disease severity (the gating strategy is shown in Supplementary figure 1). First, we focused on changes in the myeloid cell compartment in post‐COVID‐19 patients (mild n = 41, severe n = 27, critical n = 21), acute COVID‐19 patients (n = 8) and age‐matched controls (n = 10). FACS analysis of whole blood samples showed dysregulation in the frequency of total neutrophils (CD45⁺CD3⁻CD56⁻CD66b⁺) after COVID‐19 (Figure 1a), particularly in post‐mild COVID‐19 patients (38.1%) compared to the acute COVID‐19 patients (61.2%, P‐value 0.003) and controls (49.0%, P‐value = 0.017). In contrast, the frequency of total monocytes (CD45⁺CD3⁻CD56⁻CD66b⁻CD14^+/dimCD16^+/−) remained unaffected (Figure 1a). Interestingly, at the monocyte subset level, we observed a reduction in classical monocytes [CD14⁺CD16⁻; post‐severe (92.9%) vs. post‐mild (96.4%) COVID‐19, P‐value < 0.003] and an increase in non‐classical [CD14^lowCD16⁺; post‐severe (2.86%) vs. post‐mild (1.09%) COVID‐19, P‐value < 0.001] monocytes in post‐severe COVID‐19 patients compared to the post‐mild COVID‐19 group (Figure 1b), raising the possibility that COVID‐19 severity may be linked to long‐term alterations in monocyte populations.

Figure 1.

Long‐term impact of COVID‐19 and its severity on myeloid immune cells. (a) Changes in frequencies of monocytes and neutrophils observed in patients who recovered from mild (n = 41), severe (n = 27) and critical (n = 21) COVID‐19. An age‐ and comorbidity‐matched control cohort (Cntrl, n = 10) and acute COVID‐19 patients (Acute C19, n = 8) were also included. (b) Frequency of monocyte subsets: classical (CD14⁺ CD16⁻), intermediate (CD14⁺ CD16⁺) and non‐classical (CD14^low CD16⁺) monocytes found in post‐mild/severe/critical COVID‐19 patients. (c) Expression levels of functionality‐related markers in neutrophils. (d) Expression levels of functionality‐related markers in monocytes. Data were tested using Kruskal‐Wallis test followed by post‐hoc Dunn's test with Bonferroni correction. Statistically significant differences are indicated as follows: *P‐value < 0.05, **P‐value < 0.01, ***P‐value < 0.001. Cntrl, Control group; C19, COVID‐19.

To determine how COVID‐19 severity affects the activation and functional status of monocytes and neutrophils, we performed a detailed characterization based on expression (geometric mean of fluorescence—GMF) of relevant immune‐cell activation and functional markers. For function, we evaluated the expression of CD36, CD11b, CD85d, CD39, CD64 and HLA‐DR. In neutrophils (Figure 1c), we detected significantly increased CD11b expression in all post‐COVID‐19 patients compared with controls (post‐mild [704.0]/severe [594.0]/critical [806.0] COVID‐19 vs. control [155.0], P‐value < 0.001/ = 0.006/ < 0.001) and in post‐mild/critical COVID‐19 patients compared with acute COVID‐19 patients (post‐mild [704.0]/critical [806.0] COVID‐19 vs. acute COVID‐19 [374.0], P‐value < 0.001/ < 0.001). By contrast, CD39 was significantly decreased when comparing post‐COVID‐19 patients with controls (post‐mild [108.0/]/severe [100.0]/critical [120.0] COVID‐19 vs. control [176.0], P‐value < 0.001/ < 0.001/ = 0.002). Finally, CD64 expression was significantly decreased in post‐severe and post‐critical COVID‐19 patients compared with post‐mild COVID‐19 patients (post‐severe [898.0]/critical [898.0] vs. post‐mild [1108.0] COVID‐19, P‐value < 0.001/ = 0.013) and controls (post‐severe [898.0]/critical [898.0] COVID‐19 vs. control [1157], P‐value < 0.001/ = 0.023) (Figure 1c).

In monocytes (Figure 1d), we observed significantly increased CD36 (post‐mild [953.0]/severe [989.0]/critical [901.0] COVID‐19 vs. control [441.0], P‐value < 0.001/ < 0.001/ = 0.003) and CD11b expression (post‐mild [117.0]/severe [156.0]/critical [127.0] COVID‐19 vs. control [20.5], P‐value < 0.001/ < 0.001/ < 0.001) and decreased CD85d expression in all post‐COVID‐19 patients when compared to controls (post‐mild [129.0]/severe [201.0]/critical [169.0] COVID‐19 vs. control [342.5], P‐value < 0.001/ = 0.033/ = 0.033). Similarly, CD39 was significantly increased on monocytes from post‐mild COVID‐19 patients compared with controls (post‐mild COVID‐19 [349.0] vs. control [229.0], P‐value = 0.001). CD64 and HLA‐DR remained unchanged when compared to controls. These data suggest that dysregulated myeloid cells persist in post‐COVID‐19 patients regardless of COVID‐19 severity.

Having detected persistent alterations of myeloid cells in post‐COVID‐19 patients, we next aimed to deeply characterise the impact of COVID‐19 severity on lymphoid cell frequencies and function by FACS analysis (the gating strategy is shown in Supplementary figure 1). We showed that the frequencies of total NK cells (CD45⁺ CD3⁻ CD56⁺; acute COVID‐19 [1.6%] vs. control [3.8%], P‐value = 0.004) and T cells (CD45⁺ CD3⁺; acute COVID‐19 [12.7%] vs. control [21.3%], P‐value = 0.023) were dysregulated during the acute state of COVID‐19 compared controls (Figure 2a and b). In agreement with our previous observation, ³⁵ the frequencies of NK cells (Figure 2a) and T‐cells (Figure 2b) normalised in majority of convalescent patients several months after COVID‐19. However, we noticed persistently reduced NK cell frequencies in post‐critical COVID‐19 patients in comparison with controls (post‐critical COVID‐19 [2.5%] vs. control [3.8%], P‐value = 0.037). In addition, we evaluated the frequencies of specific immune cell subset (Figure 2c and d). We observed a significantly increased frequency of CD56^dim CD16⁺ NK cells in all post‐COVID‐19 groups in comparison with the acute COVID‐19 (post‐mild [93.3%]/severe [90.6%]/critical [91.9%] vs. acute [72.4%] COVID‐19, P‐value < 0.001/ = 0.040/ = 0.005 and vs. control [80.5%], P‐value < 0.001/ = 0.022/ = 0.002). In T cells, we showed persistently reduced CD4⁺ (post‐mild [60.7%] vs. post‐critical [53.3%] COVID‐19, P‐value = 0.035; post‐severe [59.4%] vs. post‐critical [53.3%] COVID‐19, P‐value = 0.047), Tregs (CD4⁺ CD127⁻ CD25; post‐mild [3.7%] vs. post‐critical [2.8%] COVID‐19, P‐value = 0.002; post‐severe [3.6%] vs. post‐critical [2.8%] COVID‐19, P‐value = 0.026) and increased CD8⁺ (post‐mild [27.0%] vs. post‐critical [38.4%] COVID‐19, P‐value = 0.007). Looking specifically at CD8⁺ T cells, we also reported an increased frequency of CD57⁺ CD45RA⁺ CD8⁺ T cells in post‐critical COVID‐19 in comparison with post‐mild/severe COVID‐19 patients (post‐critical [22.1%] vs. post‐mild [5.9%] COVID‐19, P‐value < 0.001, and vs. post‐severe [8.3%], P‐value = 0.002).

Figure 2.

Long‐term impact of COVID‐19 and its severity on lymphoid cells. Changes in frequencies of (a) NK cells and (b) T cells observed in patients who recovered from mild (n = 41), severe (n = 27), critical (n = 21) COVID‐19. An age‐ and comorbidity‐matched control cohort (Cntrl, n = 10) and acute COVID‐19 patients (Acute C19, n = 8) were also included. Frequency of main (c) NK cell and (d) T cell subsets, and CD8⁺CD57⁺CD45RA⁺ T cells (e) in post‐mild/severe/critical COVID‐19 patients. Expression levels of relevant functionality‐related markers in (f) NK cells, (g) CD4⁺ T cells and (h) CD8⁺ T cells. Data were tested using the Kruskal‐Wallis test followed by post‐hoc Dunn's test with Bonferroni correction. Statistically significant differences are indicated as follows: *P‐value < 0.05, **P‐value < 0.01, ***P‐value < 0.001, ****P‐value < 0.0001. C19, COVID‐19; Cntrl, control group.

Taken together, these results showed alterations in frequencies of NK and T cells in convalescent patients. In particular, we observed persistently lower total NK cell counts accompanied by the increase in CD56^dimCD16⁺ NK cell subset characterised by potent cytotoxic functions, suggesting a compensatory mechanism to restore NK cell function. We also demonstrated COVID‐19 severity‐dependent impact on frequencies of CD4⁺, Tregs and CD8⁺ T cells. However, further investigation will be needed to precisely determine the implications of these results.

To further evaluate activation and exhaustion status of lymphoid cells, we focussed on expression of CD57, TIGIT and CD39. In NK cells (Figure 2f), CD57 expression represented as GMF was significantly increased in post‐critical COVID‐19 patients compared with post‐mild‐COVID‐19 (post‐critical [957.0] vs. post‐mild [432.0] COVID‐19, P‐value = 0.019), acute‐COVID‐19 (post‐critical [957.0] vs. acute [395.5] COVID‐19, P‐value = 0.016) and control group (post‐critical COVID‐19 [957.0] vs. control [320.0], P‐value = 0.001). Meanwhile, TIGIT expression was significantly increased in all post‐COVID‐19 groups compared with acute COVID‐19 (post‐mild [307.0]/severe [284.0]/critical [302.0] vs. acute [192.5] COVID‐19, P‐value = 0.009/ = 0.030/ = 0.036). However, we found persistently increased TIGIT expression only in post‐mild COVID‐19 patients compared with controls (post‐mild COVID‐19 [307.0] vs. control [210.0], P‐value = 0.024). Overall, the higher expression of CD57 and TIGIT on NK cells in all convalescent COVID‐19 patients might indicate NK cells exhaustion and inhibition of their functions.

In CD4⁺ T cells (Figure 2g), we observed significantly lower TIGIT expression in post‐mild COVID‐19 patients than in acute COVID‐19 patients (post‐mild [163.0] vs. acute [222.5] COVID‐19, P‐value = 0.003) and in post‐mild and post‐critical COVID‐19 compared with controls (post‐mild [163.0]/critical [192.0] COVID‐19 vs. control [233.5], P‐value < 0.001/ = 0.022), while CD39 expression was significantly increased in post‐critical COVID‐19 patients compared with controls (post‐critical COVID‐19 [372.0] vs. control [239.0], P‐value = 0.0172). Expression of CD57 on CD4⁺ T cells was unaffected.

Finally, we analysed CD8⁺ T cells (Figure 2h), where we detected significantly increased CD57 expression in post‐critical COVID‐19 patients compared with all other groups (post‐critical [944.0] vs. acute [345.0] COVID‐19, P‐value < 0.001; vs. post‐mild COVID‐19 [413.0], P‐value < 0.001; vs. post‐severe COVID‐19 [402.0], P‐value = 0.003; vs. control [351.0], P‐value = 0.001), as well as TIGIT in post‐severe and post‐critical COVID‐19 patients compared with post‐mild COVID‐19 patients (post‐severe [344.0]/critical [337.0] vs. post‐mild [249.0] COVID‐19, P‐value = 0.002/ = 0.037). Meanwhile, CD39 was significantly increased in all post‐COVID‐19 patients compared with controls (post‐mild [246.0]/severe [258.0]/critical [286.0] COVID‐19 vs. control [141.5], P‐value = 0.006/ = 0.002/ < 0.001). The altered expression of these markers in T cells, particularly in CD8⁺, implies an immunosuppressive state of these populations.

Overall, NK and T cells are influenced by previous COVID‐19 infection, most notably the CD8⁺ compartment where perturbed TIGIT, CD39 and CD57 expression suggest dysfunctional immune status and a possible link to exhaustion and immunosuppression.

Association of plasma anti‐BCG, T. gondii and CMV IgG levels with COVID‐19 severity and sepsis survivorship

Based on extensive studies, BCG vaccination is suggested to provide non‐specific protection against unrelated pathogens. Here, we aimed to determine whether BCG seropositivity as well as latent infections (CMV, T. gondii) are associated with COVID‐19 severity. Since COVID‐19 shares many similarities with sepsis, as evidenced by the fact that critically ill COVID‐19 patients meet the Sepsis‐3 criteria ³⁶ and present infection‐associated organ dysfunction, ³⁷ we also aimed to investigate and compare the seroprevalence of BCG, CMV and T. gondii in a cohort of sepsis patients. To do so, we recruited 97 patients who had recovered from mild to critical COVID‐19 (mild n = 41, severe n = 35, critical n = 21) (Table 1) and 64 patients with sepsis (Table 2). From these cohorts, we first aimed to investigate the relationship between persistent plasma anti‐BCG IgG levels and latent infections (T. gondii, CMV) with COVID‐19 severity. We observed differences in the BCG/CMV/T. gondii seropositivity between the COVID‐19 patients' groups (Figure 3a). Specifically, more patients with undetectable anti‐BCG in plasma were reported in the severe and critical COVID‐19 groups than in the mild COVID‐19 group (60.0%, 57.1% and 34.1%, respectively). We also observed a higher frequency of patients with latent T. gondii infection in the severe (28.6%) and critical (42.9%) COVID‐19 patient groups vs. the mild COVID‐19 patient group (9.8%). We saw a similar pattern for those with a latent CMV infection (68.6% and 95.2% vs. 58.5%). In the sepsis cohort, the frequency of patients seropositive for BCG, CMV and T. gondii mirrored the rates observed in critically ill COVID‐19 patients. Although we tested statistical significance by comparing all groups, only some differences were found to be significant. (CMV: post‐critical COVID‐19 vs. post‐mild COVID‐19, P‐value 0.02; T. gondii: post‐critical COVID‐19 vs. post‐mild COVID‐19, P‐value < 0.021) or close to significant in case of anti‐BCG antibodies (P‐value < 0.053). These results should be interpreted with caution, as we found a statistically significant difference in age between the groups (post‐mild COVID‐19 vs. post‐severe COVID‐19, P‐value < 0.001; post‐mild COVID‐19 vs. post‐critical COVID‐19, P–value < 0.001). To better assess the impact of age on our findings, we calculated the effect size specifically for age. The resulting value [η² (H) = 0.41] indicates that age has a substantial effect on the observed group differences, underscoring the need for further research to validate these results. As the post‐mild COVID‐19 group is significantly younger than the other groups, we further compare only the age‐matched groups (post‐severe COVID‐19, post‐critical COVID‐19 and sepsis). The statistical analysis was performed using the Chi‐square test with Yates' continuity correction, followed by a post hoc Chi‐square test with Bonferroni correction and the effect size was calculated using Cramér's V. As anticipated, we have not found any statistically significant difference between post‐severe/post‐critical/sepsis groups.

Table 1.

Demographic characteristics of post‐COVID‐19 cohort

Recruited post‐COVID‐19 patients
	Post‐mild C19	Post‐severe C19	Post‐critical C19
Patients
n (%)	41 (42.3%)	35 (36.1%)	21 (21.6%)
Age
Median (IQR)	35.0 (30.0; 39.0)	63.0 (50.5; 75.5)*	63.0 (48.0; 69.0)**
Sex
Male [n (%)]	18 (43.9%)	14 (40.0%)	8 (38.1%)
Female [n (%)]	23 (56.1%)	21 (60.0%)	13 (61.9%)
BMI
Median (IQR)	—	26.6 (23.1; 29.9)	30.1 (25.4; 36.5)
C19 vaccination
Yes [n (%)]	40 (97.6%)	19 (54.3%)	5 (23.8%)
No [n (%)]	1 (2.4%)	16 (45.7%)	16 (76.2%)
Months after C19
Median (IQR)	6.0 (5.0; 6.0)	7.0 (6.0; 13.0)*	10.0 (8.0; 11.0)**
Charlson comorbidity index
Median (IQR)	0.0 (0.0; 0.0)	3.0 (1.0; 5.0)*	3.0 (1.0; 3.0)**
Number of comorbidities
0 [n (%)]	35 (85.4%)	16 (45.7%)	8 (38.1%)
1 [n (%)]	4 (9.8%)	5 (14.3%)	5 (23.8%)
2 [n (%)]	1 (2.4%)	11 (31.4%)	5 (23.8%)
3 [n (%)]	1 (2.4%)	1 (2.9%)	1 (4.8%)
4 [n (%)]	0 (0.0%)	1 (2.9%)	2 (9.5%)
5 [n (%)]	0 (0.0%)	0 (0.0%)	0 (0.0%)
6 [n (%)]	0 (0.0%)	1 (2.9%)	0 (0.0%)
C19 symptoms
Increased temperature
Yes [n (%)]	4 (9.8%)	1 (2.9%)	0 (0.0%)
No [n (%)]	37 (90.2%)	34 (97.1%)	21 (100.0%)
Fever
Yes [n (%)]	19 (46.3%)	24 (68.6%)	14 (66.7%)
No [n (%)]	22 (53.7%)	11 (31.4%)	7 (33.3%)
Cough
Yes [n (%)]	20 (48.8%)	23 (65.7%)	12 (57.1%)
No [n (%)]	21 (51.2%)	12 (34.3%)	9 (42.9%)
Fatigue
Yes [n (%)]	30 (73.2%)	20 (57.1%)	15 (71.4%)
No [n (%)]	11 (26.8%)	15 (42.9%)	6 (28.6%)
Loss of taste/smell
Yes [n (%)]	5 (12.2%)	7 (20.0%)	7 (33.3%)
No [n (%)]	36 (87.8%)	28 (80.0%)	14 (66.7%)
Sore throat
Yes [n (%)]	20 (51.2%)	6 (17.1%)***	3 (14.3%)
No [n (%)]	21 (48.8%)	29 (82.9%)	18 (85.7%)
Headache
Yes [n (%)]	22 (53.7%)	10 (28.6%)	10 (47.6%)
No [n (%)]	19 (46.3%)	25 (71.5%)	11 (52.4%)
Diarrhoea
Yes [n (%)]	1 (2.4%)	5 (14.3%)	2 (9.5%)
No [n (%)]	40 (97.6%)	30 (85.7%)	19 (90.5%)
Respiratory difficulty
Yes [n (%)]	8 (19.5%)	20 (57.1%) $$$	15 (71.4%)**
No [n (%)]	33 (80.5%)	15 (42.9%)	6 (28.6%)
Rash
Yes [n (%)]	0 (0.0%)	1 (2.9%)	0 (0.0%)
No [n (%)]	41 (100.0%)	34 (97.1%)	21 (100.0%)
Red irritated eyes
Yes [n (%)]	3 (7.3%)	0 (0.0%)	0 (0.0%)
No [n (%)]	38 (92.7%)	35 (100.0%)	21 (100.0%)

Continuous variables are presented as median (interquartile range—IQR). Categorical variables are presented as numbers (n) with percentages (%). Continuous variables for three groups were tested using the Kruskal‐Wallis test followed by Dunn's post‐hoc test with Bonferroni correction. Categorical variables were tested using the Chi‐square test with Yates' continuity correction.

ARD, Anaesthesiology and Resuscitation Department; C19, COVID‐19; ICU, intensive care unit.

^* P‐value < 0.05 for post‐mild C19 vs. post‐severe C19.

^** P‐value < 0.01 for post‐mild C19 vs. post‐severe C19.

^*** P‐value < 0.001 for post‐mild C19 vs. post‐severe C19.

^$$$ P‐value < 0.001 for post‐mild C19 vs. post‐critical C19.

Table 2.

Demographic characteristics of sepsis patients

Recruited septic patients	n = 64
Age
Median (IQR)	71.0 (60.5; 77.0)
Sex
Male [n (%)]	44 (68.8%)
Female [n (%)]	20 (31.2%)
BMI
Median (IQR)	28.6 (24.7; 32.40)
Charlson comorbidity index
Median (IQR)	4.0 (3.0; 6.0)
Number of comorbidities
0 [n (%)]	12 (18.8%)
1 [n (%)]	19 (29.7%)
2 [n (%)]	13 (20.3%)
3 [n (%)]	10 (15.6%)
4 [n (%)]	5 (7.8%)
5 [n (%)]	3 (4.7%)
6 [n (%)]	2 (3.1%)
Smoking
Current smoker [n (%)]	22 (34.4%)
Ex‐smoker [n (%)]	12 (18.8%)
Non‐smoker [n (%)]	23 (35.9%)
NA [n (%)]	7 (10.9%)
SOFA
Median (IQR)	10.0 (8.0; 12.0)
CRP
Median (IQR)	255.6 (161.8; 335.5)
PCT (n = 21)
Median (IQR)	6.5 (1.6; 79.7)
Leu
Median (IQR)	14.9 (10.3; 21.7)
Catecholamine dose (standard dilution)
Median (IQR)	5.5 (2.0; 12.0)
Catecholamine dose (recalculated)
Median (IQR)	0.104 (0.037; 0.269)
Lactate
Median (IQR)	1.7 (1.3; 2.9)
Bilirubin (n = 53)
Median (IQR)	14.3 (8.6; 22.9)
Kreatinin
Median (IQR)	133.0 (92.0; 219.8)
FiO2
Median (IQR)	40.0 (30.0; 55.0)
paO2
Median (IQR)	12.0 (9.9; 13.3)
Horowitz
Median (IQR)	163.5 (90.4; 263.2)
Source of sepsis
Abdominal infection [n (%)]	12 (18.8%)
Cholangitis [n (%)]	2 (3.1%)
Soft tissue infection [n (%)]	5 (7.8%)
Neuroinfection [n (%)]	4 (6.3%)
Pneumonia [n (%)]	27 (42.2%)
Pneumonia + COVID‐19 [n (%)]	3 (4.7%)
Urosepsis [n (%)]	9 (14.1%)
Other [n (%)]	2 (3.1%)
Anti‐BCG IgG
Negative [n (%)]	40 (62.5%)
Positive [n (%)]	24 (37.5%)
Median (IQR)	0.8 (0.6; 1.2)
Anti‐CMV IgG
Negative [n (%)]	9 (14.1%)
Positive [n (%)]	55 (85.9%)
Median (IQR)	110.3 (32.1; 171.9)
Anti‐T. gondii IgG
Negative [n (%)]	36 (56.2%)
Positive [n (%)]	28 (43.8%)
Median (IQR)	0.32 (0.16; 2.58)

Continuous variables are presented as median (interquartile range – IQR). Categorical variables are presented as numbers (n) with percentages (%).

BCG, Bacillus Calmet‐Guérin; BMI, body mass index; CMV, cytomegalovirus; CRP, C reactive protein; FiO₂, the fraction of inspired oxygen; PaO₂, the partial pressure of oxygen in the arterial blood; PCT, procalcitonin; SOFA, Sequential Organ Failure Assessment; T. gondii, Toxoplasma gondii.

Figure 3.

Detection of plasma antibodies against BCG/CMV/T. gondii in post‐COVID‐19 and sepsis patients. (a) Seroprevalence of anti‐BCG/CMV/T. gondii IgG in patients who recovered from mild/severe/critical COVID‐19 and in the sepsis cohort. (b) Correlation between anti‐BCG/CMV/T. gondii IgG levels and age of post‐COVID‐19 patients (Spearman's correlation). (c) Kaplan–Meier survival curves (solid lines) and 95% confidence intervals for the group of sepsis patients with detectable levels of anti‐BCG/CMV/T. gondii IgG (red) and the group with undetectable levels (blue). Log‐rank test was used to test statistical significance. BCG, Bacillus Calmet Guérin; C19, COVID‐19; CMV, cytomegalovirus; T. gondii, Toxoplasma gondii.

In addition, we found that sepsis patients who were positive for anti‐BCG IgG in plasma (IgG level > 3 U mL⁻¹) had higher 28‐day survival rates compared to those negative for BCG antibodies (P‐value 0.059; Figure 3c). In contrast, no such trend was evident in survival outcomes between patients positive or negative for plasma T. gondii and CMV antibodies (Figure 3c). Notably, further analysis identified a significant positive correlation between antibody levels for CMV and T. gondii and age (Figure 3b) but not between BCG positivity and age (Figure 3b). Thus, it is tempting to speculate that BCG seropositivity may be associated with improved survival in sepsis.

Monocyte response to SARS‐CoV‐2 remains unaffected by previous COVID‐19 or BCG

Given the observed altered expression of activation and functional markers by monocytes and a potential association of BCG seroprevalence with better disease outcomes, we next wanted to understand whether the monocyte response to microbial challenge is affected. To do so, we isolated monocytes from post‐COVID‐19 patients (n = 24) 4–10 months (median = 8 months) after COVID‐19, and stimulated them with LPS or SARS‐CoV‐2. Relative gene expression was quantified using qPCR, and the expression levels of each gene were normalised to GAPDH. Monocytes significantly upregulated IL‐1β (P‐value < 0.001), TNF‐α (P‐value < 0.001) and IL‐6 (P‐value < 0.001), gene expression after 24 hours of stimulation with LPS in comparison with unstimulated control (Figure 4a). Similarly, the expression of IL‐1β (P‐value < 0.001), and IL‐6 (P‐value = 0.0104), was upregulated after infection with SARS‐CoV‐2 in comparison with unstimulated control. These data suggest that the ability of monocytes to respond to microbial stimuli remains intact post COVID‐19. The magnitude of ex vivo response of patients’ monocyte to stimulation was not affected by COVID‐19 severity (Supplementary figure 3a).

Figure 4.

Long‐term impact of COVID‐19 and BCG vaccination on monocyte response to microbial triggers measured at mRNA level. (a) Inflammatory gene expression measured by qPCR in post‐COVID‐19 patients triggered with LPS and SARS‐CoV‐2. (b) Impact of BCG vaccination in childhood on monocyte response to LPS, native‐ (NAT), opsonised‐ (OPS), inactivated‐ (IA) SARS‐CoV‐2, and zymosan (ZYM). Data were tested using Kruskal Wallis followed by post‐hoc Dunn's multiple comparison test. Statistically significant differences are indicated as follows: *P‐value < 0.05; ***P‐value < 0.001. BCG, Bacillus Calmet Guérin; BCG, individuals have been vaccinated with BCG vaccine in childhood; non‐BCG, individuals without previous BCG vaccination.

To investigate the possible link between BCG‐induced unspecific protection against SARS‐CoV‐2, several early epidemiological studies found that inhabitants of countries lacking a BCG vaccination policy were more prone to SARS‐CoV‐2 infection and succumbing to COVID‐19. Thus, we aimed to determine the association between previous BCG vaccination given in childhood and the monocyte response to microbial challenge. Because monocytes are short‐lived immune cells, the long‐term effects of TI arise from the reprogramming of progenitor cells. ³⁸ Therefore, we simultaneously isolated monocytes and CD34⁺ haematopoietic stem and progenitor cells (HSPCs) from peripheral blood mononuclear cells (PBMCs) of BCG‐vaccinated (n = 8) and non‐vaccinated (n = 8) individuals. We stimulated isolated monocytes with either native, opsonised, or inactivated SARS‐CoV‐2, LPS and zymosan and measured IL‐6, IL‐1β and TNF‐ɑ expression levels. We found no differences in the monocyte response to any stimuli when comparing BCG‐vaccinated and non‐vaccinated groups (Figure 4b). However, when we evaluated the monocyte response regardless of the BCG vaccination, we found that TNF‐ɑ was significantly upregulated by monocytes exposed to inactivated SARS‐CoV‐2 (P‐value = 0.002) but not native or opsonized SARS‐CoV‐2 in comparison with control (Supplementary figure 3b). This finding implies that mechanism of virus recognition could be important in shaping the magnitude of the immune response.

Finally, we were interested in determining the differentiation capacity of isolated HSPCs. We saw no significant differences in terms of numbers of the colonies formed (Supplementary figure 4a and b) or their phenotypes (Supplementary figure 4c) when comparing the HSPCs from the vaccinated vs. unvaccinated individuals. As such, we can conclude that the hypothesised TI effect of BCG vaccination administered in childhood is not preserved in adulthood.

Overall, our results revealed long‐term alterations of myeloid and lymphoid immune cells in post‐COVID‐19 patients, underscoring a profound long‐term impact of COVID‐19 on the immune system. Furthermore, we observed lower BCG seropositivity together with a higher rate of latent infections (CMV, BCG) in severe/critical post‐COVID‐19 patients, as well as sepsis patients, although these findings may be confounded by age differences between the post‐COVID‐19 groups. Moreover, our data do not show a significant relationship between previous BCG vaccination and improved responsiveness of monocytes to SARS‐CoV‐2 infection.

Discussion

Currently, the understanding of how COVID‐19 severity impacts long‐term immunity remains limited. Here, we provide a comprehensive characterisation of immune cells in the patients who recovered from mild, severe, and critical COVID‐19 (Figure 5). We observed that immune dysregulation persisted months after COVID‐19. While our analysis showed only minor changes in frequencies of innate cells among the post‐COVID‐19 cohorts, the detailed analysis of activation markers revealed persistent significant differences in comparison to controls, showing that regardless of the severity the COVID‐19 has a profound impact on the immune system several months after recovery. Specifically, profound changes in the neutrophil population among severe COVID‐19 patients were associated with increased CD11b and decreased CD39 and CD64 expression. These changes in CD11b indicate dysregulation in phagocytic capacity as well as in adhesion and migration, ³⁹ while decreased CD39 indicates alleviation of anti‐inflammatory potential. ⁴⁰ We hypothesise that reduced CD64 expression on neutrophils could be associated with receptor shedding, since this receptor was reported as highly upregulated during acute COVID‐19. ⁴¹

Figure 5.

This study examined long‐term immune alterations in individuals who recovered from COVID‐19 and explored potential links between BCG vaccination, latent infections, and disease outcomes. Persistent immune dysregulation was observed in post‐COVID‐19 patients 3–17 months after COVID‐19, characterised by altered expression of activation and functional markers in both myeloid and lymphoid cells, even in those with mild disease. While BCG seropositivity was associated with a trend toward improved outcomes in COVID‐19 and sepsis, no effect of previous BCG vaccination on monocyte responsiveness to SARS‐CoV‐2 infection was observed.

Monocytes from all COVID‐19 patients exhibited dysregulated expression of several markers in comparison to controls. Specifically, increased expression of CD11b suggests an impact on phagocytic function, adhesion, and migration, and elevated CD36 expression points toward impaired lipid metabolism ⁴¹ and senescence, ⁴² while decreased CD85d might indicate persistent activation status. ⁴³ Notably, rescued HLA‐DR expression in convalescent patients indicates functional antigen presentation capacity post‐COVID‐19 regardless of disease severity.

We have also analysed the phenotype of adaptive immune cells showing long‐term alterations. T cell frequency returned to those of healthy controls in the majority of convalescent COVID‐19 patients within several months, which aligns with our previous findings. ³⁵ However, the frequency of NK cells remained lower in patients who recovered from critical COVID‐19 in comparison to the control group. Similarly, the reduction in NK cell frequencies during acute COVID‐19 was followed by slow recovery in post‐COVID‐19 patients. ⁴⁴ In addition, we showed a persistently higher frequency of CD56^dimCD16⁺ NK cells across all post‐COVID‐19 patients in comparison to the control, which is in agreement with a recent study. ⁴⁴ Regarding T cell subsets, we reported lower CD4⁺ and Tregs shifting toward increased CD8⁺ T cell frequencies in post‐critical‐COVID‐19 patients compared to post‐mild COVID‐19 patients, aligning with studies showing the vital role of CD8⁺ T cells in COVID‐19 resolution. ⁴⁵ , ⁴⁶

Furthermore, we observed changes in the expression of activation and exhaustion markers in lymphoid cells. Elevated expression of CD57, TIGIT and CD39 on CD8⁺ T cells suggests terminal differentiation and exhaustion, indicative of potential immune dysfunction. ⁴⁷ , ⁴⁸ , ⁴⁹ Notably, CD57⁺ expression on CD8⁺ T cells was exclusively elevated in post‐critical COVID‐19 patients in comparison to post‐mild/severe COVID‐19 patients and controls, indicating a significant impact of disease severity on CD8⁺ T cells and their function. In addition, these findings suggest a possible link with enhanced immune ageing in survivors of critical COVID‐19. ⁴¹

While the role of BCG in TI is well established, ³⁵ the possible contribution of CMV and T. gondii infections to TI development remains elusive. We observed differences in seropositivity for CMV, T. gondii and BCG antibodies across all studied cohorts. Interestingly, patients who suffered from severe or critical COVID‐19 were most frequently positive for CMV and T. gondii; however, they exhibited the lowest BCG seropositivity. We observed a similar trend in CMV, T. gondii, and BCG seropositivity in sepsis patients. Further analysis revealed a positive correlation between CMV or T. gondii positivity and age, confirming previous reports. ⁵⁰ , ⁵¹ , ⁵² , ⁵³ However, we were unable to distinguish between the effects of disease severity and age because of the significant age difference between the post‐mild COVID‐19 group and the other groups. With the overall seroprevalence of T. gondii in the Czech Republic reported at ~23–32%, ⁵² , ⁵⁴ we speculate that latent T. gondii infection may similarly influence COVID‐19 severity, as previously noted by Flegr et al., ⁵⁵ although indirect effects cannot be excluded. ⁵⁶ CMV and T. gondii seroprevalence varies by country and age, ⁵⁰ , ⁵¹ , ⁵² , ⁵³ reaching 50–90% ⁵⁰ and 10–80% globally, ⁵⁷ , ⁵⁸ respectively. Thus, throughout the lifetime, CMV affects most individuals, causing lifelong latent infection with occasional episodes of reinfections, and T. gondii follows a similar aetiology. In contrast, we showed that anti‐BCG IgG levels did not correlate with age, indicating a potential age‐independent protective effect. Rather, our data point to a potential association between persistent anti‐BCG antibodies in plasma and improved survival of sepsis patients, which corresponds with findings that BCG vaccination may protect against severe neonatal sepsis. ⁵⁶ , ⁵⁹ , ⁶⁰ However, similar studies in adults are lacking, and larger clinical trials are needed to confirm these observations and inform preventive strategies. Both infections (CMV, T. gondii) have been studied in the context of sepsis severity ²⁶ , ²⁸ , ⁵⁸ or TI ⁶¹ and also for their potential risk of reactivation and reinfection of vulnerable immunosuppressed sepsis patients. ⁶² Moreover, CMV has a profound impact on immune system composition and function, including the accumulation of CMV‐specific memory T cells (memory inflation), particularly terminally differentiated T cells associated with immunosenescence. ⁵⁰

To assess the impact of TI induced by SARS‐CoV‐2 and BCG vaccination, we exposed monocytes from post‐COVID‐19 patients (mild n = 8, severe n = 5, critical n = 11, median time after COVID‐19 = 8 months), BCG‐vaccinated (n = 8) and unvaccinated volunteers (n = 8) to LPS, zymosan, and various forms of SARS‐CoV‐2. No significant differences in IL‐1β, TNF‐α, or IL‐6 production were observed in response to these stimuli in comparison to unstimulated control, suggesting that neither COVID‐19 nor previous BCG vaccination has an effect on monocyte responsiveness. Our findings are supported by a recent study showing that administration of BCG vaccine at birth does not have a protective effect against COVID‐19. ⁶³ Interestingly, TNF‐α levels were elevated in monocytes exposed to inactivated vs. opsonised SARS‐CoV‐2, emphasising the importance of SARS‐CoV‐2 recognition pathways in monocyte responses. This finding might be particularly relevant for vaccine development. ⁶⁴

This single‐centre study has also some limitations. The major limitation of our study is a relatively small sample size of sub‐cohorts based on COVID‐19 severity, as well as the significantly younger age of the post‐mild COVID‐19 group compared to the other groups. Furthermore, the patients were enrolled over a relatively long period (from October 2021 to September 2024), during which different variants of SARS‐CoV‐2 were circulating, which may have had an impact on the heterogeneity of the (sub)cohorts. We also do not have available clinical data assessing post‐COVID‐19 syndrome. Further studies evaluating the link between our findings showing immune dysregulation and long COVID‐19 are needed. Another limitation is that we assessed the effect of BCG administered in childhood on adults. The long‐term duration of innate immune memory (trained immunity) is still not fully understood, but it is currently assumed to last at least several months to years. Nevertheless, a clear advantage of our approach is the correlation of the outcomes directly to the BCG antibody plasma titres, as most published studies to date have relied on information about the vaccination itself without data on persisting antibody titres. ¹⁸ , ¹⁹ , ²⁰ , ²¹ , ²² , ²³ , ⁶³

In conclusion, our findings demonstrate that immunomodulatory changes in both myeloid and lymphoid cells can persist for up to 17 months following COVID‐19, even in individuals who experienced only mild illness. These alterations may be particularly relevant for all COVID‐19 survivors, especially those suffering from long‐term post‐COVID‐19 sequelae. While causal relationships cannot be established, BCG seropositivity appeared to be potentially associated with improved outcomes in both COVID‐19 and sepsis. Further research is needed to clarify the immunomodulatory roles of BCG vaccination and latent infections in disease progression.

Overall, we hope these results stimulate further research and discussions on how such accessible diagnostic markers could inform COVID‐19/sepsis patient risk stratification.

Methods

Cohort demographics

The post‐COVID‐19 patient cohort composed of 104 adults with mild to critical COVID‐19 who were recruited between October 2021 and September 2024. Seven post‐COVID‐19 patients were excluded from the analysis because of incomplete data. Disease severity classification followed the WHO guidelines. ⁶⁵ Immunophenotyping of whole blood samples was performed on 78 patients. The cohort demographic characteristics are summarised in Table 1 and a comparison of the clinical characteristics of patients with severe and critical COVID‐19 is shown in Supplementary table 1. In total, 64 patients with sepsis were recruited to the study between January 2023 and September 2024 (Table 2).

Post‐COVID‐19 patient cohort

Adult patients hospitalised at St. Anne's University Hospital in Brno, Czech Republic, with severe (n = 35, Table 1) or critical (n = 21, Table 1) COVID‐19 between October 2021 and September 2024, as well as non‐hospitalised volunteers with moderate COVID‐19 (n = 41, Table 1) were recruited 3–17 months (median = 6 months) after infection. Patients undergoing ongoing chronic immuno‐suppression therapy or oncological disorders were excluded from the study. SARS‐CoV‐2 variants were not identified. Written informed consent was obtained from all participants, and all procedures were approved by the institutional ethics committee of St. Anne's University Hospital Brno (10G/2021).

Sepsis patient cohort

Adult patients hospitalised at the Intensive Care Unit (ICU) of St. Anne's University Hospital in Brno, Czech Republic, with early septic shock were enrolled (n = 64, Table 2). Patients with chronic immuno‐suppression therapy or active oncological disease were excluded. Written informed consent was obtained from all participants, and all procedures were approved by the institutional ethics committee of St. Anne's University Hospital Brno (10G/2021).

BCG‐vaccinated and non‐vaccinated volunteer cohort

Adult volunteers vaccinated or not with BCG during childhood were recruited between August 2022 and July 2023 [BCG vaccinated: n = 8, median age (min–max) = 34 (28–41), sex (male/female) = 4/4; BCG non‐vaccinated: n = 8, median age (min–max) = 27 (23–34), sex (male/female) = 4/4]. Patients with chronic immuno‐suppression therapy or active oncological disease were excluded. Written informed consent was obtained from all participants, and all procedures were approved by the institutional ethics committee of St. Anne's University Hospital Brno (10G/2021).

Control cohort

An age‐ and comorbidity‐matched control cohort to post‐severe/critical COVID‐19 patients and sepsis cohort, consisting of individuals hospitalised at St. Anne's University Hospital in Brno 1 day before elective orthopaedic surgery [n = 10; median age (min–max) = 75 (64–83); sex (male/female) = 5/5; median BMI (min–max) = 29.2 (23.5–35.9); median Charlson comorbidity index (min–max) = 3.5 (2–8)] was also included. As a positive control for activation‐induced markers, we used stabilised blood samples from severe COVID‐19 patients in the acute phase [within 15 days of hospitalisation; n = 8; sex (male/female) = 4/4; median age (min–max) = 59 (41–85); median BMI (min–max) = 30 (23–39); median Charlson comorbidity index (min – max) = 2 (0–6)], obtained from our previous study. ³⁵ Written informed consents were obtained from all participants, and all procedures and protocols were approved by the institutional ethics committee of St. Anne's University Hospital Brno (control cohort: 11G/2021; acute COVID‐19: 6G/2022).

Blood sample processing and plasma preparation

Blood samples were processed within 2 h of collection. Stabilised blood samples were prepared by incubating 0.2 mL of whole blood with an equal volume of Whole Blood Cell Stabiliser (Cytodelics AB, Stockholm, Sweden) at room temperature for 15 min, then stored at −80°C until processing. ³⁵ Plasma was separated by centrifuging whole blood at 2500 g for 15 min at 4°C and immediately frozen at −80°C.

Detection of anti‐CMV, anti‐BCG, anti‐T. gondii

Anti‐CMV, anti‐BCG and anti‐T. gondii IgG levels in plasma were measured using commercial ELISA kits (Cytomegalovirus IgG ELISA Kit [DEIA326, Creative Diagnostics, New York City, USA], Human Anti‐Tuberculosis [BCG] IgG and Ig ELISA kit [Alpha Diagnostic Intl Inc., San Antonio, TX, USA], and EIA Toxoplasma [TestLine Clinical Diagnostics, Brno, Czech Republic]) according to the manufacturer's instructions.

Immunophenotyping of stabilised whole blood

Stabilised blood samples were processed as described above. Fc receptor (FcR) binding sites were blocked with FcR blocking solution (Miltenyi Biotec, Bergisch Gladbach, Germany) for 15 min at 4°C. Samples were stained with antibody cocktails (Supplementary table 2) for 30 min at 4°C. Sample acquisition was performed on the BD FACSymphony™ A1 (BD Biosciences, Franklin Lakes, NJ, USA). To account for batch effects, a common sample was included in each batch, and only markers without batch effects were used for downstream analysis. Data were analysed using FlowJo® v10.10 (FlowJo, LLC, Ashland, OR, USA). Gating strategies are presented in Supplementary figure 1.

Monocyte isolation from fresh blood

Peripheral blood from post‐COVID‐19 patients was collected in Lithium‐heparin tubes. Monocytes were isolated using the RosetteSep Human Monocyte Enrichment Cocktail (STEMCELL Technologies, Vancouver, BC, Canada) and gradient centrifugation with Lymphoprep (density 1.077 g mL⁻¹; STEMCELL Technologies) according to the manufacturer's recommendations. Isolated monocytes were cryopreserved until use.

Monocyte/HSPCs enrichment from PBMCs

PBMCs were isolated from the blood of BCG‐vaccinated and non‐vaccinated volunteers using gradient centrifugation with Lymphoprep (density 1.077 g mL⁻¹; STEMCELL Technologies). Monocytes were enriched with CD14⁺ microbeads (MACS, Miltenyi Biotec), and CD34⁺ haematopoietic stem and progenitor cells (HSPCs) were enriched from the negative fractions using a CD34 microbeads kit ultra‐pure (MACS, Miltenyi Biotec). Isolated cells were used for stimulation assays and methylcellulose colony‐forming unit (CFU) assays.

Monocyte stimulation

Monocytes were seeded at 1 × 10⁵ cells per well in RPMI‐1640 (Gibco, Thermo Fisher Scientific, Waltham, MA, USA) with 10% FBS and incubated overnight at 37°C with 5% CO₂. Cells were then stimulated with LPS (1 μg mL⁻¹ LPS from Escherichia coli O55, InvivoGen, San Diego, CA, USA), SARS‐CoV‐2 (omicron variant B.1.1. 529, isolate: 17577/21, MOI 0.1), opsonised SARS‐CoV‐2 (VNT serum, titre 1:320, 1 h at 37°C), inactivated SARS‐CoV‐2 using 0.1% β‐propiolactone (overnight incubation at 2–8°C, MOI 0.1), or zymosan (zymosan from Saccharomyces cerevisiae, 5 μg mL⁻¹, InvivoGen) for 24 h. Harvested cells were lysed in TriReagent (Molecular Research Center, Cincinnati, OH, USA) for RNA isolation.

RNA isolation and gene expression analysis

RNA was extracted from monocytes using TriReagent and purified with RNeasy spin columns (Qiagen, Hilden, Germany). RNA concentration and quality were assessed using a NanoDrop spectrophotometer (Agilent, Santa Clara, CA, USA). RNA was transcribed into cDNA using the High‐Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific, Waltham, MA, USA). qPCR was performed with TaqMan Gene Expression Assays (Thermo Fisher Scientific) on a LightCycler II (Roche, Basel, Switzerland). Relative gene expression was calculated as 2^−ΔCt, normalising to GAPDH. TaqMan probes used included TNF‐ɑ (Hs00174128_m1), IL‐1β (Hs01555410_m1), IL‐6 (Hs00174131_m1) and GAPDH (Hs02758991_g1).

Methylcellulose CFU assay

CD34⁺ cells (3 × 10² cells mL⁻¹) were seeded in methylcellulose media (MethoCult™ GF⁺ H4435, STEMCELL Technologies) and incubated at 37°C with 5% CO₂ for 16 days. Colonies were counted, harvested, washed and stained with antibodies targeting CD34 (APC‐eF780, clone 4H11, eBioscience, Thermo Fisher Scientific), CD11b (PE‐Cy7, clone ICRF44, Biolegend, San Diego, CA, USA), CD13 (APC, clone WM‐15, eBioscience, Thermo Fisher Scientific), CD15 (eF450, clone HI98, eBioscience, Thermo Fisher Scientific) CD235ɑ (PE, HIR2 [GA‐R2], Invitrogen, Thermo Fisher Scientific), CD14 (BV510, clone M5E2, Sony Biotechnology, San Jose, CL, USA), CD45 (PerCP‐eF710, clone HI30, Invitrogen, Thermo Fisher Scientific) and a LIVE/DEAD fixable dead stain kit (Green, Invitrogen, Thermo Fisher Scientific). Data were acquired on FACS Canto II (BD Biosciences) and analysed with FlowJo^® v10.10. Gating strategies are presented in Supplementary figure 2.

Statistical analysis

Prism^® (GraphPad Software LLC Ltd, La Jolla, CA, USA) software and R v4.2.3 (R Core Team 2021, R Foundation for Statistical Computing, Vienna, Austria) were used for statistical analyses. Data were tested for normality using the Shapiro–Wilk test and graphically by Q‐Q plots and histograms. Continuous variables were presented as median (interquartile range). Categorical variables were presented as numbers with percentages. Continuous variables for three or more groups were tested using the Kruskal‐Wallis test followed by Dunn's post‐hoc test with Bonferroni correction. For comparisons done between two groups, the Wilcoxon test was used. Categorical variables were tested using the Chi‐square test with Yates' continuity correction. Correlation was calculated using Spearman's correlation coefficient. For comparing dependent variables, the Friedman test was applied, followed by pairwise Wilcoxon post‐hoc tests with Bonferroni correction. Any deviation from the abovementioned statistical tests is described in the figure legend or the appropriate section of the results. The level of statistical significance was determined as follows: *P‐value < 0.05, **P‐value < 0.01, ***P‐value < 0.001 and ****P‐value < 0.0001. The effect size for the Kruskal‐Wallis test was calculated using eta squared based on the H statistic (small effect 0.01–< 0.06; moderate effect 0.06–< 0.14; large effect ≥ 0.14). Cramér's V was used to calculate the effect size for the chi‐square test of independence.

Author contributions

Kamila Bendíčková: Conceptualization; data curation; funding acquisition; investigation; methodology; project administration; resources; supervision; validation; visualization; writing – original draft; writing – review and editing. Ioanna Papatheodorou: Data curation; methodology; software; writing – review and editing. Gabriela Blažková: Investigation; methodology; validation; writing – review and editing. Martin Helán: Investigation; writing – review and editing. Michaela Haláková: Data curation; investigation; writing – review and editing. Petr Bednář: Investigation; methodology; writing – review and editing. Erin Spearing: Writing – review and editing. Lucie Obermannová: Data curation; investigation; writing – review and editing. Julie Štíchová: Investigation; writing – review and editing. Monika Dvořáková Heroldová: Investigation; writing – review and editing. Tomáš Tomáš: Investigation; writing – review and editing. Roman Panovský: Investigation; writing – review and editing. Vladimír Šrámek: Supervision; writing – review and editing. Marco De Zuani: Conceptualization; funding acquisition; writing – review and editing. Marcela Vlková: Investigation; writing – review and editing. Daniel Růžek: Resources; supervision; writing – review and editing. Marcela Hortová‐Kohoutková: Investigation; methodology; supervision; writing – original draft; writing – review and editing. Jan Frič: Conceptualization; funding acquisition; resources; supervision; writing – original draft; writing – review and editing.

Conflicts of interest

MDZ is an employee and owns stock in Ensocell Therapeutics (Cambridge, UK). The other authors declare no conflict of interest.

Supporting information

Supporting information

Data availability statement

The data supporting this study are available from the corresponding author upon reasonable request. Because of legal and ethical restrictions, only aggregated data can be shared.