Systemic lupus erythematosus (SLE) is the prototypic multi-organic autoimmune disease, and it implies high morbidity and early death in young and productive people (1). The immunopathology of SLE encompasses multiple innate and adaptive immunologic alterations, including hypocomplementemia, higher levels of TNF-α, IL-4, IL-6, IL-10, as well as type I and II interferons, with a consequent skewing toward a Th17 response, persistent B-cell activation with sustained auto-antibody secretion and a deficient regulatory T cell profile (2).

Patients with SLE also have higher amounts of low-density granulocytes (LDGs), which infiltrate tissues, secrete pro-inflammatory cytokines and spontaneously produce neutrophil extracellular traps (NETs) (3). Aside from the role of NETosis in the SLE pathogenesis, the NETs are an innate defense mechanism since they contain antimicrobial proteins including LL-37 (3).

Moreover, previous studies have shown that patients with SLE have expansion of the pro-inflammatory intermediate monocytes in peripheral blood (4). Toll-like receptor 2 (TLR2) is one of the mainly expressed pattern recognition receptor in monocytes, has diverse ligands (5), and it is pivotal in the SLE pathophysiology since it promotes an interferogenic response and has been shown to augment disease activity in an animal model of SLE through the recognition of the bacterial curli (5).

Infections are one of the main causes of hospital admissions in patients with SLE (6), and they represent one of the three major causes of death, along with renal and cardiovascular diseases (7). In patients with SLE, infections are traditionally considered a complication of immunosuppressive therapy, as was confirmed in a recent retrospective cohort study (8) and therefore, most of the infection prophylactic measures in SLE are directed at patients using immunosuppressive therapy (9). Nevertheless, 25.9% of severe infections in patients with SLE occur at diagnosis in the absence of immunosuppressive therapy (10) and previous studies have described disease activity and number of flares as independent risk factors for infections (11), which suggests an inherent infection predisposition that may be dependent upon the immunologic abnormalities that characterize the disease (2).

Currently, there is not an available tool to identify SLE patients at high risk of infection, regardless of steroid treatment in the routine clinical scenario. Thus, the development of an infection prediction index that includes the clinical and immunological features of SLE patients is crucial to identify a group at higher risk to develop this complication. The aim of this study was to prospectively create and validate a compound clinical-immunological index that was useful to predict the development of infections in SLE patients in the following year.

Seventy-five percent of patients were women. Median time since SLE diagnosis was 13 months (IQR 4-28). The main clinical manifestations at SLE diagnosis were articular (84%), mucocutaneous (81%), renal (50%), hematologic (37%), serositis (37%), constitutional (35%), neurological (6%), and 14% had secondary anti-phospholipid syndrome.

As previously described, when compared to healthy controls, patients with SLE had a diminished count of lymphocytes (25) and regulatory T cells (26) and a higher percentage of LDGs (27), Th1, Th2 and Th17 cells (28). Moreover, patients with SLE had lower absolute number of B cells and a lower expression of TLR2 in total monocytes (Table 1).

During 12 months of follow-up, 18 patients (32%) developed 19 infectious events in a median time of 21.5 weeks (IQR 4-24), 5 (26.3%) were severe. The main types of infections were community-acquired pneumonia (23%), upper respiratory tract (23%) and urinary tract infections (17%), herpes zoster virus (17%), gastroenteritis (11%), and cellulitis (5%). During the study period, 6 (10.9%) patients were receiving cyclophosphamide, 4 (66.6%) of them had an infection.

As shown in Table 2 and Figure 2, patients that developed infection during the follow-up period had higher amounts of anti-dsDNA antibodies, LDGs, Th17 cells, and diminished absolute number of B cells at baseline. We found no association between the vaccination status, tobacco use and prophylactic antibiotic administration with the development of infection (data not shown).

In comparison to their baseline immunological parameters, there was a higher expression of LL-37 in LPS induced NETs (Figure 3) and a higher amount of LDGs during the infectious events (Table 3).

At the time of infection, 10 (55.5%) patients presented an SLE flare. Nevertheless, there were not statistically significant differences in the immunological parameters in comparison to patients who did not flare.

Univariate and multivariate logistic regression analyses were performed to address the association of infection with diverse clinical and immunological parameters at baseline Table 4 and in the cohort-nested case-control study Table 5. The variables associated with infection in multivariate analyses included anti dsDNA antibodies, absolute number of B cells and LDGs and the expression of TLR2 in total monocytes. Nonetheless, as described in Methods, the variables selected for the LIPI were those that were able to predict infection based on the ROC analyses and include the following: the use of cyclophosphamide in the repeated measure analysis, the absolute number of B and Th17 cells and the MFI of TLR2 in total monocytes. The cutoff points for the index were calculated with ROC curves as described in methods and are depicted in Table 6. After the cutoff points were determined, a score was assigned for each variable according to the RR Table 6 as described in methods, and we tested different variable combinations to assess their predictive capability comparing the AUC with the ROC test Table 7.

As shown in Table 7 and Figure 4, the selected index was the one with the best AUC, which was the LIPI including the 4 variables at baseline. Although the AUC of the LIPI at baseline was not different from the AUC of the combination of cyclophosphamide use and absolute number of Th17, the LIPI showed a specificity of 100% and a higher negative LR than the former combination of variables with the baseline dataset. The AUC of the LIPI at baseline, was compared with the ROC test and was higher than the LIPI in the cohort nested case control study (P = 0.0025) and the combination of CYC+Th17 in the referred dataset (P = 0.035). Also, the latter combination of variables showed lower sensitivity as well as lower negative LR.

To our knowledge, this is the first prospective study to include the assessment of clinical and immunological parameters to develop an index for infection prediction in patients with SLE. Our most relevant finding is that the imbalance of Th17 and B cells numbers and a lower expression of TLR2 in monocytes along with cyclophosphamide use are the major risk factors for infection in SLE patients.

Infections are a burden in SLE, with a frequency ranging from 19.3 to 58.7%, and a mortality of 24.5% (6, 8), which agrees with our findings. Diverse risk factors for infections have been reported, such as age at diagnosis, Hispanic ethnicity, any use of glucocorticoids (≥10 mg/day), immunosuppressors, hospitalization by SLE, renal involvement and SLE damage index (SDI); while time on anti-malarials has been found as protective (8). Among these, we only corroborated the use of immunosuppressors (cyclophosphamide) as a major risk factor for infection, similar to the study by Bosch et al. (29). The higher risk of infection in patients receiving cyclophosphamide may be secondary to its cytotoxic effect, particularly upon proliferating lymphocytes (30). Furthermore, in animal models, cyclophosphamide use promotes translocation of intestinal bacteria (30), which could be another mechanism to explain its association with infection. In contrast with previous cohort studies, we did not find a protective effect of anti-malarials, probably because of the distinct follow-up time (8) and the methodological differences, since we performed a repeated measure analysis for every immunosuppressive drug and did not take into account the time on anti-malarials.

Interestingly, although we were not able to find association between the disease activity scales and the occurrence of infection, among our relevant findings are that patients with immunological signs of high disease activity including higher levels of Th17 cells, anti-dsDNA antibodies (31) and LDGs (27) were more prone to develop infections.

Two different subsets of Th17 cells have been identified in humans. The non-pathogenic subtypes whose main function is the defense against pathogens in the mucous membranes (32) and the pathogenic Th17, which are elevated in peripheral blood in lupus (31). In SLE, disease activity could be related to an imbalance between these subsets, with a predominance of the pathogenic Th17 and lower amounts of those cells devoted to defense toward infection.

Besides, we found that patients with lower levels of B-cells developed infection during the follow-up. In SLE, naïve B cells may be initially hyper-activated and produce higher levels of anti-dsDNA antibodies but this eventually leads to necroptosis (33), which would explain the lower B cell numbers in our study. Furthermore, in comparison to healthy controls, patients with SLE have lower frequency of BCR sequences with somatic hypermutations (SHM) (34). In animal models, the reduction in naïve B cells promotes Cryptococcus dissemination (35) and the lower biodiversity of BCR sequences with SHM in patients with SLE (34) may diminish the B cell repertoire to combat infections. Also, the lower pool of naïve B cells could lead to less class-switched memory B cells, which are known to be fundamental in the prevention of infections in other clinical conditions (36). The combination of a lower naïve B cell repertoire and diminished SHM may explain the increased risk of infection in SLE patients, although further studies are required to assess if different B cell subsets are involved in infection development in these patients.

We were not able to find the amount of NETs as a predictor of infection. However, we did find a differential protein cargo characterized by increasing amounts of LL-37 upon infection, probably as a defense mechanism. Besides, this antimicrobial peptide augments the expression of co-stimulatory molecules in dendritic cells (37) and promotes the production of type I IFN (38), which could be related to the association of SLE flares and infection. Indeed, the prospective design of this study allowed us to detect a disease flare in more than 50% of patients during infections.

Regarding TLR2, patients with SLE had a 56.6% lower expression in comparison to healthy controls. Although the regulating factors of the expression of TLR2 in SLE are unknown, glucocorticoids diminish its expression in keratinocytes and respiratory epithelial cells (39, 40). Previous studies had shown that patients with SLE had lower expression of TLR2 in monocytes in comparison to healthy controls (41), but this is the first study to describe it as a major risk factor for infection in lupus patients. Besides, the reduction of TLR2 has been acknowledged as a prognostic factor in infections, such as pneumonia (42) and its diminished expression may predispose SLE patients to infections in a similar way of genetic variants. The TLR2 Arg753Gln and T597C polymorphisms are related to Gram-positive septic shock, S. aureus and cytomegalovirus infections, as well as pulmonary and meningeal tuberculosis (43).

In summary, we propose that SLE patients have lower expression of TLR2 in monocytes. As a consequence, they have lower capacity to recognize and combat pathogens, since TLR2 is fundamental in the immune response against many bacterial, fungal, and viral infections. Besides, higher disease activity leads to Th17 expansion and B cell hyper-activation, with the consequent activation-induced necroptosis. The use of cyclophosphamide along with B cell necroptosis may promote a diminished pool of circulating B cells with a consequent limitation of the B cell repertoire to recognize, neutralize and combat pathogens.

Our data suggest that the combination of the use of cyclophosphamide with the compound measurement of B cells, Th17 cells, and TLR2 expression in monocytes is useful as an infection predictive index in SLE patients. All of the parameters included in the index are readily available using conventional flow cytometry, a technique that has demonstrated diagnostic utility as a compound measurement tool (44) as well as a cost-effective strategy (45) in many other diseases in the clinical setting, particularly at third level referral centers. Also, this is the first study to emphasize the pathogenic autoimmune response in SLE as a pivotal risk factor for infections, which could explain the concomitant occurrence of high disease activity and infections at lupus diagnosis even in the absence of immunosuppressive therapy. By alerting clinicians about patients who are more prone to develop infections, there could be a prompt and closer follow-up, with a multidisciplinary approach, including counsel by experts in infectious diseases. Additionally, since one of the most relevant prophylactic measures is the use of antibiotics, it is fundamental to detect patients with higher risk of infections to lower the generalized use of this preventive action in order to avoid the selection of resistant strains as it has been shown in other chronic diseases (46).

Our study has many limitations, including that our cohort is unicentric, solely composed by Hispanic patients with < 5 years since diagnosis and without any treatment with biologic anti-rheumatic disease drugs. Besides, our sample size is limited, and this could influence the lack of association between infection and some of the variables, such as the NET amount and the use of anti-malarials. Even though, a longer follow-up time would have been desirable, our study allowed us to find association of some of the variables with infection since the outcome developed early in the follow-up period. Furthermore, we should acknowledge that some of the variables included in the LIPI, as individual items, did show a small effect size for the prediction of infection. Nevertheless, the compound use of all these variables was able to increase their predictive capacity. Also, even though flow cytometry has gained access to the routine clinical setting, still it is not globally available and its cost-effectiveness should be addressed in future studies. Finally, this tool needs to be prospectively validated to identify a group of SLE patients with high risk of infection throughout the disease course to apply prophylactic measures regardless of immunosuppressive therapy and a closer follow-up.

DG-M participated in the conceptualization, funding acquisition, design, investigation process, project administration, supervision, data curation, formal analysis, validation, and visualization of the work. JT-R participated in the investigation process, data curation, formal analysis, validation and visualization of the work, and writing the original draft. NM-D participated in the formal analysis. AZ-D and MT-R provided software and resources for confocal mycroscopy. SM-P, RV-R, MA-L, RR-d-l-G, GJ-V, JM-C, and AB-V participated in data curation. JA-V participated in conceptualization, supervision, review, and editing the manuscript. AP-d-L participated in the supervision of the project.