Impaired respiratory function reduces haemoglobin oxygen affinity in COVID‐19

Using homology modelling and molecular docking algorithms, Liu and Li predicted the existence of interactions between SARS‐Cov‐2 and haemoglobin (Hb) and suggested that these interactions may change haemoglobin‐oxygen (Hb‐O₂) affinity. ¹ Hb‐O₂ affinity has been extensively studied in intensive care unit (ICU) patients with COVID‐19 (usually intubated and mechanically ventilated) showing normal to increased affinity. ² , ³ , ⁴ Higher Hb‐O₂ affinity facilitates Hb oxygenation in the lungs while reducing tissue O₂ unloading; a reduced Hb‐O₂ affinity has the opposite effect. Only one study has investigated patients with COVID‐19 outside the ICU. ⁵

In a retrospective, cross‐sectional study, we analysed Hb‐O₂ affinity in COVID‐19 patients hospitalised in general wards of our institution. The local ethics committee approved the study. Details on study protocol, methods and additional results are reported in Data S1. All patients had a laboratory‐confirmed diagnosis of COVID‐19 and an arterial blood gas analysis performed on admission, with a Hb‐O₂ saturation (SO₂) of less than 97%. At the time of analysis, no patient was intubated, mechanically ventilated, or had received a COVID‐19 vaccination. We determined oxygen tension at half‐saturation of Hb (P₅₀), an index of Hb‐O₂ affinity, using two equations for the determination of in vivo and standard P₅₀ (Figure S1). ⁶ In vivo P₅₀ reflects actual arterial blood P₅₀, as determined by Hb structure, red‐cell 2,3‐diphosphoglycerate (DPG), pH and arterial pressure of CO₂ (PaCO₂). Standard P₅₀ is the P₅₀ value measured at pH 7.4 and PaCO₂ 40 mm Hg; it is an index of Hb‐O₂ affinity in which the influences of pH and PaCO₂ have been removed. ⁷

General characteristics of study patients are shown in Table S1. Systemic inflammation was common (elevated C‐reactive protein and serum ferritin), but only 12% of cases had procalcitonin concentrations greater than 2.00 ng/ml, suggestive of superimposed bacterial infection. Standard P₅₀ was higher than in vivo P₅₀ (p < 0.001), although median values of both parameters fell within the normal range. Table S2 shows correlations of in vivo and standard P₅₀ with clinical and laboratory parameters. Multivariable analysis demonstrated that a compromised respiratory function, expressed by reduced SO₂ and a low ratio of arterial O₂ pressure to fraction of inspired oxygen (PaO₂/FiO₂) and high lactate and lactate dehydrogenase, were associated with higher P₅₀ (Table 1). pH had a negative correlation with in vivo P₅₀ and a positive correlation with standard P₅₀. Ninety‐five patients (33%) died within 28 days from hospital admission. Patients who died had more elevated in vivo P₅₀ than those who survived, while no difference was observed for standard P₅₀ (Table S3). However, on multivariable logistic regression, the effect of in vivo P₅₀ disappeared and only advanced age, reduced PaO₂/FiO₂ ratio and increased lactate dehydrogenase were associated with mortality (Table S4).

TABLE 1.

Multiple regression analysis of factors related to P₅₀

Variables	In vivo P50		Standard P50
	PCor (95% CI)	p	PCor (95% CI)	p
SO2	−0.581 (−0.659 to −0.490)	<0.001	−0.563 (−0.645 to −0.469)	<0.001
pH	−0.675 (−0.739 to −0.599)	<0.001	0.399 (0.285 to 0.502)	<0.001
Lactate	0.287 (0.166 to 0.400)	<0.001	0.289 (0.167 to 0.402)	<0.001
Multiple regression r	0.769 (0.711 to 0.816)	<0.001	0.679 (0.603 to 0.742)	<0.001

The below analysis was performed excluding SO2 from independent variables; pH and PaCO2 were additionally excluded in standard P50 analysis
PaO2/FiO2	−0.314 (−0.428 to −0.190)	<0.001	−0.146 (−0.278 to −0.009)	0.038
pH	−0.603 (−0.681 to −0.512)	<0.001	—	—
Lactate	0.324 (0.200 to 0.438)	<0.001	0.153 (0.016 to 0.284)	0.030
LDH	—	—	0.361 (0.235 to 0.475)	<0.001
OxyHb	−0.209 (−0.332 to −0.079)	0.002	−0.148 (−0.280 to −0.011)	0.036
Multiple regression r	0.501 (0.395 to 0.594)	<0.001	0.508 (0.399 to 0.603)	<0.001

Abbreviations: LDH, lactate dehydrogenase; OxyHb, arterial oxyhaemoglobin; PaO₂/FiO₂, PaO₂ to fraction of inspired oxygen; PCor, partial correlation coefficient.

Figure 1 shows SO₂ values as a function of measured PaO₂, or PaO₂ adjusted for pH and PaO₂; distribution of SO₂ was similar to the oxygen dissociation curve (ODC) obtained from Hill's equation as adapted by Doyle (Figure 1), in which normal P₅₀, measured at 37°C, pH 7.40 and PaCO₂ 40 mm Hg, is assumed to be 26.6 mm Hg, ⁸ showing an overall agreement between observed and expected SO₂. This curve almost perfectly overlapped the standard human Hb ODC reported by Severinghaus ⁹ (data not shown). Lower levels of SO₂ (<85%) were often shifted rightward with respect to the reference SO₂ curve expected for a given PaO₂, indicating that these patients had reduced Hb‐O₂ affinity. Patients with SO₂ of less than 85% had a reduced PaO₂/FiO₂ ratio, high lactate (>1.7 mmol/L) and increased P₅₀ (Table S5). In these patients, standard P₅₀ was significantly higher than 28.0 mm Hg, the upper limit of normal (p < 0.001), and the elevation was not justified by pH or PaCO₂ changes. These results show that a severely compromised respiratory function is associated with reduced Hb‐O₂ affinity.

FIGURE 1.

Distribution of observed and predicted SO₂ values in COVID‐19 patients. (A) SO₂ is reported as a function of observed PaO₂. and (B) as a function of PaO₂ corrected for arterial blood pH and PaCO₂. Red marks represent the expected SO₂ value for any given PaO₂ according to Doyle equation; blue dots correspond to experimental data.

Our data suggest a preeminent role of patients’ respiratory status and acid‐base balance in the regulation of both in vivo and standard P₅₀. pH was negatively related to in vivo P₅₀, since acidosis reduces Hb‐O₂ affinity; ⁹ in contrast, pH had a positive correlation with standard P₅₀, as described in critically ill patients. ⁶ PaCO₂ behaved in the opposite way. The correlations with standard P₅₀ are likely to represent the indirect effects of pH (and PaCO₂), possibly due to the upregulation of 2,3‐DPG production by alkalemia. ¹⁰ Our results are partially at variance with previous reports showing increased or normal Hb‐O₂ affinity in COVID‐19. ² , ³ , ⁴ , ⁵ , ⁶ , ¹¹ Most of these studies were conducted in ICUs with determination of standard P₅₀; in these settings a permissive hypercapnia is often allowed when low tidal ventilation is adopted to reduce ventilator‐associated lung injury ¹² and may explain the reported reduction in standard P₅₀. Otherwise, reduction of standard P₅₀ could result from 2,3‐DPG downregulation in the presence of respiratory or metabolic acidosis. Compared with our data, ICU studies reported higher PaCO₂ and lower pH values, ³ , ⁴ which can cause standard P₅₀ reduction, but also demonstrated that elevated P₅₀ is usually associated with more severe disease and increased mortality. ⁴ , ⁵

Some patients in our study had respiratory distress and were subsequently transferred to the ICU, but at the time of analysis they were spontaneously breathing, and most had normal blood pH with low PaCO₂. Median Hb was 129 g/L, confirming that COVID‐19 patients in internal medicine units usually have only mild anaemia. ¹³ In addition, Hb was similar in patients with SO₂ of 85% or more and those with a SO₂ below 85%, indicating that anaemia was not involved in the upregulation of P₅₀ associated with worsening respiratory function. Our data in patients with SO₂ below 85%, showing increased P₅₀ and lactate concentration, confirm previous observations in acute myocardial infarction, suggesting that P₅₀ elevation represents a compensatory change aimed at preventing tissue hypoxia when oxygen delivery is compromised, ¹⁴ whereas high plasma lactate is a marker of the O₂ debt associated with decreased oxygen transport and consumption. ¹⁵ Although the reported results are preliminary and require confirmation, we document that a severely compromised respiratory function is accompanied by significant elevations of P₅₀ in COVID‐19 that is likely to be an expression of the organism's adaptation to hypoxaemia. This agrees with previous observations showing that a left‐shifted ODC is associated with lower mortality in COVID‐19 patients at hospital admission. ⁵

AUTHOR CONTRIBUTIONS

Gaetano Bergamaschi designed the study, analysed data, performed statistical analysis and drafted the manuscript. Chiara Barteselli, Virginia Del Rio, Federica Borrelli de Andreis, Ivan Pellegrino, Caterina Mengoli, Emanuela Miceli, Marta Colaneri and Valentina Zuccaro set up the database, collected the data and critically reviewed the manuscript. Michele Di Stefano, Raffaele Bruno and Antonio Di Sabatino designed the study, analysed the data and critically reviewed the manuscript.

CONFLICTS OF INTEREST

The authors declare no competing conflicts of interest.

Supporting information

Figure S1.

Table S1.

Table S2.

Table S3.

Table S4.

Table S5.

Table S6.

Table S7.

Table S8.

Table S9.

Data S1.

APPENDIX A.

Internal Medicine Covid‐19 Collaborators

Giulia Maria Abruzzese, Giovanna Achilli, Eleonora Alimenti, Giacomo Alunno, Valentina Antoci, Marco Aprile, Alice Argelli, Nicola Aronico, Alessia Ballesio, Giulia Bellini, Alessandra Berbenni, Giampiera Bertolino, Paola Ilaria Bianchi, Silvia Biserni, Alessandra Bonfichi, Tommaso Bosoni, Michele Brattoli, Francesca Calabretta, Marco Callisti, Ginevra Cambiè, Roberta Canta, Pietro Canu, Chiara Cebrelli, Aurora Cecco, Carolina Cicalini, Marina Clemente, Luca Clerici, Silvia Codega, Federico Conca, Luigi Coppola, Filippo Costanzo, Maria Elisa Cremonte, Roberto Cresci, Mariangela Delliponti, Virginia Del Rio, Claudia Delogu, Giuseppe Derosa, Davide Dionisi, Francesco Di Terlizzi, Michele Dota, Francesco Falaschi, Erica Fazzino, Letizia Ferrami, Serena Ferrara, Maria Giovanna Ferrari, Nicola Ferruccio, Giulia Freddi, Carmine Frenna, Chiara Frigerio, Federica Fumoso, Alessandra Fusco, Alice Galeazzo, Vera Gaspari, Antonella Gentile, Alessio Giangreco, Giulia Gori, Giacomo Grandi, Virginia Gregorio, Paolo Grimaldi, Andrea Italia, Francesco Lapia, Mario Andrea Latorre, Marco Vincenzo Lenti, Federica Lepore, Antonio Lobello, Elisabetta Lovati, Pietro Carlo Lucotti, Francesca Lusetti, Stiliano Maimaris, Jacopo Mambella, Alessandra Martignoni, Federica Melazzini, Chiara Mercanti, Stefania Merli, Giusy Cinzia Moltisanti, Maria Elena Monti, Luisa Morbegno, Francesco Mordà, Amedeo Mugellini, Chiara Muggia, Roberta Muscia, Alba Nardone, Lucia Padovini, Ilaria Palumbo, Ignazio Marzio Parisi, Alessandro Pecci, Giulia Paola Peroo, Clarissa Petrucci, Carla Pieresca, Gianluca Pino, Lavinia Pitotti, Silvia Poma, Paola Stefania Preti, Andrea Quadrelli, Alessandro Rascaroli, Giovanni Rigano, Carlo Maria Rossi, Giorgio Rotola, Damiano Ruggeri, Maria Concetta Russo, Umberto Sabatini, Simone Saglio, Giovanni Santacroce, Gabriele Savioli, Jessica Savioli, Simone Salvator Scalia, Davide Scalvini, Yiftach Shoval, Carlo Soffiantini, Simone Soriano, Dario Spadaro, Andrea Staniscia, Debora Stefani, Federico Vai, Matteo Varallo, Marta Vernero

DATA AVAILABILITY STATEMENT

The datasets used in this study are available from the corresponding author upon reasonable request. The data are not publicly available due to privacy or ethical restrictions.